University of Virginia Library

The Hengwrt and Ellesmere Manuscripts of the Canterbury Tales: Different Scribes
by
Roy Vance Ramsey

A vexed question in Chaucer studies is the exact relationship of the two best manuscripts of the Canterbury Tales, Hengwrt (Peniarth 392 D) and Ellesmere. Although the two manuscripts have been shown by John M. Manly and Edith Rickert to be only occasionally affiliated in their text,[1] paleographers in the past have identified the handwriting of the two as that of the same scribe. An adaptation of the techniques of compositor identification, however, has provided the means for subjecting the manuscripts to statistical tests. The results not only seem to settle the question of the number of scribes copying the two manuscripts but also seem to offer at least one reason why the two differ materially in the quality of their texts: Hengwrt and Ellesmere were copied by different scribes with different habits of registering certain accidentals, different habits of proofing finished copy, and different habits of registering substantive readings.

The present study was begun more than six years ago in the midst of working on the Variorum Chaucer project. The appearance of the Hengwrt manuscript (hereafter called 'Hg') reproduction, transcription, and full registration of the variants in the Ellesmere manuscript (hereafter called 'El') provides the means for a general testing of the present study's rejection of the claim that the two manuscripts were copied by the same scribe.[2] In their exhaustive study of the manuscripts of the Canterbury Tales, Manly and Rickert say of the scribal hand in Hg, "A large, clear book hand, probably the same throughout. . . . Paleographers believe the scribe to have written El also; the same variations in appearance occur in both" (I, 268). This belief that the same scribe copied Hg and El throughout is reiterated by A. I.

Doyle and M. B. Parkes in their "Paleographical Introduction" to the First Fascicle (pp. xix-xlix).

Taken by itself, this reinforcement by two scholars as expert in fifteenth-century paleography as Doyle and Parkes might seem to put the matter to rest. In another part of their study, however, Manly and Rickert say of the paleographical evidence: "Several scribes trained by the same master may seem to be a single scribe. The writing in El is so much like that in Hg and the writing in each is subject to such variations that some of the best experts cannot decide whether both are the work of the same scribe, or both the work of the same two, or each the work of a different writer. Even if two scribes working on the same MS normally write very differently, one may try to imitate the other when he continues to work" (I, 23). Doyle and Parkes also indicate the possibility of such differences in a scribe going undetected in a footnote which states in part, "R. Vance Ramsey has shown us some statistics of the variations, which need further discussion."[3] The present study begins this "further discussion" by providing much fuller sets of statistics than those seen by Doyle and Parkes and by arguing that, as undeniably expert as they are in English medieval paleography, the statistics and some other differences to be discussed later refute their identification of the Hg and El scribes as one. The importance of this question for the textual criticism and editing of the Canterbury Tales will be discussed in the last part of the study.

Basic to any attempt to study two manuscripts whose handwriting seems identical are the points made simply by Charlton Hinman concerning compositors: ". . . if two or more compositors set type for a given book, two or more sets of spelling habits are likely to be discoverable in it. . . . The later text always reproduces some copy spellings, but many changes are made, and these changes generally leave no doubt whatever as to the spelling preferences of the compositors who alone can be responsible for them."[4] Similarly, Manly and Rickert discovered in the manuscripts of the Canterbury Tales "consistency in adhering to a spelling system" and noted: "Even the less regular scribes show some points of preference; and others, especially some of the best and earliest, are characterized by a fairly complete standardization. [¶] The reasons for this lie in the obvious fact that writers were taught in schools or trained in shops, and that they inevitably formed habits in spelling as in writing" (I, 558). While taking the Manly-Rickert evidence and analyses into account, the present study shares certain premises with such compositor-studies as that of Hinman: both recognize the likelihood of variability in spelling-habits by different compositors and scribes; both recognize

that the spellings of the exemplars will at times be brought forward unchanged and that the percentage of such spellings left unchanged will vary from one man to another and even from one locus to another; thus, both kinds of study depend heavily on the law of averages. The present study, however, depends far more heavily on this law and on sheer numbers. This need for large numbers is explained after presenting the first chart.

The chart below follows six words through the 500 pages (250 leaves) of Hg and the corresponding spellings in El. The chart differs from those in compositor studies in not detailing the occurrences page by page or quire by quire. This is for the simple reason that analysis of the various ways copy might be apportioned to scribes (leaf by leaf, gathering by gathering, or tale by tale) has shown that the systematic variations in spelling persist from one leaf to another, from one gathering to another, and from one tale to another (although there are occasional anomalies in the counts from the latter which are discussed later). A demonstration of the point about the differences in spelling persisting from one tale to another is provided for the first word on the chart, thou. After that word the others are simply given as totals for reasons of space; however, a comparable regularity obtains for all.

Characteristic Words in Hg/El (Order of Variants)[5]

                                                       

1.	thow/thou	thow/thow	thou/thou	thou/thow
Pro
KtT	30	24	3	1
MiT	17	3	6
ReT	4	7	3
CkT
WBT	38	8
FrT	24
SuT	16		1
MkT	22	2	2
NPT	8	1
McT	14	3
MLT	19	5	5
SqT	4	1
MeT	17	2	1
FkT	6
SNT	36	10
CIT	15	2
PhT	10		1
PdT	22	1	1
ShT	1
PrT	2
Th	8
Mel	83	4	9
PsT	12	1
---
	408 (79%)	74 (14%)	32 (6%)	1 (.2%)
TOTAL: 515

                             

2.	ellis/elles	ellis/ellis
	102 (88.7%)	13 (11.3%)
	TOTAL: 115
3.	thogh/though	thogh/thogh	though/though	though/thogh	OTHER
	121 (57.6%)	58 (27.6%)	20 (9.5%)	3 (1.4%)	8 (3.9%)
	TOTAL: 210
4.	doun/doun	down/doun
	93 (68%)	44 (32%)
	TOTAL: 137
5.	town/toun	toun/toun	town/town
	27 (54%)	15 (30%)	8 (16%)
	TOTAL: 50
6.	at the/atte	atte/atte	at the/at the	atte/at the	OTHER
	29 (40.8%)	26 (36.6%)	7 (9.8%)	7 (9.8%)	2 (2.8%)
	TOTAL: 71

The only word on this chart with a combination of frequency of occurrence and regularity of spelling-pattern comparable to the Shakespeare Folio test-spellings of do, go, and here (Hinman, I, 182-214) is the first word, thou. This word occurs slightly more often than once per page, and four-fifths of the time Hg spells it thow at the place where El spells it thou. Putting together all of the occurrences shows an even more striking opposition: in 515 opportunities to spell the word, the Hg scribe spells it thow 482 times (or 93% of the spellings) and the El scribe spells it thou 440 times (85.4%).

The second word (else) offers even more striking testimony (allowing for its occurring only a fifth as often as thou): the Hg scribe never spells the word elles, whereas the El scribe spells it ellis only 11.3% of the time. The first point explains why the other two possible combinations (elles/ellis and elles/elles) never occur. While all possible combinations of thogh and though do occur (plus 8 other such combinations as theigh/though), once again Hg shows a marked preference for one spelling (thogh 179 times, or 85% of the spellings), while El prefers the other (though 141 times, or 67%). Also, a few of Hg's though spellings are in the rhyme position at the end of the line, and Hg almost always spells the rhyming portion of two words in such a place the same.

The words down and town have identical vowel sounds, of course; yet in 137 occurrences El never spells the former word down (compare the 85% preference for thou), while he spells the latter town 8 times out of 50 occurrences. Hg, on the other hand, uses ou 2 to 1 in doun, but ow 3 to 1 in town (compare the 93% preference for thow). The word atte, standing for at the, is familiar to readers of Chaucer because of past editors' dependence on El, which uses it 55 out of 71 times (77.5% of the time). Hg, conversely, spells out at the 37 times and registers atte 34 times (El has at once where Hg has at the and once where Hg has atte).

Although not occurring with the frequency of do, go, and here, then, the words on the chart above do seem to show marked and opposed tendencies on

the parts of the copyists of Hg and El; nevertheless, certain evidence available to the student of the First Folio is not available for the study of Hg and El. No exemplar of either manuscript exists (unless Hg did serve as El's exemplar for the Friar's Tale, as Manly and Rickert believe possible).[6] Also, there is no 'evidence of the case' which serves compositor-studies so well. Most important, while compositor-studies generally deal with different compositors working from the same copy, this study does not. (Manly and Rickert's analyses show that the use of the same exemplar for two extant manuscripts is the exception rather than the rule.)

These differences explain the need mentioned before to study large numbers of occurrences, larger than the chart above of 'characteristic words' can provide: some of the differences in spelling registered must stem from the exemplars (such as the theigh/though contrast already mentioned) but, with the possible exception of the Friar's Tale, no exemplars have survived. Some of these differences, then, might stem from the same scribe faithfully registering different spellings in each exemplar; indeed, the marked difference in the Knight's Tale spellings of thou suggests that El is copying there from an exemplar with the word spelled thow a larger-than-usual percentage of the time (25 out of 58 times, El spells thow, compared with 50 out of 457 times elsewhere). The regularity of the other instances of thou and thow, on the other hand, makes it unlikely that this change stems from a variation in habit.

The scribes of Hg and El must have trained and worked in the same shop, a point to be returned to in connection with their identical-appearing hands. The solution to this problem of a need for large numbers to counter the lack of other resources available for compositor-studies and the identical look of the hands seems to lie in studying the scribes' graphic (whether they cross a final h or flourish a final d) and graphemic habits (their ways of representing sounds and combinations of sounds such as -er- and that).

The statistics of graphic and graphemic variations between Hg and El were gathered in two ways. The first way, represented on the chart which follows, was to count all the variations noted in all the lines of Hg and the corresponding lines of El in the course of working on the transcription of Hg and the corresponding lines of El for the First Fascicle. Because the number of variants watched for was so large (forty-plus) and because the counting was done in the midst of other work, the accuracy of the statistics is somewhat less than it might have been (unlike the 'characteristic words' which were recounted twice for this study). Spot-checks have shown the percentage of errors to range from o to 20%, with perhaps 10% as an average. That this percentage of error does not negate the import of the differences registered is shown by the chart below, where the percentages of the ten variants registered

TALE GRAPHIC AND GRAPHEMIC CONTRASTS IN THE HENGWRT AND ELLESMERE MANUSCRIPTS (Dominant Contrast/Reversal)

                                                         

(Hg (order)	1. h/h	2. d/d	3. -ogh/-ough	4. -ow-/-ou-	5. &/and	6. -on/-oun	7. that/that	8. single vowel/double	9. -ay-/-eye-	10. O/-e
	---	---	---	---	---	---	---	---	---	---
Pro	64/0	15/5	5/0	8/1	1/13	11/1	20/10	27/133	1/0	18/19
KtT	239/0	10/0	39/0	23/1	3/16	17/18	89/10	81/123	17/0	136/26
MiT	41/0	2/2	9/2	12/2	0/2	5/2	36/1	80/27	4/3	30/9
ReT	15/1	15/1	5/0	4/4	1/2	0/0	29/2	40/13	6/3	30/7
CkT	6/0	0/0	1/0	2/0	1/0	0/0	2/0	5/4	0/0	8/0
WBT	103/1	11/0	6/1	8/1	1/6	8/0	22/28	78/17	4/2	56/14
FrT	18/1	10/0	2/0	1/3	2/1	2/0	2/13	26/2	0/1	17/6
SuT	64/0	16/0	6/0	14/0	0/2	3/0	4/20	44/14	7/1	26/8
MkT	144/0	36/0	10/1	3/4	5/5	9/1	36/13	81/13	0/1	32/14
NPT	43/0	37/0	3/1	4/0	10/0	5/0	39/1	113/2	4/1	10/25
McT	43/0	16/0	1/0	7/0	0/6	1/1	9/2	69/0	2/0	12/12
MLT	101/0	14/1	4/2	21/2	1/3	7/2	68/8	131/28	5/6	102/9
SqT	94/1	28/0	3/0	4/0	0/4	28/1	18/6	50/14	3/0	69/7
MeT	22/0	22/0	20/0	1/1	0/4	7/1	38/33	110/17	10/2	106/10
FkT	130/0	27/0	7/0	5/1	5/5	7/1	26/17	77/10	11/0	75/8
SNT	91/0	10/0	4/0	6/3	7/1	5/0	8/16	65/4	5/0	18/14
ClT	125/0	9/1	15/2	6/0	7/4	4/1	53/24	142/8	15/5	141/7
PhT	37/0	22/0	4/0	2/0	0/3	2/1	10/5	24/10	1/0	11/1
PdT	98/0	28/0	9/0	3/0	8/5	9/0	23/15	65/10	3/0	33/8
ShT	35/1	22/0	5/0	6/0	1/4	2/0	23/15	39/2	2/0	21/6
PrT	27/0	12/0	1/1	0/1	0/1	1/0	7/6	42/5	3/0	20/1
Th	22/1	15/0	2/0	5/0	5/1	3/0	2/1	24/7	1/0	16/1
Mel	269/8	59/10	18/2	92/8	429/19	46/9	371/32	263/29	5/6	90/37
PsT	145/0	36/0	4/0	15/0	287/6	31/3	341/3	266/23	3/0	52/92
	---	---	---	---	---	---	---	---	---	---
TOTAL:	1976/14	472/20	183/12	258/32	768/113	213/42	1276/281	1942/515	112/31	1129/341
% of dominant contrast:	99.3	95.93	93.85	88.97	87.17	83.53	81.95	79.04	78.32	76.8

fully range from 76.8 to 99.3%. The range of the four variants listed at the bottom of the chart is from 60.3 to 72%. Another way of counting was to take 100 lines from each tale in Hg (the third fifty from the beginning and from the end where possible) plus the General Prologue (Pro) and the Wife of Bath's Prologue (listed with WBT). This second way of counting, like the spot-checks of individual variants throughout a tale, showed that the statistics on the chart had less of an incidence of error than the occasion of their gathering might have suggested.

Of the fourteen contrasts on the chart above, some require explanations because print cannot reproduce them: the italicized h stands for a 'crossed h' at the end of a word; d stands for a 'flourished d'—a 'd' with a tilde attached to the top of the riser—at the end of a word; '&' is the still-used Tironian sign for et; -oun stands for 'ou' in a word followed by a nasal mark; that stands for the Old English letter 'thorn' followed by an elevated 't'; -er- stands for the brevigraph used in place of that syllable; and with stands for 'w' followed by an elevated 't'. Except for the O standing for absence where a final -e or final -n stands, the other variations seem self-explanatory.

The most obvious difference between this chart and the first one is that all occurrences are not listed. This difference rests on both the feasibility and the desirability of such a listing. Although it is simple enough to count when Hg has a single vowel in a word where El has two or vice versa, any attempt to list all occurrences would come very close to listing all the words having the relevant one of the five basic vowel letters (graphemes) in English. Final -e and final -n occurrences and absences would be similar cases. Beyond this difficulty there is a very real question about the need or even the desirability of listing all the words where the two manuscripts are spelling alike. The Hg scribe's total rejection of elles is the unique exception in this study (he must know that it is an option, his identical hand showing him to be working in the same shop with the El scribe). Given the very probable spelling-variation in his exemplars and the certain information which El provides that the variation between ellis and elles was in keeping with shop-practice, even this one invariable spelling is more striking than it might otherwise seem. Nevertheless, Hinman is right that some of the spellings (such as theigh) will be brought forward from the exemplar, so that some variability in spelling is guaranteed by several scribes' separately and in varying combinations making the intevening copies between Chaucer's holographs of various tales and the copying done for Hg and El.

The question, then, is not variability (which was inevitable) but tendency: if the same scribe did copy the many thousands of lines in each manuscript, then whatever the reasons for his variations in spelling, his tendency to use a particular grapheme for a particular sound or sound-combination should be apparent in any very large count. If El much more often has a final -e or a final -n on the same word in the same place where Hg does not have it, then whatever the reason for its presence in one manuscript and absence in the other, a marked difference in tendency is clear.

Even more important than the numbers and percentages recorded at the bottom of the chart is the overall consistency in the ratios from one tale to another of the variants and their reversals. Early in this study, one consideration which prompted the counting of variants in all the lines rather than sampling lines was the possibility that the work on either or both of the manuscripts had been done by more than one scribe. Such apportionment of the copying (the forerunner of printing-house practice) would presumably have shown up in smaller patterns within the larger ones. The striking reversal of the single vowel/double vowel pattern in the General Prologue (Pro) and the Knight's Tale (KtT) looked as though it might be such an indication. A look at the other variants in the two pieces, however, shows the same ratios as those prevalent elsewhere. Some of the smaller anomalous ratios on the chart may well stem from the variability of scribal habit, but such reversals as the single vowel/double vowel ones seem to involve enough instances and a complete enough change in the dominant patterns as to stem mainly from the exemplars being copied rather than from some variability in scribal habit.

A difference in exemplar having been admitted as the probable cause of the anomalous patterns (or at least of those where relatively large numbers are involved), the question may well be raised of why this difference in exemplars does not explain the variant ratios generally. There are three major reasons and a minor one why the patterns point to a difference in scribes rather than a difference in exemplars as the primary explanation of the ratios: the nature of accidentals, the regularity of the ratios tale by tale, the shifting textual affiliations of El, and a local but striking instance where both scribes are probably copying from the same exemplar.

The nature of such accidentals as spelling, the use of a brevigraph, the use of & in place of and, and so forth is that a writer gives much less thought to them than to such questions about substantives as whether the word in his exemplar is fore or sore (the initial f and long s being particularly easy to confuse). The chart seems to have more than enough data to show that different scribes with different habits were involved in copying the two manuscripts; on the other hand, such complete reversals of the dominant patterns as that involving the single and double vowels in the General Prologue and the Knight's Tale and that involving the spelling -on or -oun in the latter (18 of 42 reversals occur in this tale) seem to indicate that at times the spellings of the exemplars were registered more often than the scribe's own habitual ones. Similarly, the Corpus Christi 198 (Cp) and Lansdowne 851 (La) manuscripts of the Canterbury Tales, which are very close textually, show mixtures of exemplar patterns and scribal habits of spelling and of registering other accidentals. Such an alternation between a scribe's following his exemplar's spellings and his own spelling-habits should cause a fairly constant minority (stemming from the exemplar) of reversals of the dominant patterns, a smaller number of complete reversals, and should rule out more than one or two invariables in manuscripts of this length—just the pattern which the chart displays.

The second basic reason for rejecting the exemplars as the source of the variations between Hg and El in accidentals is the regularity of these variations. The chart demonstrates this regularity tale by tale, and, as noted, samplings were made throughout for smaller patterns—which never appeared. This study relies, as did Manly and Rickert's classification of manuscripts, upon "the regularity of the operations of chance in mass phenomena," upon "the doctrine of probability" (II, 23). The case offered by the 'characteristic words' chart was undermined by the relatively small numbers involved. The combination of number of contrasts, number of tales, and regularity of results in the second chart presents in itself a case of high probability.

The third basic reason that the differences on the chart cannot stem from differences in the exemplars of Hg and El is the latter's shifts of textual affiliation. This is a point more amply demonstrated by Manly and Rickert in the tale-by-tale analyses of variants in Volume II, but a brief account may be useful. Manly and Rickert discovered that even the early manuscripts of the Canterbury Tales tend to change textual affiliation from one tale or group of tales to another (Hg and El date with the earliest). The textual affiliations of El in the Man of Law's Tale and the Wife of Bath's Prologue and Tale can illustrate such shifts. For the Man of Law's Tale, the El scribe copied an exemplar much like the ones used by the scribes of the Fitzwilliam (Fi), Phillips 8136 (Ph2), British Museum Additional 35286 (Ad3), Christ Church CLII (Ch), Bodleian 686 (Bo2), and the a family of manuscripts. For the Wife of Bath's Prologue, on the other hand, the El scribe copied from an exemplar related to those used for Cambridge Gg.4.27 (Gg), Sion College (Si), and the Bodleian 414 (Bo 1) and Egerton 2864 (En 3) groups of manuscripts up to line 387. Then, at line 388 the El scribe's exemplar shifted (or he changed exemplars) to a textual relationship with the exemplars of the Royal College of Physicians manuscript (Py) and the a and augmented b* groups of manuscripts. In the Wife of Bath's Tale, the El exemplar remained textually affiliated with the a and augmented b* groups, but the Py scribe's exemplar shifted to another affiliation, while the Bo2 exemplar shifted to affiliation with El again.

Needless to say, shifts like these must have disconcerted Manly and Rickert. They concluded that such shifts even in early manuscripts of high textual quality such as El must have been caused by the earlier piecemeal issuance of tales and groups of tales during Chaucer's later life (with subsequent losses of leaves to explain the shifts in the early part of the General Prologue and such a shift as that at line 388 of the Wife of Bath's Prologue). This shifting and piecemeal issuance explain why, although Hg and El contain the two best extant texts of the Canterbury Tales virtually throughout, these two manuscripts are textually affiliated only six times: the Friar's Tale, the Summoner's Tale (through line 1991), the Monk's Tale, the Nun's Priest's Tale (but not the prologue—and Manly and Rickert vary about the tale itself), the Manciple's Tale, and the Clerk's Tale.

An instance where the Hg and El scribes are copying either the same exemplar or ones separated by no more than one copying provides the opportunity

to see the respective habits overbearing the accidentals of the exemplar. For this purpose the accidentals of the copies of the same exemplar will be compared with those made of exemplars which have no relationship short of the original scribal copy of Chaucer's holograph (O). Of the Reeve's Tale, Manly and Rickert state, "Hg and El probably have no common ancestor short of O" (II, 159). Of the Friar's Tale, on the other hand, they say, "For the first time Hg and El are together. . ." (II, 225). They think that the exemplar is so close to O, in fact, that they venture this comment at the end of their analysis of the Friar's Tale variants in Hg and El: "From the mistakes in the ancestor of Hg-El and related MSS, it is clear that Chaucer did not correct his text with much care" (II, 225). The fact that El changes its textual closeness to Hg at line 1991 puts them of two minds about how probable it is that the exemplar used by the Hg scribe for the first half of the Summoner's Tale is the same one used by the El scribe. The two tales seem to have been issued as a unit (naturally enough in view of their being the expressions of a quarrel between the two men): "Hg and El are apparently from an V Hg-El [common ancestor] in FrPT . . . and SuPT to 1991 . . ." (I, 150). Later, they state, "Through 1991, then, aside from the editorial efforts in El, Hg and El are not very different in quality; from 1991 the difference is enormous" (II, 239). While Manly and Rickert stop short of affirming that Hg and El used the same exemplar up to line 1991, they do state of the misreadings which Hg and El share, "It may be observed . . . that the six misreadings are suggestive of the use of the same exemplar" (II, 239).

At any rate the closeness of the text of Hg and El in the Friar's Tale and the Summoner's Tale to line 1991 and the distance of their texts in the Reeve's Tale and the Summoner's Tale, lines 1992-2294, should manifest themselves in the accidentals if such differences do stem from a single scribe's having copied very different exemplars for Hg and for El. The range of differences in substantives which allows decisions about nearness or distance of textual affiliation should be matched by a similar range of differences in accidentals if the scribe is simply registering those he finds in very different exemplars. The comparison follows:

                         

VARIANT	SuT, lines 1665-1991	SuT, lines 1992-2294	SuT, TOTAL	FrT	ReT
---	---	---	---	---	---
thow/thou	7/0	9/0	16/0	23/0	2/0
1. h/h	31/0	33/0	64/0	18/1	15/1
2. d/d	4/0	12/0	16/0	10/0	15/1
3. -ogh/-ough	5/0	1/0	6/0	2/0	5/0
4. -ow-/-ou-	5/0	9/0	14/0	1/3	1/4
5. &/and	0/1	0/1	0/2	2/1	1/2
6. -on/-oun	2/0	1/0	3/0	2/0	0/0
7. that/that	1/14	3/6	4/20	2/13	29/2
8. single vowel/double	24/7	20/7	44/14	26/2	40/13
9. -ay-/-ey-	6/1	1/0	7/0	0/1	6/3
10. O/-e	16/2	14/11	30/13	17/6	28/9

As with the other charts, there are anomalies in the distribution, notably in the distribution of that/that, which may stem from the exemplars; but, given the matter of closeness or distance of textual affiliation, they probably stem from variation in scribal habit. The important point about this chart is that the preponderance of the ratios is the same whether the Hg and El scribes are copying from the same or very similar exemplars in the Friar's Tale and the Summoner's Tale to 1991 or whether they are copying from very different ones in the Reeve's Tale and the Summoner's Tale after line 1991.

Returning for a moment to the earlier chart, one pattern there is not immediately evident: Hg has the sign & where El writes out the word and 768 times to 113 of the reverse. The most striking anomalies of all occur with this variant: while the overall ratio is 768/113, the ratio in Melibeus is 429/19 and in the Parson's Tale, 287/6. When these two counts are subtracted, the ratio reverses to 52/88. This is a case where a major break in the pattern actually reinforces the point of different scribes: Melibeus and the Parson's Tale are the two prose works in the Canterbury Tales; therefore, when the Hg scribe is copying poetry, he is slightly less likely than the El scribe to use the sign for and (52/88), but when he is copying prose, he is twenty-eight times more likely to use the sign (716/25).

The evidence seems clear, then: the major source of the differences in accidentals between Hg and El is not in the exemplars of each but in the scribe of each. If we accept that these differences stem from the scribe, the question still may be raised of a possible change of habits. Once again, the figures can only point to probabilities; but, once again, the probabilities are quite high against a change in habits as consistent and widespread as this would have been. In the first place, no more than a decade can have intervened between the copying of Hg and El (the loss of part of the Summoner's Tale between the two copyings being a portion of the evidence for the sequence); a much shorter interval is likely. Second, Hg and El are in a professional book hand which was somewhat old-fashioned in the early fifteenth century. To put down the difference charted to changes in habit would require us to imagine that a perhaps middle-aged scribe who had earned his livelihood by daily copying for a number of years greatly changed fourteen-plus of his copying habits during the few years after he copied Hg. On the contrary, as engrained as middle-aged habits tend to be generally, the copying-habits of a middle-aged scribe should be especially set. The source of the widespread, systematic differences between Hg and El in their accidentals is not to be found in their different exemplars nor in a single scribe's greatly changing his habits in mid-life: the source is the difference in scribes.

There are still two questions raised by this rejection of the paleographical evidence in favor of different scribes' having copied the two manuscripts: (1) how the writing of different scribes can be so alike as to escape the detection of expert paleographers; (2) given this possibility that different scribes can escape detection, whether more than one might not have been involved in the production of either or both Hg and El. The answer to how the handwriting

of different scribes can be in effect identical relates to the aims of the best professional scribes: "Several scribes trained by the same master may seem to be a single scribe," as Manly and Rickert state (I, 23), because they were trained to seem so. The point of training for such identity of hand was so that when the work of copying a single manuscript was apportioned among two or more of them, no differences would be evident after their portions were put together. This is why Manly and Rickert discovered that "some of the best experts" could not settle the question of the number of scribes copying Hg and El.

What, then, of the second question: if more than one scribe evaded detection by expert paleographers between Hg and El, might not more than one have copied one or both manuscripts and still have evaded the present count of variants? This seems possible but not probable. That same regularity of contrasts tale after tale which points to separate copyists of Hg and El also points to a single set of habits operating upon each manuscript. The few anomalies involving relatively large numbers may give us pause, but their kind and occurrence seem too random to point to smaller patterns. Particular attention was paid in the various counts within tales to those with such anomalies, but no smaller pattern emerged. If two scribes did copy either Hg or El, then one was the imitator of the other (perhaps the apprentice rather than the partner) not only in handwriting but in his habits of registering accidentals as well—and thus quite possibly not detectable at all. This question, however, is of much less moment than that of different scribes' having copied Hg and El because of one last difference in scribal habits—the treatment of substantives.

The discovery that different scribes with significantly different habits copied Hg and El is important in much the same way that similar studies of compositors' habits are: the differences in scribal habits extended beyond the accidentals to substantives and resulted in significant differences in the textual quality of each manuscript, apart from the textual quality of their respective exemplars. Let it be clear that this difference in the textual quality of Hg and El was well understood by Manly and Rickert and that they tried to convey the import of the difference: "Although El has long been regarded by many scholars as the single MS of most authority, its total of unique variants, many of which are demonstrable errors, is approximately twice that of Hg, as is also its total of slips shared with other MSS by acco [accidental coincidence of variants in unaffiliated manuscripts; the Manly-Rickert abbreviation is used throughout this study]. And again, while it has a few lines not in any other MS, and shows some editorial changes that could have been made by Chaucer, it has many others that are questionable and some distinctly for the worse, even involving misunderstanding of the context. Since it is very clear that an intelligent person, who was certainly not Chaucer, worked over the text while El was copied, the unsupported readings of this MS must be scrutinized with the greatest care" (I, 150). Of Hg, on the other hand, Manly and Rickert state: "Because of its great freedom from accidental

errors and its entire freedom from editorial variants, Hg is a MS of the highest importance" (I, 276). These are important statements with important implications, not only for the present study but also for the general understanding and editing of the text of the Canterbury Tales. Editors still far too freely use El variants, given the Manly-Rickert analyses and their "Corpus of Variants." While such is a study in itself, a measure of the neglect of Manly-Rickert's analyses and data by editors is the use of El as the base-text of the Canterbury Tales in one of the latest editions of Chaucer's works.[7]

In light of work done on the Variorum Chaucer and given Manly and Rickert's analyses of the manuscript variants tale by tale in Volume II, even the above-quoted characterizations of the two manuscripts seemed restrained, so a count was made. Going through these Volume II analyses revealed that Manly and Rickert's estimate of the "total of unique variants" in El as "twice that of Hg" was apparently based on impression, on an early count, or on the ratio of the total number of variants, by acco as well as unique. Leaving out the Canon's Yeoman's Tale (which is not in Hg) and the Parson's Tale (only partially present in Hg) and counting the rest revealed the following contrasts: El's total variants (unique and by acco, leaving out those by affiliation) is 1440 to Hg's total of 809; El's unique variants number 249 to Hg's total of 47. While El's total number of variants which comparison of affiliated manuscripts shows cannot stem from its exemplar, that is, almost matches the estimate of double the number of Hg, the actual ratio of unique variants is more than five times the number in Hg. As this difference in ratio suggests, the number of unique variants in a manuscript has different implications than does the number of variants by acco.

Many of the variants in a manuscript stem from its exemplar: the more changes which were made by the scribe of the exemplar and by the scribes of the ancestor-manuscripts, the more variants which will be copied unintentionally by the present scribe ('variant' has been chosen over 'error' as the more neutral word). In addition, each copying of an exemplar of any length inevitably introduces new variants because of mistakes of the copyist. The Canterbury Tales perhaps had no more careful copyist than the scribe of Hg, yet the manuscript contains a total of 876 variants, unique and by acco (including the Parson's Tale). Even attributing half of these to his exemplar (an unlikely percentage) leaves a remainder of 438, almost an error a page. Because the exemplars of El were at times demonstrably farther from the original scribe copy of Chaucer's holograph (O) than the exemplars of Hg (e.g., the Summoner's Tale after line 1991), a larger proportion of the El variants by acco must stem from problems with its exemplars. This still leaves the Hg scribe the more careful in proofing, but not in copying by the 14-to-8 ratio of the total number of variants by acco. On the other hand, the 5-to-1 ratio of unique variants is unassailable and implies very different scribal habits.

In the first place, a unique variant is almost always attributable to the scribe of the manuscript (in the case of the large number and many affiliations of the manuscripts of the Canterbury Tales). There may be cases where the unique variants stem from earlier scribal mistakes in manuscripts which served as exemplars only for a single copy or where other scribes recognized errors and corrected, but for early manuscripts of good textual affiliation, the great majority of unique variants are the scribe's own. Some or all of these unique variants may be unintentional, depending upon the scribe's attitude toward the text: in this regard, Manly and Rickert attribute only 5 of Hg's 47 unique variants to editing, whereas they attribute 89 of El's 249 to this cause. While such a comparison does not quite demonstrate Hg's "entire freedom from editorial variants" (though the 5 so termed are at least arguable), the El scribe's greater willingness to edit his text is clear from these figures alone. As there are four tales in the El analyses where Manly and Rickert make no statement about editing, the count could probably be even higher.

By their nature, unique variants should also shed some light on the proofing of a manuscript. Variants by acco are mostly mistakes common to scribes as a profession (rather than stemming from the common ancestor of a manuscript tradition). That they were also mistakes which are easy to over-look in proofing is shown by their continuance and number. Unique variants, on the other hand, are mistakes peculiar to one copying of the text; a careful proofing should catch a number of them. In line with his greater care in copying is the Hg scribe's greater care in proofing: despite the greater number of variants in El, Manly and Rickert list twenty times when the Hg scribe corrected a variant (twelve of them unique variants) to only four times when the El scribe corrected (all four unique variants). The Hg scribe's greater care in proofing and correcting is shown clearly by a final set of numbers: taking the 12 corrected unique variants from the total of 47 leaves 35 unique variants in Hg; taking the 4 corrected from El's 249 leaves 245 (a ratio of uncorrected variants of 7-to-1).

A look at the two manuscripts themselves shows why this difference in proofing cannot stem from the conditions under which each was produced. If either copyist was rushed in his work, it must have been the Hg scribe: the inexpensive vellum used, the lack of supervision, and especially the blank spaces for words and even lines all suggest that he may have been under some time-constraints. El, on the other hand, is the most elaborate manuscript of the Canterbury Tales: copied on expensive vellum, provided with the famous illuminations, this manuscript was prepared with care and at great expense. So it is not the press of time but a difference in attitude toward the received text which is the operative element in the substantive differences between the texts of Hg and El.

A look at the Friar's Tale where, as we have seen, the Hg and El scribes are copying from the same or a very similar exemplar will show specifically how the difference in attitude toward the text of the exemplar is revealed by the kinds of unique variants in each manuscript. Hg and El each has in its

text of the tale three unique variants, meaning that no other manuscript has them of the eighty-four extant. But the reasons for the variants in each manuscript are quite different. The first two Hg variants occur in line 1282, where the line in the Manly-Rickert reconstruction reads, "I praye that noon of yow be yuel apayd." It may be that the exemplar had a brevigraph for 'yuel' (evil) which the Hg scribe mistook, or it may be that mental fatigue caused him to write 'y payd' instead of 'yuel apayd'. Whatever the explanation, it must fall into Manly and Rickert's category of "unintentional errors" rather than "editorial variants." The other unique reading in Hg is in line 1571, and this is one which the scribe caught and corrected (an example of his greater care in proofing).

Manly and Rickert say of El's three unique variants, on the contrary, that they are "all probably editorial" rather than unintentional (II, 226). A look at the three variants in their contexts will show both why Manly and Rickert call them "editorial" and why they seem to point to the scribe himself rather than to a separate editor or supervisor. The first unique reading in El occurs at line 1479 and is given in brackets:

The second unique reading is in line 1496: The third unique reading occurs in line 1502: Remembering that these variants were certainly not in the El scribe's exemplar, the first point to be made is that two of the changes cannot be the result of misreading. The closest of the three to such a possibility is the second, which may be an "eye-skip" from the line before.

As for the possibility that an editor other than the scribe was responsible for the changes, the very misunderstanding of, in all three cases, a very limited context speaks against this (as do other variants by acco, to be discussed below). The first passage is the expression of that favorite Chaucerian villain's ideal of keeping mum and looking out for Number One (all the time blabbing to someone). Very little thought would have been needed by an editor to show why, whatever the momentary confusion of the switch from the singular I of the preceding line to the plural oure, greater confusion results from changing oure to hir (their). Similarly, someone (Manly and Rickert's "intelligent person") who was familiar with the idea of a fiend's being allowed to afflict the body but not the soul might have changed body to soule in line 1496; however, if he had been looking at only so large a

part of the context as three lines, he would have seen that the soule in the preceding line had to be changed to body for the lines to make any sense. On the other hand, if we are to see in the variant not intentional change but an unintentional "eye-skip," then a signal instance of faulty proofing is evident. The third change, in line 1502, of erchebisshop to Bisshop looks like the attempt of someone to correct what he thought was a factual error; however, the change turns the line into one of four beats, also a questionable change for anyone with the time to read it over. (The possible attention of the El scribe elsewhere to meter will be discussed below.)

In lines 1735-44 of the Summoner's Tale (where Hg and El are still closely affiliated), the Summoner is describing how the friar goes around begging while his companion writes down the names of the givers; then he switches at line 1746 to mimicking the friar as he begs (El variant in brackets):

Apart from the fact that El's hym is a unique variant, the we in line 1748 underlines the switch to indirect quotation and shows how another El editorial change has neglected a very limited context. Quite simply, the he of line 1745 seems to have led to the hym of line 1746, while the we of line 1748 was not noticed. Two unique variants in the Clerk's Tale (where Hg and El are also affiliated) show hasty editorial decisions for other reasons (lines 341-342 and 425-426, respectively): Once again, these changes are understandable, but once again they show the same misunderstandings of context found in many of the tales in El. In line 341 the scribe remembers that a tale is in progress but does not remember (realize?) that it still has two-thirds of its length (to line 1212) to go. That a marquis is doing the seeing apparently caused him to change the lowe to heigh in line 425, but little thought would have been needed to reject the change, both because the virtue of the lowly Griselda is the focus and even more because the change violates the point of the immediate context and of the tale as a whole.

These variants are unique to El, they are obviously editorial rather than unintentional, and they violate very immediate contexts. Numerous other unique variants in Manly and Rickert's "Corpus of Variants" (Volumes V-VIII) amply demonstrate why they labelled such changes "questionable" and

"some distinctly for the worse, even involving misunderstanding of the context" (II, 150). Other editorial changes, on the other hand, and even some of those listed above also demonstrate why Manly and Rickert describe the one who made the changes as "an intelligent person." An intelligent scribe who is copying line by line and occasionally deciding in the midst of copying a line to change the reading of his exemplar seems a far more likely candidate for the role of editor in these circumstances than some such person as an intelligent supervisor who is reading over the whole text and deciding which of its readings should be changed. The number and kind of El's variants by acco give perhaps stronger support yet to this identification of the El scribe as the editor.

More than a third of the unique variants in El are probably due to intentional change, whereas the great majority of the variants by acco are the result of failures of care; nevertheless, both causes operate in both kinds of variant (a point neglected, for the most part, in the Manly-Rickert analyses in Volume II where, most of the time, only unique variants are analyzed for editing). As noted, the ratio of variants by acco in El comes close to the Manly-Rickert estimate of "approximately twice" the number of such variants in Hg. Not only the larger number but the parallel with what was going on in the copying of other manuscripts make these variants by acco in El an even stronger basis for identifying the El scribe as the source of editorial variants: most of the variants by acco in those other manuscripts, after all, had their sources in both scribal accidents and scribal intentions.

In the Friar's Tale, besides its three unique variants and the fifteen which it shares with El (and which thus stem from the exemplar), Hg has three other variants; by contrast El, besides its own three unique variants and the fifteen shared ones, has eighteen other variants, three shared with affiliated manuscripts (but also probably by acco, given the shared exemplar of Hg and El and the general accuracy of the Hg copying). The other fifteen variants are shared by acco with other manuscripts. Unlike variants shared with affiliated manuscripts, variants by acco are traceable to errors and to decisions to edit which are common to the copying process itself; therefore, these are the kinds of variants we need not hesitate to call errors. When such errors by acco are added to those seven times as many uncorrected unique variants in El, the difference in its textual quality relative to Hg is seen to be greater still.

Of the other tales where Hg and El are affiliated, in the Summoner's Tale (to line 1991), Hg has no unique variants and has 5 by acco, while El has 4 unique variants (2 labelled "editorial") and 7 by acco. In the Clerk's Tale (apart from 9 variants shared with El), Hg has 4 unique variants, 4 others labelled "editorial" (of the total of 5 so labelled), and 19 others by acco with other manuscripts; El, conversely, has 22 unique variants (6 editorial) and 39 others by acco. In the Monk's Tale, apart from 14 variants shared with El, Hg has no unique variants and only 15 by acco; El has 41 other variants, 8 of them unique (of which 3 are called "editorial"). In the Nun's Priest's Tale,

Hg and El share 10 variants; in addition, Hg has "11 trivial variants," of which none is unique, while El has 11 unique variants and 36 others. Finally, in the Manciple's Tale, "Hg has ten variants not in El, one unique," and "El has 14 variants not in Hg, four unique" (II, 453). The list need not be extended to tales where the two manuscripts are not affiliated: in some of those the ratio is much larger, as in the unaffiliated portion of the Summoner's Tale after 1991 where Hg has 9 variants, none unique, while El has 15 unique variants and "more than 60 others" (II, 240).

Rather than use space to demonstrate how El's editorial variants by acco are like the unique variants at times in showing the same kinds of misunder-standings of limited contexts, the examples of variants by acco which follow will be used to make another point: many are very trivial pieces of editing indeed. In the Clerk's Tale, for example, the affiliated Hg records the received reading "knaue child" in lines 444, 447, and 612; El erased the knaue (boy) each time and replaced it with man. In the Nun's Priest's Tale, line 4084, Chauntecleer begins to relate his dream: "By god me mette [dreamed] I was in swich meschief . . . ." El changed mette to thoughte. In her reply Pertelote explains that all such dreams about red come from a superfluity of "rede cholera," as when people (or chickens) dream "of rede bestes that wol hem byte" (line 4121). Apparently because of the word byte, El changed the reading rede bestes to grete bestes. In line 4132 Pertelote begins her prescription for her husband, "Now sire quod she whan we fle fro the bemes . . . ." For we El substituted ye. And finally, when the dastardly fox has nabbed our hero, the Nun's Priest picks out the shriek of Pertelote from those of the other six wives (line 4552): "But souereynly dame Pertelote shrighte." For souereynly (in this context 'most loudly') El substituted sodeynly (suddenly). In the Manciple's Tale, line 105, El changed erthe to worlde.

Such a list could be expanded, but it makes another point about editorial changes made by El: many of the variants in El are by acco with unaffiliated manuscripts because they reflect the kinds of changes which contemporary scribes made in manuscripts when they thought a word might be misleading or unfamiliar to their readers (knaue to man, mette to thoughte), when they thought they had caught an error in their exemplar (we to ye, rede to grete), or when they simply were not sure about the word which they found (souereynly to sodeynly). Of course, a separate editor might have decided on such editoral changes down to the most trivial (e.g., the numerous changes of preposition, such as to to in, Clerk's Tale, 941; of article, such as a to the, Nun's Priest's Tale, 4072; or of verb, such as gan to began, Nun's Priest's Tale, 4068); nevertheless, the fact that such variants and many more of them are in acco with unaffiliated manuscripts where the scribes are the undoubted sources plus the fact that so many of them are too trivial to tempt an intelligent editor provide added evidence for the case that the scribe is the source of much of the editing in El.

A last piece of evidence for ascribing at least the majority of the editorial changes in El to the scribe is the almost total lack of signs that the scribe was

supervised by someone else: "A few marginal crosses mark errors; a few corrections are written in the margin, and there are also a few over erasures" (I, 148). Hg has no such signs at all: "Alterations of the text, including the filling-in of words and lines in spaces left by the scribe, were made by several hands, some much later, and perhaps none by a supervisor" (I, 269). Still, Manly and Rickert are probably right that El "must have been carefully supervised," given its expense and elaborateness. As they state elsewhere, "there were originally many more marks of supervision [in the Canterbury Tales manuscripts generally] than have survived" (I, 24). That many of the editorial changes in El were directed by such supervision, however, is much less likely for the simple reason that the overwhelming majority of such changes have no corresponding supervisorial mark.

This study has focussed so far on the substantive variants in Hg and El in tales where they are affiliated because there can be no question that the numbers and kinds of unique variants and variants by acco are attributable to the writing-process itself rather than to the lost exemplars. More indicative of the changes made in El, however, are the 15 unique variants and more than 60 others by acco in the half of the Summoner's Tale after the scribe lost the use of the Hg exemplar (for whatever reason). This is because the exemplars away from Hg were so often of a diminished quality which was apparent to the scribe himself: he used two different exemplars after line 1991 in his search for a good one. Manly and Rickert comment: "It looks as if the excellent MS used by Hg-El had become inaccessible to the El scribe at some point between 1991 and 2015 and he had turned to the copy used by d*1 [one version of the augmented d group] and Gg-Si Ad 3 Ch Ra3-Tc1-Gl, and when this gave out at 2158, used Vcd*2 [the ancestor of the c and another version of the augmented d groups]" (II, 241). Given the expense of the El project, inaccessibility seems less of a possibility than loss of leaves for the change after 1991. After the change to other exemplars, El changed the text of each far more freely than he had done in the first half.

In addition to the parallels with other scribal changes and the variation in variants according to the nature of the exemplar is El's lack of editorial marks. While it may be true that manuscripts which were closely supervised have lost many marks of that supervision, they still have enough marks left to be able to discern the closeness of that supervision: two supervisors can be discerned on Corpus Christi 198 (Cp), as can the places where the scribe failed to make the directed changes; even the eccentric Harley 7334 (Ha4) shows "many signs of supervision" (I, 220). Usually, directions by a supervisor would be marked on the page itself (by stylus, crayon, chalk, or even pen and ink); if they had been given separately (perhaps on sheets of paper or, less likely, on the rare and treasured manuscript being copied), there still should be places where the scribe failed to make the directed change and which would have been marked by the supervisor in his proofing for correction. On the contrary the unsupervised Hg scribe, as we have seen, caught a much greater percentage of his errors than were caught of El's greater number. The

lack of supervisorial marks plus the scantness of the proofing and correcting make very close supervision of the text, as opposed to the physical layout of the manuscript, very unlikely. Once again, the evidence points to the scribe as the source of the editing, or at least a great portion of it.

One last matter relating to the chart demonstrates how shadowy at times the boundary between accidental and substantive variants can be: Variant #10 (O/-e) and "the question of final e." For almost a hundred years, Chaucer scholars have debated whether Chaucer's meter is or is not iambic pentameter. Those who believe that it is pronounce the final e in appropriate places and, if they are editors, put in the final e enough times to make for a ten-syllable line (though 'headless' nine-syllable lines are increasingly accepted). In a series of articles and books after World War II, J. G. Southworth pointed out, in the course of attacking the prevailing metrical theory, that a great many of the final e's in modern texts of Chaucer's poetry had been introduced by editors with no supporting evidence from the manuscripts.[8]

Believers in the "final e theory" themselves, Manly and Rickert commented upon the "unusual care" with which El treated this matter of putting in final e. Certainly the ratio of 1129 times when the El scribe spelled with a final e and the Hg scribe did not to 341 times when the reverse occurred shows that El had a more consistent tendency to spell with final e; however, this study should have raised doubts about this registration's inevitably stemming from the exemplars which the El scribe copied. While such final e's (and their lack) may be explainable as stemming from the El exemplars, the shifting of such exemplars, the lack of corresponding evidence in contemporary manuscripts, and especially the contrasting lack of such final e's in the text of the more careful Hg scribe raise serious doubts that this is evidence of anything more than one scribe's habitual spelling. When the Hg and El scribes are copying the Friar's Tale from the same or a very similar exemplar, their registration of no final e in Hg to one in El is 17, the reverse 6; a similar ratio holds when they go on to the Summoner's Tale up to line 1991 (16/2), and it is close when El switches at line 1992 first to one exemplar and then to another (14/11). In the Reeve's Tale, where the scribes are copying from exemplars no closer to each other than the original scribal copy (O), the ratio holds (28/9).

These differences may stem from the El scribe's copying final e's from his exemplars when the Hg scribe fails to, but this study casts doubt on such a conclusion. The Hg scribe tries to be faithful to his exemplar and leaves blanks when he is in doubt; the El scribe is less careful in copying and in proofing, and he is much more willing to make regular changes in his text. Whether he makes the changes in final e's for the same reason that modern editors do can hardly be recaptured, but that they are his changes and not a faithful registration of his exemplars seems very likely.

Here, then, is the evidence that different scribes with different habits and

different views of their roles produced Hg and El. The likeness between their hands indicates either that one trained as the apprentice of the other or that both trained as apprentices of a third (the old-fashioned nature of the hand perhaps suggests the latter). On the other hand, the two scribes had different habits with such accidentals as spelling thow or thou, using & or spelling out and, using the brevigraph for -er-, and so forth. Of greater importance for the text of the Canterbury Tales, the El scribe had a different understanding of his role with the substantives of his exemplars: rather than leave space for a word or a line to be filled in later, as the Hg scribe did, the El scribe at times changed unfamiliar words for familiar ones, reversed less usual orders of words (e.g., is he to he is), and corrected what struck him as mistakes in his exemplar. In addition to the intentional changes which the El scribe made, the greater number of variants than Hg throughout, both unique and by acco, and especially those where the two manuscripts are affiliated textually show that he was also less careful in copying and in proofing his copy.

These differences in scribal habits between the scribes of Hg and El are more important than for the manuscripts of the Canterbury Tales in any other combination because both have texts of much higher quality than any other extant manuscript (Cambridge Gg.4.27 is probably in third place) and because of the role which El continues to play in editorial decisions. Manly and Rickert's acknowledgement of the much higher quality of the texts of the two manuscripts is shown by the way they devote part of each analysis of the variants and of the manuscript-relations of a tale to a comparison of the characteristics of Hg and El—whether they are affiliated or not. Their awareness of the very large role played by El in the half-century preceding their own edition (when a lightly emended El was virtually unanimous) prompted their warning that because of the evidence of systematic editing in El, "the unsupported readings of this MS must be scrutinized with the greatest care" (I, 150). Although their warning was applied only to unique variants, the present study has shown that it should be extended to the many more variants which El shares with other manuscripts by accidental coincidence (acco). The fact that an El variant can be found in other manuscripts for which there is no possibility of a common ancestor simply means that similar intentional and unintentional changes were common among fifteenth-century scribes: such changes produced not only the unique variants in El but also the variants by acco.

The availability of the First Fascicle of the Variorum Chaucer provides a convenient means of testing the statistics presented in this study. Such a testing will be welcome for refining the totals but is unlikely to threaten conclusions based on a great number of accidentals, large percentages of variation, and a regularity largely unaffected by the textual affiliations of the two manuscripts. While the identification of the scribe as the source of the editing in El does not rest on such numerical bases as does the differentiation of the Hg and El scribes, this also may be tested by a search of Manly and Rickert's analyses in Volume II and their "Corpus of Variants" in Volumes V-VIII. A

general return to the study of Manly and Rickert's monumental edition can only help the current understanding of the text of the Canterbury Tales. Such a return will certainly reinforce two of the most significant discoveries of the Manly-Rickert project: the very high quality of the text of the Hengwrt manuscript and the many dangers of too great a reliance on and trust in the text of the Ellesmere manuscript.

Notes