by Hiroshi Yamashita and Toshiyuki Suzuki
No autograph of The Faerie Queene survives and its text has been transmitted to us through the printed editions. Bks I-III were first published in 1590 in a quarto volume, together with some supplementary matter including the >Letter to Raleigh= (LR). In 1596, two quarto volumes were published, consisting of the second edition of Bks I-III and the first edition of Bks IV-VI. In 1609, the first folio edition of the poem included the first edition of the >Two Cantos of Mvtabilitie=, which are assumed to be part of Bk VII. Later editions seem to be derivative reprints without any independent authority.(For bibliographical details, see F. R. Johnson 1933, the Pforzheimer Catalogue, Yamashita et al 1990, and Yamashita et al 1993.)
Since J.C. Smith=s Oxford edition of the poem in 1909, modern editions have been based on the two quarto volumes of 1596 for Bks I-VI and on the 1609 folio for Bk VII. The choice of the copy-text for Bks IV-VII is indisputable because the 1596 quarto and the 1609 folio are the only substantive editions. The anomaly is that the copy-text for Bks I-III has been the second edition of 1596 rather than the first edition of 1590, which certainly deserves more attention than has hitherto been given. Editors from Smith onward have thought lightly of the merits of the first edition, claiming that >the text of 1596 shows sufficient alteration for the better to justify the opinion that Spenser was responsible for an incidental revision= (Var 1.516). Yet the first edition was very probably set from Spenser.=s own manuscript, while the second edition is for the most part a mere page-for-page reprint of the first edition.
Clearly, the 1590 text has preserved more of the generally accepted Spenserian characteristics, particularly his spellings, which have been established through studies of extant documents written by his hand. (See R.M. Smith 1958.) The correction of the 1590 errors by the 1596 edition was not thoroughgoing. The second edition failed to correct nearly half of the errors listed in >Faults escaped in the Print= (F.E.), the errata printed on the last page (Pp8v ) of the first issue of the first edition. An analysis of the 1596 corrections agreeing with F.E. suggests that it was not consulted in the reprinting of Bks I-III. (For an analysis of F.E., see T. Suzuki 1997.) The second edition ignored about 48 corrections out of the 110 listed in the errata. It would be reasonable to suppose, as did both J.C. Smith and F.M. Padelford, the textual editor of the Variorum edition, that neither Spenser nor the printer attempted to make a systematic correction using F.E.
Although no fewer than 83 misprints in 1590 were corrected independently of F.E., 183 new misprints were introduced in 1596 (Yamashita et al 1993: xii, 269-72). In addition, confusions of personages, apparently unintentional discords in rhyme and even incomplete lines (II iii 26.9 and II viii 55.9) survived the second edition. In view of the quantity and quality of its errors, we doubt that the poet himself supervised the printing of the second edition of the first three books. As F. B. Evans 1965:62 suggests, >Spenser may well have sent ahead the necessary copy [for the revised second edition] and entrusted his publisher with the reprinting=. Probably he could neither proofread the second edition nor supervise its printing because he may have been in Ireland at that time, and this might also explain why F.E. was not consulted in the printing of the second edition. A collation of 1590 and 1596 suggests that the authorial alterations made in the latter were limited to a small number and not as extensive as it has appeared to modern editors. Major revisions, which indeed seem to be authorial, are the addition of a stanza (I xi 3) and the replacement of the last five stanzas at the end of Bk III of 1590 by three newly composed ones. There are some substantive changes which may have been revised by the poet himself, but others could well have been caused by a compositor=s or proof-reader=s tamperings. (For an analysis of the variants between the two editions, see Yamashita et al 1993.)
It seems that the influence of the second edition on contemporary readers could have been less than the first, for the number of copies was probably fewer than that for the first edition of 1590 or the Second Part (Bks IV-VI) of 1596, as F. R. Johnson 1933:19 points out in his Bibliography: >probably only one-half or one-third as many copies were printed of the second edition of the First Part as were printed of the first edition of the First Part=. This edition seems to have been no more than a supplement to make good the shortage of the 1590 quarto copies, although the text includes some additions and revisions. The book appears to have been very hastily and perfunctorily made, for it reprinted, apparently just to fill up the final leaf of the volume (Oo8r-Oo8v), only three commendatory poems, omitting LR found in the 1590 edition. This omission has been variously interpreted (see LR, n.), but it was possibly made for economical reasons: the printer wanted to save the cost and time of printing the new gathering Pp, which would have been necessary if he had intended to reprint the letter. On the other hand, the publication of the first edition of Books I-III, which included LR plus seven commendatory poems to Spenser and seventeen dedicatory sonnets by Spenser (ten in the first issue of the quarto), was indeed monumental in many ways.
The present edition, therefore, bases its text for Bks I-III on 1590, for Bks IV-VI on 1596, and for Bk VII on 1609. What follows is a description of each edition together with related textual matters and editorial procedures we have adopted for this edition. For the entries of the copies for the early editions in the Stationers= Register, see the Chronological Table.)
(1) Books I-III
For the title page of 1590, see Facsimiles. Owing to the varying states of the dedicatory sonnets at the end, extant copies differ in the collation of the last few leaves. The one most commonly found in libraries and markets today collates as 4o in eights: A-Z8, Aa-Pp8, Qq4; 308 leaves. The text runs to approximately 18,500 type-lines. On the basis of the skeleton pattern and the records of John Wolfe=s printing house, Johnson concluded that probably at least two presses were used for the printing of this quarto. However, his analysis of the pagination errors in gathering F led him to the erroneous conclusion that >one compositor or pair of compositors set both the outer formes and a different pair set both the inner formes= (17). As he says, four different skeleton-formes can be identified, which precludes the supposition of printing with one press; and if we hypothesize two presses, it follows that there must have been more than one compositor.
Inquiring into the number of compositors and their stints, we have examined spelling variations, typographical characteristics such as the spacing of words and the punctuation marks (Yamashita 1981), recurring impression of identifiable types, ornamental boxes surrounding canto arguments, and so on. Among these, the first two afforded particularly useful information to identify the compositors. The space test, examining use or non-use of space before colons, semicolons and question marks, has suggested division of work between 1r-3r (the first five pages) and 3v-8v (the remaining eleven pages) in many gatherings. The spelling test has not only corroborated the results of the space test but also strongly suggested that a third compositor was probably involved in the setting of pages 6r- or 6v-8v. 
The compositor who set most of the first five pages in gatherings C-Oo (designated as Wolfe=s Compositor X) can be distinguished from the second (Compositor Y) and/or the third compositor (Compositor Z) by his frequent use of the following spelling forms: else, forrest, foorth, little, whyles and whylome as against els, forest, forth, litle, whiles and whilome. Among these, litle, whiles and whilome are Spenser=s well-established preferences. Compositor X is more clearly identified by his persistent rejection of such Spenserian final voiceless stops as -att, -ett, -itt and -ott in favour of -at, -et, -it and -ot and of such initial or medial long vowels and diphthongs as aid, maid, pain, spoil, grownd, rownd and sownd in favour of ayd, mayd, payn, spoyl, ground, round and sound. Compositor Y, who was mainly responsible for 3v-5v in each gathering, is distinguished from Compositor Z by his use of medial i as in daies, eie(s), guide and noise as against Compositor Z=s dayes, eye(s), guyde and noyse. Compositor Z set all of 6v-8v in gatherings C-Oo except for Y6v, which was probably set by Compositor X as the evidence from spacing suggests.
As for the 6r pages, evidence available at present is not sufficient to assign them with confidence to any of the three compositors, though it might be said that all were involved in setting these >extra= pages. To turn to the four gatherings A, B, Pp and Qq, both the space and spelling tests suggest that the first eight pages of gathering A including the title page (A1r), Dedication (A1v) and the last page (A8v) were set by Compositor X alone, while the remaining seven pages were done by Compositor Z alone. Gathering B was possibly divided into three sections; 2r-3r and 8v were undertaken by Compositor X, 3v-5v by Compositor Y, and 6v-8r by Compositor Z. The first two pages could not be identified. As for the last two gatherings Pp and Qq, suffice it to say that LR (Pp1r-Pp3r) may have been set by Compositor X. (For a fuller description of the three compositors, see Yamashita et al 1990: ix-x.)        For the title page of 1596, see Facsimiles. The collation of this edition is 4o in eights: A-Z8, Aa-Oo8; 296 leaves. Evans identified two pairs of skeletons and a pair of compositors, designated as Richard Field=s Compositors A1 and B1, who alternated in setting by gatherings.
In determining the text for Bks I-III we have collated twelve extant copies of 1590 and have chosen the copy in possession of Yamashita as the base text because it was not only the easiest of access but also it proved to contain more corrected formes than any other copy. (For the collated copies and the states of variants, see the Textual Notes.) Except for the British Library copy (C12h17), all copies we have collated have blank spaces in stead of the Welsh words on X7v (II x 24.8-9), which have been filled by consulting the British Library copy.
Of the two major variants between 1590 and 1596, the stanza (I xi 3) added in 1596 has been inserted into our text possibly because it was an omission by the 1590 compositor, but we have kept the five stanzas at the end of Bk III as in 1590 and laid out the three stanzas newly composed for 1596 after the 1590 ending. As for the two eight-line stanzas (I x 20 and III vi 45) in 1590 and 1596, we have adopted the half lines added in 1609. Other revisions possibly made by Spenser for 1596 are not always adopted but are recorded in the Textual Notes.
We have adopted the corrections of F.E. appended to 1590, for it is probable that Spenser was concerned in the preparation of this errata, which includes not only corrections of compositorial errors but also authorial revisions. A considerable number of revisions involve stylistic changes and adjustment of syllables to the metre. However, F.E. appears to have been made hurriedly and haphazardly, for it overlooked 83 obvious misprints, as noted above, and no fewer than 48 possible misprints and doubtful readings. It is notable that the number of the F.E. corrections in Bk III, where 33 misprints and 22 doubtful readings remain uncorrected, is far fewer than those in Bks I and II. (See Yamashita et al 1993.)
In addition, the list has its own faults; erroneous citation of pages, improper quotation and misprint of words sometimes make it difficult for the reader to locate the errors in the text. Simple citations of the definite article >the= or the demonstrative pronoun >that= as well as citations of spellings and capitalizations that differ between the text in the quarto and F.E. puzzle the reader when they occur more than once on the same page of the text.
It is probable that Spenser himself marked the corrections on the printed sheets, and a proof-reader of the printing-house compiled the corrections out of the marked sheets, noting only page numbers, errors and corrections as they were marked on the sheets.
(2) Books IV-VI
The title page of >The Second Part= of The Faerie Queene, see Facsimile.
The edition collates as 4o in eights: A-Z8, Aa-Ii8, Kk4; 260 leaves. Evans 58, 65-67clarified two pairs of skeletons different from those used in setting the first part and also a different pair of compositors, whom he names Field=s Compositors A2 and B2. From the results of his spelling test, he infers that Compositor A1, who shared the work on the first part with Compositor B1, and Compositor A2 are possibly one and the same compositor, whereas Compositor B2 is different from Compositor B1. His bibliographical evidence shows that this quarto edition was probably printed from Spenser=s own manuscript and that he saw the printing through the press. We have chosen it as copy-text for Books IV-VI and collated sixteen copies including the one in possession of Yamashita. We use the Yamashia copy as the base text, adopting into it the corrected state of readings where we find press variants.
In determining the text for Books IV-VI, we have followed the editorial procedures adopted in Books I-III and tried to reproduce the copy-text as closely as possible.
(3) Two Cantos of Mutabilitie, Book VII
The collation of the first folio edition in 1609 in which the two Cantos of Mutabilitie appear for the first time is 2o: A-Y6, Aa-Hh6 , Ii4; 184 leaves. The stanzas are numbered throughout each canto for the first time. Like the 1596 second edition, it prints only three commendatory verses at the end of Bk III. For the head title of the two cantos, see General Introduction. Except for these cantos, which run Hh4r-Ii3r, this edition was set from the two-volume edition of 1596, but both spelling and punctuation were considerably modernized. Some substantive variants between 1596 and 1609 may have been due to Spenser=s marginal interpolations in the printer=s copy. We have chosen the Yamashia copy as the base text and collated seven copies but have found few press variants.
(4) Editorial Procedures
The stanzas are numbered throughout each canto and the numbers for the last five stanzas of III xii are marked with an asterisk to distinguish them from those of the current editions based on the 1596 edition. All substantive and accidental emendations made to the copy-texts and substantive variants of some significance in the early texts (1590, 1596, 1609 and occasionally 1611) are recorded in the Textual Notes.
The original spelling of the copy-texts has been basically preserved, but in a few cases where the old interchangeable spelling is likely to produce unnecessary confusion (e.g. of/off, to/too, there/their), spelling has been changed according to modern usage. Though some proper nouns are variously spelled in the copy-texts, they are not standardized except for a few cases where misprints are suspected. Contractions such as >qd= and >&= are expanded. The tilde is also expanded and thus the form th., for instance, is spelled out as them or then. As for the initial u and medial v, we have followed the copy-text. Ornamental capitals and display initials are normalized. The long s is silently replaced by the short s. So are >VV= by >W= and >!= by >l=. The initial letter of the first word in conversations, normally capitalized to demarcate the beginning of a quotation, is sometimes in minuscule in 1590 and 1596. We have capitalized such minuscules in the present text. The spacing of words in the quartos is so variable that we have often found it difficult to discern between a one-word form and a two-word form (e.g. himselfe/him selfe, howeuer/how euer, tomorrow/to morrow). In such cases we had to use our own judgement. Compound characters such as ligatures, diereses and accents are retained where they are used in the copy-texts.
In editing punctuation, we basically adhered to the pointing of our copy-texts and accepted their inconsistencies, unless they are obvious errors. In other words, we strictly refrained from standardizing or normalizing punctuation as we did in editing other >accidentals=, for we believe that any attempt to sweep out inconsistencies in these >accidentals= in an old-spelling edition is historically inappropriate and editorially interpolating. Observing this principle inevitably requires accepting what J.C. Smith deemed puzzling to the modern reader and rejected in his edition, such as absence of the comma after vocatives and occurrence of the comma between modifying phrases and nouns they modify.
Padelford 1938 analyzes the punctuation of the FQ. Since most of the 1590 punctuation has been transmitted intact to 1596, his description of the punctuation of the latter is for the most part applicable to the earlier edition, but some noticeable traits typical of 1590 are as follows: (1) The most lightly punctuated of all the early editions, it economizes punctuation and often places no marks for setting off additive, adversative, concessive and relative clauses, and for setting off vocatives. (2) On the other hand, this edition frequently employs the comma for breath between subject and verb, verb and object, and noun and prepositional phrase, particularly at the end of the line, whereas later editions tend to omit these commas. (3) The comma is used for demarcating almost every kind of clause and for marking quotations. In the latter case, 1596 often adopts the semicolon. (4) The parentheses breaking quotations are less often used. (5) The exclamation mark is not used in exclamatory sentences and rhetorical questions but the interrogation mark is used instead. (6) The semicolon as well as the colon is used at the end of complete sentences. The 1590 compositors seem to have considered this mark of intermediate value closer to the colon than to the comma, though the distinction between the colon and semicolon is not clear and they were used to some extent interchangeably. 
In an attempt to formulate principles of editing punctuation, we have re-examined the 1596 changes of punctuation, restricting the definition of obvious errors to the following cases: (1) omission of the period, colon, semicolon, or interrogation mark at the end of complete sentences; (2) use of the comma at the end of complete sentences; (3) use of the period or the colon where sentences are yet to be completed; and (4) use of the period or other marks at the end of interrogative sentences. Though these cases are concerned only with logical or syntactical aspects of punctuation, they are what the compositors of both editions would have commonly recognized as erroneous. The results showed that out of the total of 741 punctuation variants between the two editions, 174 obvious errors in the 1590 edition were corrected and 38 fresh errors were introduced in 1596.
To see the characteristics of the changes made in 1596, we have further analyzed the variants excluding the obvious errors and classified them into three categories established purely from a syntactical viewpoint. In other words, contribution to the interest of logical value is our criterion for judging whether a change is for the better or for the worse. Accordingly, if a change clarifies the structure or meaning of a sentence to any extent, we regard it as improving the original punctuation, and conversely, if a change obscures the structure or meaning, we regard it as deteriorating. A change that has no conceivable significance in clarifying meaning is treated as indifferent. Thus, omission or addition of commas placed for breath is taken to be indifferent, regardless of the consequent effect on the metrical pattern. _
While 137 changes turned out to be improvement, 350 proved indifferent and 42 deteriorating. A further analysis showed that 87 out of the total of 350 indifferent changes involved semicolons replacing commas followed by quotations, coordinate clauses, relative clauses and other subordinate clauses. On the other hand, there are 12 instances of substitution of semicolons with commas without influencing syntactic clarity. It should be also noted that while 103 commas were removed at the expense of pauses for breath, particularly at line endings, 55 additions of the comma were made both within the line and at the end of the line. Thus the indifferent changes reveal the compositors= contradictory behavior. The figures also show that a considerable number of changes served to clarify the syntax. (See Suzuki 1999.) In view of this and the freqent omission of the comma placed for breath, we can say that the 1596 changes in punctuation on the whole reveals a shift from rhythmical to grammatical pointing.
We have confined our emendations of punctuation to obvious errors, adopting neither metrical value nor thought units as criterion for judging whether an emendation is necessary or not, for fear that subjective judgement contingent to these kinds of emendations would be misleading, even if they are meant to improve the reading. In emending errors, we consulted corrections made in the early texts up to 1609 and adopted them if we found them proper, on the ground that early compositors or editors were closer to and more familiar with the contemporary punctuation system. Otherwise, we emended them according to the normal practice of the copy-texts.