University of Virginia Library

F. N. Robinson's Editing of the Canterbury Tales
by
Roy Vance Ramsey [*]

After its appearance in 1933, F. N. Robinson's text of the Canterbury Tales steadily took the place of W. W. Skeat's as the one scholars taught from and cited in their professional articles.[1] This continued to be true despite the appearance in 1940 of The Text of the Canterbury Tales, ed. by John M. Manly and Edith Rickert (Manly-Rickert), with full collations and analyses of and a text based upon all of the manuscripts (Robinson had avowedly based his text upon ten), and despite data and analyses in Manly-Rickert which proved the faultiness of earlier analyses of manuscript relations to which Robinson had subscribed in his textual introduction. Not only have that first Robinson text and the little-changed second one of 1957 been cited as standard for well over fifty years, but the appearance in 1987 of a very lightly emended "Third Edition," under the general editorship of Larry D. Benson, makes the continued scholarly use of it as the standard text of the Canterbury Tales a possible, if not a likely, prospect. Such past, present and possible future uses of this text make particularly urgent a better and more general understanding of the methods by which Robinson edited because, as the present study will seek to indicate, careful analysis fails to support the belief of others and of Robinson himself in his textual methods and in his text of the Canterbury Tales.

A. The Initial Reception of Robinson's Text

As George Reinecke noted of the reviews of the first edition, "the Chaucer

student who has grown up in a world where Robinson is something of a fixture may well be surprised at the inconclusive nature of these evaluations, the rather small number of which is also surprising."[2] The one who wrote of the text in the friendliest terms, J. S. P. Tatlock, lauded the editor for choosing the via media between the eclecticism of "the literary man," Skeat, and the automatic registration of the consensus of the best manuscripts by the textual genealogist, John Koch. But in his lavish praise of the text, Tatlock clearly came down on the side of the habit by 'the literary man' of overbearing manuscript testimony; nonetheless, he raised a question about Robinson's full knowledge of the manuscript evidence, noting "that the editor shows little awareness of the inaccuracy in minor details of the Chaucer Society's printed editions of the manuscripts,"[3] the avowed primary sources of Robinson's decisions. What Tatlock himself failed to note was the role played in Robinson's text by the manuscript which Tatlock had earlier attacked so vehemently, Harley 7334 (Ha4).[4]

In a much less sympathetic review, Dorothy Everett noted how very large metrical considerations had loomed in Robinson's decisions: ". . . he has frequently adopted readings because of their greater metrical regularity, even making use for this purpose of the suspect MS. Harleian 7334. This practice will appear a dangerous one to many Chaucerians in view of the uncertainties which still exist in regard to Chaucer's metre and in view of the fact that some medieval scribes certainly 'corrected' their texts in order to produce smoother lines" (YWES [1934]: 103-105). In a more extensive "review" published five years later, Everett focussed upon "deficiencies . . . connected with what may be called the 'linguistic' side of Mr. Robinson's work" (MA 7 [1938]: 204-213). Thus, in her two pieces on Robinson's edition, Everett was the only reviewer to remark at length upon the two guides which, as Reinecke pointed out, seem to have loomed much the largest in Robinson's editorial decisions: "Robinson . . . had marked preconceptions about Chaucer's meter and did not view the poet's grammatical usage either as tolerant or as marked by flux" (242). As did the few others who ventured to write about Robinson's textual method at all, Everett surmised, "In the case of the Canterbury Tales . . . like the Globe editors, he has used the Ellesmere manuscript as his basis. . ." (104). Everett and others who have shared this surmise have had to read between the lines because Robinson's discussion of his textual method was very sketchy indeed.

B. Robinson's Discussion of His Editorial Method

In an otherwise friendly review, M. B. Ruud complained of Robinson's failure to note all of his departures from the text of Ellesmere (El) (MLN 50 [1935]: 329-332), and in noting this criticism Reinecke attempted to put the best possible face upon the scarcity of textual notes and the guardedly brief description of textual method:

In this Reinecke does seem to have fathomed at least a part of the reason both for the very guarded explanations which Robinson provided of his editorial method and for his "thin selection of textual notes." If this is the principal part of the explanation, however, it is hardly a defense. After all, those who use a text which claims the title of "critical edition" have the right to know the critical methods by which the editor arrived at it. Furthermore, they need enough of such evidence as alternative readings and textual notes which explain the reasoning behind the choices to be able to follow the editor's methods in their execution. We may first turn to what we may gather from what Robinson said about his achievement and methods and then proceed to what can be learned from charts comparing his choices with those of other editors.

With a very minor reservation, Robinson claimed for his entire text of Chaucer's works the status of a critical edition:

Reinecke's comment after quoting this passage certainly seems just: The passage alluded to about "archetypes" will be discussed after a look at Robinson's expansion of his one reservation about having produced a critical edition.

In alluding to the forthcoming Manly-Rickert "great critical edition," Robinson showed how little he was concerned by his reservation that "the manuscript materials [of the Tales] accessible to the editor have not been exhaustive." He initially pointed to the textual knowledge which the publication of Manly-Rickert would make available: "Their work, which is eagerly awaited by all Chaucerians, will shed new light on doubtful readings, and will probably make it possible for the first time to reconstruct the successive stages in the composition of the Canterbury Tales" (xxxiii). Yet Robinson quickly undermined the force of this and implicitly reasserted his claim to having produced something very close to definitive by his next sentence: "But

it does not appear likely that a text based upon the complete collation would be materially different from one that can be constructed from the eight published manuscripts which include the best copy, the Ellesmere MS., and are so distributed as to represent all the important groups of authorities" (Ibid.). The remainder of the two and a half pages of introduction devoted to the text of the Tales is taken up primarily with the claims of two manuscripts, Ha4 and El. The role of Ha4 in Robinson's text provides important information about his editorial method. Moreover, the exact relation of El to his text lies at the heart of the question of his method.

Reinecke's allusion to Robinson's speaking of "archetypes" referred to a passage less noted by reviewers than the one about presenting a "critical edition" but no less important:

Probably the chief reason that this passage has drawn less comment is that Robinson clearly was not using the word "archetype" in its usual sense. In textual studies the usual meaning of "archetype" is that given by those who practice the genealogical method of Lachmann, yet, as Reinecke noted, the review of Alois Brandl "praises the editor for his introductions and explanatory notes, only to attack him severely for not having been possessed by the spirit of Lachmann and pretending to a 'critical edition' while his true basis of textual judgement was his common sense" (235). Scholars who have attempted to understand what Robinson meant by an archetype or archetypes that might contain readings superior to those in the "critical text" have quite naturally turned to the best-text method, the chief rival to the genealogical method when Robinson edited his text.

In a section called "The Text (Theory)," Reinecke followed the most general surmise about Robinson's method from the first reviewers to the general editor of the "Third Edition": "Was Robinson aware of Joseph Bédier's devastating critique of the strict Stammbaum method in the Introduction to his Lai de l'ombre (1913)? This French scholar's analysis of the German method led him to adopt a procedure very similar to that which Robinson actually used for his Chaucer: choose a 'best' manuscript and edit it in the light of the others, using common sense and scholarly knowledge and experience" (239). Reinecke did not specify that El was this "best" manuscript, but because El was the only one which Robinson discussed at any length in positive terms (as opposed to Ha4 in negative ones) and because he said that he generally followed it in "mere matters of orthography" (xxxv), Reinecke clearly took for granted that El was Robinson's base, as did others who attempted to understand his method, such as Everett. For

example, in the textual introduction to the "Third Edition" of Robinson, its general editor, Larry Benson, first noted of the "over 160 changed readings" in Robinson's Second Edition, "That there were not more shows that Robinson was not convinced by Manly and Rickert's theory that the Ellesmere is an 'edited' manuscript whose scribe (or 'editor') frequently 'corrected' the meter. That theory has now been refuted by George Kane in his essay on the Manly-Rickert edition, which appears in Editing Chaucer (ed. Paul G. Ruggiers, 1985)" (Ibid.).

Although a comparison of the discussions of El by Variorum editors such as Thomas Ross and Derek Pearsall with the critique by Kane will show how far Kane is from having refuted the theory of editing in the production of El,[5] what is important about this quotation for the present study is that in defending the text of El, the editor of the Third Edition of Robinson seemed to believe that he was defending the basis of Robinson's text. A look at Robinson's own discussion of El shows some warrant for this general belief about his having used El as his base-text but also presents (especially when taken with other comments and the charts to be discussed below) stronger grounds for thinking that Robinson had another procedure in mind.

After noting that he had "no such means of testing" the variants of El "as was afforded for the Harleian manuscript by Professor Tatlock's study" and looking to the forthcoming Manly-Rickert to "show just how much scattered support such readings may have," Robinson continued,

Particularly in its rejection of El "as an independent witness," such a description gives very little support to the notion that Robinson used El as his substantive base; however, there is perhaps a little in his closing comment upon the El accidentals: "In mere matters of orthography, when verbal variants are not involved, the Ellesmere copy has been followed, as representing a good scribal tradition. But throughout all Chaucer's works . . . the spellings of the manuscripts have been corrected for grammatical accuracy and for the adjustment of rimes" (Ibid.). Although Robinson's discussion of El might arguably offer a little support to the very widespread assumption that he used El as base, his rejection of it as an independent witness and his discussion of such matters as "a superior archetype" seem to argue otherwise, as does the evidence of charts to be discussed below.

The key to Robinson's conscious editorial method seems to be in just what he meant by the phrases "the 'critical text'" and "a superior archetype." The usual meaning of "critical text" is that which an editor aims to construct on the basis of extant or reconstructed documents thought to be the most authoritative (though subject to emendation). In the absence of

the copy from which extant ones derive, the usual meaning of "archetype" is that which a textual critic aims to reconstruct by his studies of descendant copies. Since there is nothing in his discussions nor in such analyses as that of Reinecke to indicate that Robinson himself attempted such a task as reconstructing an archetype and since he seems to refer to a "critical text" apart from his own, the question must arise of where Robinson turned to find "the 'critical text' or a superior archetype" whose readings at times "appeared unsatisfactory or manifestly inferior." Whatever Robinson may have meant by putting the article "a" before "superior archetype" (seemingly allowing for the possible existence of multiple 'superior archetypes,' perhaps of the so-called 'Type A' and 'Type B' manuscripts), the remainder of his discussion seems to hint that he believed he found something like one of them in the agreements of what he called the "A type of manuscripts": Any resort to Manly-Rickert or to the Variorum discussions will show how various are the manuscript traditions lumped together in the two categories of "Type A and Type B," particularly the former.[6] In the event this is hardly surprising: one of Bédier's most trenchant challenges to the genealogical method was that far more often than credible it resulted in bifid stemmata, and Manly-Rickert's much more rigorous than usual application of the method showed how thoroughly misplaced had been earlier analyses which had resulted in a bifid stemma for the manuscripts of the Tales.[7]

Because the Manly-Rickert data and analyses so thoroughly disproved the earlier bifid stemma of the 'A type-B type' sort and because Robinson's second edition appeared seventeen years after Manly-Rickert, we might have expected a revision of this description of method and of the method itself; however, in the Preface Robinson showed the opposite of an awareness of any conflict:

Such an apparent unawareness of the degree to which the Manly-Rickert data and analyses had shown the opposition of the full manuscript evidence to the earlier bipartite grouping reinforces the point made by Reinecke about Robinson's neglect of manuscript relations: "Robinson's emphasis on 'the relation of the manuscripts' often in application seems secondary" (240). Reinecke illustrated the point with reference to the role of Ha4 in the text.

Well over half of Robinson's textual introduction was devoted to a discussion of the defects in Ha4 with the following explanation:

In illustration of Robinson's neglect of manuscript relations, Reinecke pointed to the way he nevertheless "tended to cling to this manuscript [Ha4]," including at times accepting some of its unique readings. Because he focussed, as others have done, upon the relations of Robinson's text to various manuscripts, Reinecke had no way of recognizing the full import of Robinson's neglect of 'the relation of the manuscripts" (240) nor of the source of his Ha4 readings. In the next section, charts showing the relations of Robinson's emendations to those of earlier printed editions will help to clarify the matter.

Because of the critical mysteries in Robinson's description of his method, because he was editing in the heyday of the best-text method, and because he joined earlier editors in calling El "superior to all others" (xxxv), it is small wonder that, from the first reviewers to the editor of the Third Edition, many have thought that Robinson meant that he had used El as something like a base text. After much study and correspondence with others, George Reinecke gave this description of Robinson's method: "It is therefore best to define Robinson's text as conservative, highly informed, and eclectic, though arrived at after much of the procedure for establishing a critical text had been performed. He was careful about his choice of copy text; he rarely accepted unique readings. When he did depart from his copy, it was usually for a reading connected either with grammar or with metrics (241). Reinecke and others who have understood Robinson's editorial method in this way have taken for granted that for the text of the Canterbury Tales Robinson used El as his base text or base copy (depending upon what Reinecke means by "copy text"). Also, Reinecke spoke for many in approving Robinson's "utilization of only ten manuscripts" as "a sensible decision": ". . . this number included all the manuscripts commonly termed 'good,' and going further would mean competing with Manly and Rickert, though it promised ever-diminishing textual returns" (Ibid.). The problem with this is that it neglects the manuscript relations discovered by Manly and Rickert as completely as Robinson did in his second edition. Furthermore, there is evidence that the text which Robinson emended was not that of a manuscript at all.

C. Robinson's Emendations

With the reservation that the text which Robinson emended is very much in question, what Reinecke says of Robinson's editing of the minor poems can

also be applied to the Canterbury Tales: "When he did depart from his copy, it was usually for a reading connected either with grammar or with metrics" (241). What Reinecke says of Robinson's general neglect of the Manly-Rickert analyses in his second edition applies to the neglect of manuscripts and studies in the first: "His chief preoccupation was the printing of regularly scanned, craftsmanlike, artistically significant lines conforming to his already determined opinions about Chaucer's grammar and meter" (250).

As mentioned, a good part of the reason that Reinecke and others have been puzzled about Robinson's method of editing has been because they focussed upon the manuscripts containing the readings he chose. Two charts to follow will show how the relations of the Robinson text to the texts of earlier editions are much more illuminating than such manuscript relations, including the supposed basic one with El. Skeat's text (SK) was widely regarded as standard throughout the years Robinson was editing, and Manly's edition (ML) appeared in 1928, five years before the first edition of Robinson and four years after Manly and Rickert had begun to gather and analyze all of the manuscripts. Manly chose El as base for his 1928 text in keeping with the general belief in its superiority (this belief, shared by Rickert, who also helped with the 1928 text, steadily eroded as their collations and analyses progressed). The first chart presents an analysis of the relations of the Skeat, Manly and Robinson 1 editions in three tales which they share:

Chart No. 1. Relations of the Skeat, Manly and Robinson i Editions

           

	General Prologue	Nun's Priest	Prioress	All
SK-ML-RB1 together	72	44	24	140
ML-RB1 (—SK)	27	15	2	44
SK-RB1 (-ML)	27	25	9	61
SK-ML (—RB1)	3	5	2	10
Totals	129	89	37	255

Unlike the much fuller charts upon which are based those in the present study (and for which there is no space), this study's charts list Robinson's editorial choices with those of other editors of the Tales only under the three categories of "El," "Hg," and "Others" for the following reason: with the exception of revisions of earlier texts based upon Ha4, since the publication in the mid-nineteenth century of the Six-Text Edition with transcriptions of the El and Hg texts first and second in the perceived order of their quality, the basic questions for most modern editors have been (1) which of the two to choose as having the best text, (2) when to emend the chosen base (usually El), and (3) whether to choose for such emendations the reading in the other manuscript of the pair (Hg or El) or in some other manuscript.[8]

The first point to be made about the chart above is that the editors' agreement that El has the best text is not the only reason that the readings of the three editions coincide in 129 of 255 cases. In the General Prologue, for example, the three editions agree in preferring the reading of Hg to that of El 13 of 72 times and in preferring the reading of other manuscripts 7 times.

In the Prioress's Tale, the three editors chose the Hg reading over one in El 4 times and a reading different from either 6 of 24 times. Then, in the Nun's Priest's Tale occurs the most dramatic agreement of all: in only 19 of 44 times of manuscript conflict did the three editors choose the El reading over Hg and others; 17 times they chose the Hg reading, 5 times they chose from other manuscripts, and 3 times they chose a shared El-Hg reading where other modern editors have chosen one from other manuscripts. Sheer probability quickly rules out coincidence in such cases because the agreements in readings other than those of the base-text are too regular to admit of coincidence: rather than simple decisions to emend the base-text reading in a particular locus, in all but a very few cases under the category "Others," they are decisions to emend it in exactly the same way. Most of the times when Manly and Robinson chose differently than Skeat, they chose a reading shared by Hg and El where he had chosen a reading from another manuscript, most often from Ha4.

Because he had started with a copy of El as base-text, Manly's decisions to emend were much more independent of Skeat's text than were those of Robinson: Manly chose differently from Skeat in 105 of 255 cases, 41%, whereas Robinson's text differed in 54 cases, 21.2%. When we remember the sequence of Skeat-Manly-Robinson, an even more striking fact emerges: in 44 of those 54 cases where he differed with Skeat, Robinson chose as Manly had done. In other words, in only 10 cases, 3.9% of the 255, does the Robinson text differ from the texts of both Skeat and Manly.

The next chart involves three tales either entirely absent from Manly's edition or else, in the case of the Miller's performance, limited to a few lines carefully selected for his high-school readers; however, the Globe edition came out in 1898, four year's after Skeat's Oxford one, and contains them complete:

Chart No. 2. Relations of the Skeat, Globe and Robinson i Editions

             

	Miller	Physician	Manciple	All
SK-GL-RB1 together	76	21	36	133
GL-RB1 (—SK)	10	8	8	26
SK-RB1 (—GL)	21	2	6	29
SK-GL (—RB1)	10	5	5	20
All separate	1	0	0	1
Totals	118	36	55	209

As did Manly later on, the Globe editor, A. W. Pollard, began with a copy of El; however, he was more like Robinson in being influenced by Skeat: of 209 variants involved, the Globe text differs from Skeat's 55 times and Robinson's text does so 46. Robinson was not quite as influenced by Pollard's decisions as he had been by Manly's; still, 26 of his 46 emendations in opposition to Skeat agree with Globe readings. Once again underlining the smallness of Robinson's allegiance to what others have taken to be his base is the fact that of the 20 readings which differ from both earlier editions, 12 also differ from the text of El.

In his editing Skeat demonstrated the eclecticism which characterized the nineteenth-century editing, even of a text with one manuscript deemed much superior, by choosing readings away from El more often than did later editors; yet in the Miller's Prologue and Tale, Robinson's text is away from El only one time fewer than is Skeat's, 50 to 51. Of 117 variants, the three editions are together 76 times; then, Robinson's joins Skeat's minus the Globe in 21 others and joins the Globe apart from Skeat's in 10 others, leaving only 10 instances where Robinson chose apart from these two editions preceding his. Striking evidence that he was not departing from either or both by a preference for the text of El are the two facts that (1) in 20 of the 21 cases where Robinson joined Skeat and parted from the Globe edition, the latter has the El reading whereas Skeat and Robinson have that of Hg, and (2) in 7 of the 10 instances when he differed from Skeat and the Globe, they chose the El reading and he the Hg.

All editors tend strongly to prefer the text of Hg in the Physician's Prologue and Tale, but the Globe editor, Pollard, still managed to remain a little closer to El than either Skeat or especially Robinson, preferring Hg readings by only an 11 to 8 ratio compared with Skeat's 12 to 8 and Robinson's 15 to 4. Of 36 variants, Robinson's text is with the other two 21 times, with Globe apart from Skeat 8 times, and with Skeat apart from Globe 2 times, leaving 5 times when his first edition differs from both: twice he chose with Hg when they chose El, once with other manuscripts when they chose the shared Hg-El reading, and twice with Hg-El when they chose readings from other manuscripts.

With the exceptions of the texts of Manly-Rickert and of Pratt, modern editors have tended by a slight margin to prefer the text of El in the Manciple's Prologue and Tale to that of Hg. Even so, the Globe ratio is closer to El than those of Skeat and Robinson: 18 to 9 vs. 18 to 15 and 18 to 16 respectively. Of 55 variants, Robinson is with both others 36 times, with Globe apart from Skeat 8 times, and with Skeat apart from Globe 6 times, leaving five times apart from both. The comparison here once again shows Pollard's closer relation to the text of El and Robinson's very indifferent one: in the 6 cases when Globe is apart from the other two, it is with El against Hg in 5 and with the shared Hg-El reading once when Skeat and Robinson choose a reading from other manuscripts; on the other hand, in the 5 cases where Robinson is apart from the other two, he has the Hg reading twice, a reading from other manuscripts twice, and only once has the shared Hg-El reading when Skeat and Pollard have a reading from other manuscripts.

Such close correspondences with earlier editions in choosing apart from the text of El raise particular questions about the general belief that Robinson began with El as his base, particularly given the contrast between his text's lack of great preference for El readings and the very different performances of other modern editors who clearly stated that they did use El as base, Manly, Pollard and Fisher.[9] As we have seen, Robinson's introduction is not really clear about his use of El, though he calls it the best text. Still, he

does claim to have given an important role in his editing to the Chaucer Society transcriptions and to subsequent studies of the manuscripts; furthermore, those claims not only have been generally accepted but now have been greatly reinforced by the publication of Robinson's text of the Tales with very few emendations as part of the Third Edition. As we have just seen, however, the relations of Robinson's first edition with those immediately preceding prove much more illuminating about his procedures than those with the manuscripts, including El.

A last chart comparing the major editions of the Tales will show something of their relations to the manuscripts and to each other:

Chart No. 3. Emendation Choices In Modern Editions of the Canterbury Tales

With El (—Hg)

               

	SK	ML/GL	R1	MR	R2	PR	FI	RV
GP	50	76	55	44	53	49	58	54
MiT	67	88	62	26	59	41	71	59
PhT	8	8	4	3	4	3	6	4
PrT	16	20	15	9	14	10	19	14
NPT	22	44	25	14	25	17	34	24
McT	18	25	18	12	18	14	19	18
Totals	181	261	179	108	173	134	208	173

With Hg (—El)

               

	SK	ML/GL	R1	MR	R2	PR	FI	RV
GP	38	15	34	43	36	38	31	36
MiT	39	19	45	81	48	66	36	48
PhT	12	11	15	16	15	16	13	15
PrT	9	4	10	16	11	15	6	11
NPT	39	17	37	48	37	45	28	39
McT	15	9	16	22	15	19	15	15
Totals	152	75	157	226	162	199	129	164

With Others (—Hg El)

               

	SK	ML/GL	R1	MR	R2	PR	FI	RV
GP	38	8	16	8	15	12	7	8
MiT	12	8	5	1	5	2	5	4
PhT	14	8	8	7	8	7	6	8
PrT	10	6	11	4	10	3	5	6
NPT	21	8	17	7	17	9	9	12
McT	14	6	9	10	11	9	7	9
Totals	109	44	66	37	66	42	39	47

First, the chart shows that their dependence upon the text of the second Robinson edition is at least as close as the textual introductions of Pratt and of Benson lead us to expect.[10] Second, the preference for the El text by Pollard, Manly, and Fisher supports their having begun with that as their base. Third, the unavowed relation of Robinson's editions to Skeat seems manifest in the figures for El-Hg-Other: Skeat 181-152-109, Robinson 1 179-157-66, and Robinson 2 173-162-66 (the disparities in the totals stem from shared El-Hg readings accepted by one or the other, usually by Robinson). And finally, when the General Prologue is omitted from the list, the trend begun in Skeat's edition of favoring Hg readings as much as those from El becomes even clearer: Skeat 131-114 El-Hg, Robinson 1 124-123, Manly-Rickert 64-183, Robinson 2 120-126, Pratt 85-161, Fisher 150-98, and Riverside 119-128.[11]

One last comparison of the first Robinson edition with an earlier printed text should put to rest any lingering doubts about his primary reliance upon earlier editions and not upon manuscript evidence. The data in Charts 1 and 2 left 30 instances when Robinson 1 was apart from all three of the editions which immediately preceded his, 10 apart from Skeat and Manly and 20 apart from Skeat and Globe. Following the affiliations of Robinson 1 with Trywhitt's edition of 1775 (the best one to appear before Skeat's) accounts for most of the 30: 26 of the 30 times when Robinson 1 differs from the three editions immediately preceding it, it agrees with the reading in Tyrwhitt. Although Tyrwhitt did not have access to Hg, he used as his base an earlier printed text with readings going back to manuscripts related to Hg, so that, given his use of the other editions, the seeming Robinson choices of Hg which differ from immediately preceding editions over the seven other manuscripts of the Chaucer Society transcriptions very possibly owe much to like readings in Tyrwhitt: 16 of the 26 times when Robinson 1 coincides with Tyrwhitt apart from the other editors, the reading is that of Hg in preference to those of the other seven Chaucer Society manuscripts, and 5 times is the shared Hg-E1 reading in place of a reading of the other six. Tyrwhitt listed Dd as a manuscript of "most credit,"[12] and a Dd reading accounts for 3 of the 5 remaining times when Robinson 1 is with his edition.

Although he accepted the general consensus about Robinson's use of El as base, Reinecke had several insights which anticipated the conclusions of the present study. One of the most illuminating is his discussion of why "the spellings in [Robinson's] Glossary do not always match those in the text":

Here and elsewhere, either "base copy" or "base text" (depending upon exactly how he envisioned the process) would have been preferable to Reinecke's "copy text"; however, his discovery is illuminating and fits with what the charts and other data strongly suggest was Robinson's actual way of editing.

Two decisions by Robinson to differ from Skeat's practice shed a great deal of light upon his use of Skeat's Glossary: he decided systematically to respell the Canterbury Tales following El practice, and he decided to put the

Tales at the beginning of his edition. Thus, some of the words in Skeat's Glossary are spelled differently from their first occurrence in the Canterbury Tales because earlier works in Skeat such as Troilus and Criseyde had different spellings. Differences in spelling between the first Tales occurrence in Robinson's text and in his glossary can have no such explanation, however, because he put the Tales first and because he changed the spellings to those of El.                

	Skeat	Skeat	Robinson	Robinson
Line	Glossary	Text	Glossary	Text
A 41	aray	array	aray	array
A 2046	arayed	arrayed	arayen	arrayed
I 567	areise	areysen	areisen	areysen
A 2602	arest	arest	arest	arrest
B 4090	areste	areste	arest	areest
I 580	aretten	arretteth	aretten	arrette

Because he was glossing generally and not by first occurrence, Robinson added an 'n' for the infinitive to the verbs, and he removed the 'e' from the second "areste"; otherwise, however, he left the spelling as it occurred in Skeat's glossary—despite having respelled in the text to fit El practice.

A striking example of his practice in this regard is the word "bisinesse." Both Skeat's and Robinson's glossaries spell the word "bisinesse" in the same way, and Skeat's glossary cites the first occurrence of his first definition at B 1415 of the Tales. Although Skeat has the same spelling there and in the other 31 Tales occurrences as in his glossary, Robinson spells it "bisynesse" all 32 times. This systematic spelling of the word in Robinson's text is in keeping with his claim to have followed El "in mere matters of orthography" (xxxv) in the Tales. Still, Robinson did not regularize the spelling of other texts, so that the various texts exhibit a total of 7 ways of spelling "bisinesse" (e.g., "bysynes") in the total of 61 occurrences of the word in the Tales and elsewhere; however, only one matches the glossary spelling ("Fortune," 75). Such an identity of spellings with those in Skeat's glossary and such differences between Robinson's own text and glossary suggest that Robinson's use of Skeat's glossary was akin to his use of Skeat's text of the Tales.

Putting the two glossaries side by side should convince anyone of the great debt which Robinson's glossary owes to Skeat in definitions as well as in spellings. Because Skeat attempted to cite the first occurrence of a difference in meaning of each word and because Robinson listed rather the most usual meanings of a word, Skeat's is a much fuller glossary. Still, the likenesses in definition are quite regular. For example, Skeat glossed the word "thakketh" as "strokes, pats"; glossing generally and not by first occurrence, Robinson changed to the infinitive "thakken" but kept the gloss "stroke, pat." That they were not sharing the same source in this seems clear in turning to the likely alternatives. Bosworth-Toller glosses Old English "þaccian" "I. to pat, clap, strike gently, with the open hand or the like," and the O.E.D. glosses "thack" "1. trans. To clap with the open hand or the like; to pat, slap

lightly."[13] While all agree on the definition "pat," the agreement of Skeat and Robinson in "stroke" is striking because this seems a different action from such as "strike gently" and "slap lightly." This and other evidence strongly suggests that, rather than beginning fresh with either, Robinson used as bases for emendation both Skeat's text and his glossary.[14]

In constituting his text, it is clear that Robinson did look at the Chaucer Society transcripts from time to time, just as he implied. But, just as the earlier charts made it clear that Robinson did not start, as Pollard and as Manly had done, with a transcription of El as base, it seems equally clear that he did not systematically collate the published transcripts of the eight Chaucer Society manuscripts or he would have been much clearer about the relations of the El, of the Hg, and especially of the Ha4 readings to his emendation decisions (compare Tatlock's point that Robinson showed "little awareness of the inaccuracy in minor details of the Chaucer Society's printed editions of the manuscripts"). Similarly, Robinson could hardly have studied the Manly-Rickert data and analyses very deeply for his second edition: (1) as noted, his reprinting the textual introduction of his first edition virtually intact showed that Robinson misunderstood how completely Manly-Rickert had exploded the earlier bipartite analysis of manuscript relations accepted there; (2) his reprinting of readings with little or no support beyond a single manuscript, as in the continuance of unique readings from El and even from Ha4, showed a lack of systematic collation akin to his practice earlier with the Chaucer Society transcripts; and (3) the smallness of the number of Robinson's emendations of his earlier text, the fact that much the greater portion of them were from Hg, and his apparent unawareness of how his decisions were continuing a trend begun by Skeat of moving from the text of El toward that of Hg all demonstrate how little attention he paid in his emendations to the manuscripts and their relations as revealed in Manly-Rickert. With regard to the last, Reinecke showed the surprise of one who assumes "Robinson's dedication to Ellesmere" when he noted of the second edition, ". . . fully two-thirds of the new readings (excluding those not manuscript-related) follow Hengwrt" (250).

Thus, various ways of comparing Robinson's editions with manuscripts and with other editions all agree in the same conclusion: the evident relationship of his glossary definitions and spellings to those of Skeat, the lack of regular, consciously derived nearness in text of either Robinson edition to a manuscript, and the neglect in both editions of manuscript relations all conjoin with the greater closeness of Robinson's text to that of Skeat in substantive emendations to argue very strongly that for much the greater part of the time that Robinson worked upon his first edition of the Canterbury Tales, the de facto physical base upon which he made his changes in substantives and accidentals was Skeat's printed text. When we remember how Robinson's misunderstanding of the Manly-Rickert data and analyses in his second edition fits evidence in the first of his lack of interest in the manuscripts and their relations, and when we also remember that he spent the

greater part of his career in working upon his edition of Chaucer, we can see how Skeat's edition could play such a large role in Robinson's own edition.

In the first third of the century, anyone interested in questions about the text of the Canterbury Tales would have been engaged in something like an informal version of the process of comparing Skeat's text with other information because his was much the most used in classroom and in study; thus, a man such as Robinson spending twenty-nine years in his editing (Reinecke 231), all the time teaching from Skeat's text, could hardly have avoided such a procedure. Pollard himself had a high reputation as an editor at the time, so it would have been natural for Robinson to compare the readings of the Globe text with those of Skeat. Also, Robinson would have been aware of Lounsbury's praise of Tyrwhitt as an editor of the Tales: "No more thorough and conscientious editing had ever before been applied to the elucidation of a great English classic."[15] And finally, as he mentioned in his introduction, Robinson was aware that Manly (and Rickert) had been studying all of the manuscripts in the years just before the publication of Manly's own edition of 1928. These facts and the data in the preceding charts seem to show that, while Robinson pointed in his introduction to manuscripts and to textual studies and while he tried to take them into account (particularly the division of the manuscripts between "type A" and "type B"), his heaviest and most regular reliance must rather have been not upon manuscripts and manuscript studies but upon the printed texts which were the result of previous editors' studies of the manuscripts.

His actual editorial practices and certain hints in his textual introduction seem to show that Robinson followed this procedure because he believed that among them previous editors had reconstructed something very close to the 'A archetype' in their editions, thus pointing to their collective results as the "critical edition" to which he alluded as well as the primary sources of his information about a "superior archetype" (with the Chaucer Society transcriptions of such manuscripts as Lansdowne and Petworth representing the 'B archetype'). Such a belief, together with the fact that Skeat's edition was universally accepted as standard throughout the twenty-nine years that Robinson worked on his own edition, help to explain his very heavy reliance upon that edition. Thus, whatever Robinson consciously meant in his textual introduction about his responses "when the readings of the 'critical text' or of a superior archetype appeared unsatisfactory or manifestly inferior," his actual way of editing seems to have been (1) to begin with Skeat's text, (2) to compare its readings with those of other editions, particularly those published since Skeat, (3) then to compare alternative readings of the published transcriptions of Chaucer Society "Type A" manuscripts with those of "Type B," and (4) to favor a reading from one of the A group (especially one in El or Hg) over a B reading which other editors had chosen (usually from Ha4 or Cp) only when it could be reconciled with his own ideas about Chaucer's meter and grammar. Since the earlier editors had shared less rigid forms of

the same ideas, Robinson could almost always find support from their own decisions in such matters.

D. Flaws in Robinson's Editorial Method and Why They Matter

Because for more than fifty years Robinson's text has been Chaucer's text for almost everyone, resistance to the idea that it is basically flawed is inevitable. While one reader of an earlier version of this study pointed out that Robinson's editions "have presented a sanitized, smoothened Chaucer to too many generations of students and scholars," two others spoke for what will likely be a majority in their doubts that any material gains would be made if Chaucer's text were edited in a different way. I have only the space to respond to such doubts by the description of method given above and by a brief discussion of why an alternative is preferable. Fortunately, however, ample materials are available now, as they were not in 1933, to anyone who honestly wishes to resolve the question for himself. In closing I shall briefly address the question of a preferable method.

In response to my conclusion that Robinson's text "indicates a heavy dependence upon earlier editions," a critic might ask, "Why necessarily bad?" If the editors of those editions had been in full possession of the relevant manuscript evidence, if those editors had been primarily responsive to that evidence rather than to earlier editions, if further the editor who used those earlier editions had taken both the manuscript evidence and the editions fully into account, and if finally that editor had given a full and accurate account of his procedure, then that way of editing might at least be arguable. I believe that I have presented enough evidence that such was not the case with Robinson's edition, but I would argue that such a procedure is not desirable in any case.

In the first place, the editors who published before Manly-Rickert possessed only a fraction of the manuscript evidence. Furthermore, they worked with a reading of the manuscript evidence which any resort to Manly-Rickert or to one of the Variorum fascicles will show was badly flawed. And finally, those editors were influenced by much faultier editions published earlier which further vitiated their work. Skeat very eloquently conveys the shock of learning how the editors before him had done their work:

The accretion of such changes is, after all, a version of what happened in the descent of the manuscripts themselves: not only did scribes inevitably make inadvertent errors in copying, but their suspicions that such errors existed in their exemplars led some of them to make deliberate changes of just the kinds made by later editors with similar suspicions, the changes accumulating through the years so as to move the text progressively away from what Chaucer wrote. While editors of the Tales such as Robinson who have based their editions upon earlier ones have removed some of the accretions to Chaucer's text of earlier editions, they have brought forward many others, as in Robinson's acceptance of very questionable Ha4 readings adopted by Skeat. Furthermore, when, as with Robinson, they have worked not inductively from the manuscript evidence but deductively from such grounds as rigid notions about grammar and meter, then they have necessarily added their own accretions to the 'cloak' of editorial changes which Chaucer's text wears.

Thus, this way of editing by primary resort to earlier editions performs in a way effectively the opposite of the genealogical method which strives to work back through the layers of change to the archetype. The principal reason for an initial attempt to work out the relations of the manuscripts is that by doing so an editor can have objective help in detecting and removing the changes to which an author's text has been subjected since he wrote it. By choosing among earlier editorial choices of readings with little regard for their bases in the manuscripts and their relationships, Robinson piled his own particular set of choices upon those made by earlier editors, in their turn based upon others, often in a chain all the way back to the earliest editions.[16] If nothing else, an editor who begins with the testimony of fifteenth-century manuscripts will have to deal with the accretion of only a few decades of changes in Chaucer's text rather than five centuries of them.

One very important reason that the editor's starting-point usually has such great influence upon his decisions and thus upon his completed text is simply the influence of use. A look at the hold which the very erratic Ha4 text had upon the nineteenth-century understanding of the text should be cautionary for all: as the basis of the printed texts which scholars learned from, taught from, and cited in their studies, that manuscript and its readings gained so much credibility from sheer use that some of them labored long and hard after the appearance of the Six-Text Edition to explain the vagaries of Ha4 (in clear opposition to the testimony of the six) as revisions by Chaucer himself. Robinson's text has been a like case for this century, with the added confusion that the supposed basis of his text in El has not been as most have thought: this belief has led some to argue for El's superiority without reference to the evidence in Manly-Rickert and in the Variorum fascicles that the text of El was carefully changed in its making in much the way a modern editor might do, and despite the evidence of collation that in the great majority of cases of conflict the evidence of independent lines of manuscript descent favors the Hg reading.

Thus, the latest edition of Chaucer, the Riverside or "Third Edition" of Robinson's text, follows and perpetuates a very unfortunate example. The first flaw in Robinson's way of editing is that it neglected the all-important first step of inductively examining the manuscript evidence. Rather, Robinson's first steps were deductive and based upon untenable assumptions: that the manuscripts are related by a bifid stemma, that among them Skeat and other early editions adequately represented the better branch of that stemma

("type A"), and that Robinson possessed virtually unerring guides in his ideas about Chaucer's grammar and meter. Also, Robinson's primary resort to editions rather than to manuscripts caused patterns in his choices which were at odds with his own theories: Ha4 readings adopted by Skeat persist in Robinson's text despite his acceptance of Tatlock's critique of its testimony; also, because his primary points of reference were the editions, he failed to note how his textual decisions continued the movement from El to Hg readings begun by Skeat, despite the assertion of both editors that El was clearly the superior (a matter to be developed in a separate study).

Because manuscript testimony had only secondary importance for Robinson, and because he believed that his very neat and rigid ideas about Chaucer's meter and grammar fully matched Chaucer's own throughout the fifteen or more years that Chaucer worked on the Tales, those ideas served as much more constant guides for his decisions than the manuscript testimony. Space is too limited to demonstrate the effects of Robinson's imposition of these ideas upon the text, but materials are readily available for anyone to make the comparison: for reading an 'unsanitized, unsmoothened' Canterbury Tales, the First Fascicle of the Variorum Chancer[17] presents a facsimile and transcription of the Hg text with marginal collation of El, so that anyone with a copy of Robinson's text can compare it with one copied by a scribe who consciously changed nothing (Hg) as well as with one copied by a careful scribe given to a certain amount of "smoothing" of the meter and occasionally to coming up with a 'better' word (El). Alternatively, those who wish to compare Robinson's way of editing with that of other editors and with the available manuscript evidence may turn to the other Variorum fascicles of tales which have been published to date (with collations below the text). In either case, because of the influence of the texts of El and especially of Robinson, the reader's initial tendency may be to lean toward the 'smoothened' version. Before long, however, a preference should emerge for the more lively verse, freed from five centuries' encasement in editorial decisions and supported by the best manuscript evidence.

Notes