This is a reproduction of a library book that was digitized by Google as part of an ongoing effort to preserve the information in books and make it universally accessible.

Google books

https://books.google.com

This is a reproduction of a library book that was digitized by Google as part of an ongoing effort to preserve the information in books and make it universally accessible.

Google books

https://books.google.com

.

° - - . = eed ~~ - — - - es J —_ 2 — - _ - - - _ - a -- —_ ~ —~a-* - ore - == ~——-< —— ae set = ' - _ <a ao - — = - -_ _ — a2 = ee ee — ——< -_— — ———s “= ™ - - = Pe ~ — ~ - = - — _ — — a a - ~~ * ° a - —_— — — = _ - =~ - a - ~ » << - — -< ~ = ~ = ‘ sd - = * _ — - - - - ad _ - i a — _ ~ —_ --—< - il ° - — - ns = . - a — —— es — _ — sill -_ — —— = _ ———— & a ea o —_ — pal -* = — - _——<-— —— - — at - _ a = ns — le —— “* . - a - - —— — ——— i = —_ o— - - ~ = aie a one - - -_ gaan - - — - - - . “ —— sy oo + - —— ? ~~“ ~~ - = an — - — —- ee A _ - — oe * . ea _ - _ — a mm a . ante o_o —_ EM lg eG se eee ee le -- : —— ae — — - = s om: — » - o— - - ee — ~~ -<«+ a ee a — tn — nl oe e _ —_— °-——- - ce ye I hw See - — or as —— Sa ~— — a” — eae 7 —_—- . = naan neat am - _ a _ ® ne - ’

Acesddeqvevereerrreerqesr ye Fete et ]

syenete

ease AT YT TOO AY von

Ppdeh axeeesehed cdesen th ch TRUDE Ce REN Pe here

ee ~— et gig gh ee a Aaa gE

4 te 4-8

y

‘

aii

STEN

———

“Vvorn~ Lae

Uy _ ea

° \ $f : he | ». *« "es - . . A,

a Tran, eEg@igas ape Un,,

ALAS YO _

eK << LULULDLDDD””””””””””~”~”~”~”CO

+1” A

724 OO

The Anglo-Norman VOYAGE OF.St. BRENDAN

BY BENEDEIT A POEM OF THE EARLY TWELFTH CENTURY Edited with Introduction, Notes & Glossary by

E.G.R.WATERS, M.A.

Professor of Romance Languages in the University of Oxford

is

OXFORD At the Clarendon Press

1928

c »

iV san 1

wth VVe

OXFORD UNIVERSITY PRESS LONDON : AMEN HOUSE, E.C. 4 EDINBURGH GLASGOW LEIPZIG COPENHAGEN NEW YORK TORONTO MELBOURNE CAPETOWN BOMBAY CALCUTTA MADRAS SHANGHAI

HUMPHREY MILFORD PUBLISHER TO THE UNIVERSITY

Printed in Great Britain

et-t SD cow

-_

PREFACE

HEN in my undergraduate days I perused this

venerable and fascinating text in the attractive but uncritical edition by Francisque Michel, on every page of which the reader stumbles against some unsolved problem, it became clear to me that there was no medieval French text in greater need of thorough critical treatment, and none likely to yield more valuable results. The project, then presumptuously formed, of supplying this want remained for a long time unrealizable. I count myself fortunate to see it at last accomplished after so many years of delay.

The Voyage of St. Brendan is probably the oldest extant monument of Anglo-Norman literature, except for the Cumpoz of Philippe de Thaiin. Its value and interest as a literary performance have never, I think, been sufficiently appreciated, but its importance as a linguistic document has long been recognized. Dating as it does from the first quarter of the twelfth century, it is a fundamental text for the study of early French, and above all of the Anglo-Norman dialect. Like the somewhat earlier but continental Vie de Sz. Alexis, it has been transmitted to us under peculiarly favourable conditions, in several manuscripts (one representing a later redaction) whose combined testimony enables a critical text to be established with unusual precision. We possess, moreover, its Latin source and two Latin translations. Why, then, has No critical edition been forthcoming till now? The principal

. Teason is not difficult to discern. Over fifty years ago the

Voyage of St. Brendan attracted the attention of the eminent scholars Hermann Suchier and Gaston Paris, who determined (at first, it seems, independently, then in collaboration) to bring out a critical edition. No editors possessing higher qualifications for this undertaking could have been wished for, and no other scholar would have had the temerity to intervene

iv Preface

so long as there was a possibility that they would accomplish it. But their interests were multifarious, and the project never matured. Halfadozen theses, some of them quite meritorious, were devoted to the poem by students at German and other universities ; but however useful they might be for the purpose of checking results, none of their conclusions could be accepted without strict control. The death of Gaston Paris in 1903 and of Hermann Suchier in 1914 left the field open to some less distinguished philologist. Though I cannot hope to have provided a solution of every difficulty, or to have avoided all the pitfalls that abound in this text, I trust I have contributed something to its elucidation. It is perhaps not inappropriate that the edition should, after all, be published in the country where the poem was written.

The establishing of the critical text and the solution of many of the linguistic problems depend largely on a study of the prosody. For this reason the author's versification has been examined in detail at the outset, and has been found to be based, with meticulous accuracy, upon a system (almost certainly imitated from medieval Latin verse) which is in some respects unique in Old French poetry. The analysis of the author's language has attained dimensions which may seem excessive, but are justified by the position of the poem at the fountain-head of Anglo-Norman. The lack of such detailed information as this edition attempts to supply has seriously hampered the study of the dialect, and has been responsible for many of the errors current in existing treatises. The glossary has been made as complete as possible. The Voyage of St. Brendan is the work of an author who possessed, within certain limits, a real command of his language, and is undoubtedly one of those texts of which (as was recently suggested by a committee of the ‘ Société des anciens textes’) the vocabulary and phraseology deserve to be analysed and recorded in full detail. Among the many interesting facts that emerge from the linguistic investigation, one deserves special emphasis: the dialectal features of the poem, though all consistent with later Anglo-Norman developments, show

Preface Vv

at least as close an affinity to the northern and north-eastern French dialects as to the dialect of Normandy. It is possible that the evidence here put forward may reopen the question whether the term ‘Anglo-Norman’ is justifiable. While the main purpose of the Introduction to this book is of course philological, the comparison (in ch. v) of the poem with its Latin source has necessarily involved an appreciation of the literary talent of the author. No attempt has been made here to discuss the Brendan legend and its literary manifestations in general ; this is an undertaking which I hope, after a series of preliminary investigations, to accomplish at ‘some future date.

In the course of my task I have contracted many obligations. Professors E. C. Armstrong, J. Jeanjaquet, T. A. Jenkins, and C. C. J. Webb, Dr. H. H. E. Craster, Dr. E. A. Lowe, Mr. L. F. Powell, Mr. J. R. H. Weaver, Miss J. Evans, and Miss M. K. Pope have all contributed expert advice on points of detail. Mr. E. A. Barber gave kind assistance with the Latin text. The proof-readers of the Clarendon Press have made numerous helpful suggestions. To Mr. C. T. Onions and Mr. A. Ewert I owe a special debt of gratitude for the long-suffering friendliness with which they have read through large parts of the work in manuscript or in proof; to Mr. Onions also for valuable advice in connexion with the Germanic element in Old French; to Mr. Ewert also for obtaining particulars of the York MS. at first hand. The Rector and Fellows of Exeter College gave invaluable aid and encourage- ment by electing me in 1925 to a Research Fellowship, but for which the completion of this volume would have been indefinitely delayed. The Curators of the Taylor Institution have made a generous grant from the Gerrans Memorial Fund towards the publication of this work ; and through the kind agency of Professors Joseph Wright and Sir Charles Firth a similar grant has been made by the British Academy. To all I express my very hearty thanks. It is a cause of keen regret to me that Professor P. Studer did not live to see the publication of the Voyage of St. Brendan, in which he always

vi Preface

took a great interest: only the text and glossary reached their final shape early enough to benefit by his advice. I sincerely hope that the outcome may justify his encouragement and expectations, and continue the tradition of sound scholarship which he himself so honourably maintained in the Romance

field. E.G. R. W.

OXFORD, February, 1928.

CONTENTS

INTRODUCTION I. THe MANUSCRIPTS : : II. AUTHOR AND DaTE P ,

III. VERSIFICATION . : 5

§r. § 2. § 3. § 4. § 5. § 6.

The Metre ,

The Feminine Lines . ; ; The Internal Construction of the Line The Counting of Syllables .

The Couplet . ; : , . The Rimes ‘ ;. ; : ‘

IV. CLASSIFICATION OF THE MANUSCRIPTS

§ 1. § 2. § 3. § 4. § 5.

§ 6. § 7.

The Group AB . ‘ ‘

The Group DE .

AB against DE .

Other possible Groupings .

The Common Ancestors of AB (a),

DE (8) and af (x) The Position of C_ . P ‘ ‘ Conclusion ‘ P ; : ‘é

V. THe RELATION OF THE ANGLO-NORMAN POEM TO THE ‘ NAVIGATIO SANCTI BRENDANI’

VI. Tue Latin Prost TRANSLATION .

VIII.

. THE LATIN RIMED TRANSLATION . . ‘:

THe LANGUAGE OF THE AUTHOR § 1. § 2.

Lists of Rimes .

Phonology (a) Vowels (4) Consonants.

Vill Contents

VIII. THe LANGUAGE OF THE AUTHOR (continued ) § 3. Morphology (a) Declension : (4) Conjugation .

§ 4. Vocabulary § 5. Syntax § 6. Conclusion : : : : ; TEXT . : i «4 NOTES . APPENDIX I The expansion of the feminine lines in the MSS. . ; ‘ , ; s II The spellings and forms of the Anglo- Norman MSS. . : : e : GLOSSARY . ; , ; ;

INDEX OF PROPER NAMES. . . ...

FACSIMILE OF MS. COTTON VESPASIAN B. X,

clxiv clxvili clxxix

clxxxv

FoL. 9" . é ; ; : To follow p. \xxx

ERRATA

Page cvii, note 4. The dissertation by E. Phitzner appeared also in the Zeitschrift fiir romanische Philologie, xxxv (zg7z),

pp. 31-66. 38, vartant tov. 681. Read E E dencensiers. 43, second line of Latin text. For aera read aeris.

62, variant fo v. 1163. Add E before Molt longement. 68, variant tov. 1279. Delete bracket after D quen li. » 93. Zhe variants to vu. 1820, 1821, and 1822 are wrongly

attributed to vv. 1819, 1820, and 1821 respectively.

as _S-

INTRODUCTION

I. THE MANUSCRIPTS

IVE MSS. of the Voyage of St. Brendan are extant, one, however, being but a fragment. The designations used by Bartsch’ are here adopted in preference to those of other scholars.?

A. London, Bntish Museum, Cotton Vesp. B. x (1), fo. r-rr r°. Attention was first drawn to this MS. by Francisque Michel * and Abbé de la Rue,‘ and a detailed description was published by the former.* More recently it has been described by H. Suchier * and H. L. D. Ward.’

The MS. is bound in one volume with two other MSS. of later date. It is composed of twenty-three vellum folios, the latter part being occupied by the Latin Wavigatio Sancti Brendani (fo. 11 v°- 21 r°, col. 1) and some shorter Latin texts ; fo. 23 is blank. All the texts are plainly and regularly written, in handwriting (the same throughout) of the thirteenth century, probably of the middle or second half of the century.* The Anglo-Norman text is in double

1 K. Bartsch et A. Horning, La langue et la littérature frangatses, Paris, 1887, col. 69.

3 Concordance: ABCDE = Vising C. BDA, Birkenhoff L.O.P, Hammer and Wien LAOYP, Brekke LTOYA, Calmund LAOYArs.

3 Le Roman de la Violette, Paris, 1834, p. xlvii.

* Essass historiques sur les Bardes, les Jongleurs et les Tronvéres normands a anglo-normands, ii, Caen, 1834, pp. 66 ff.

® Collection de Documents Inédits sur [ Histoire de France. Rapports au Ministre, Paris, 1839, p. 165.

® In Boehmer’s Romanische Studien, i, Heft 5, Strassburg, 1875, p. 554.

1 Catalogue of Romances in the Department of Manuscripts in the British Museum, ii, London, 1893, p. 541.

* Dr. E. A. Lowe suggests the middle of the thirteenth century. Mr. J. P. Gilson, Keeper of MSS. at the British Museum, has assigned the MS. to the end of the century. Fr. Michel dated it from the twelfth or beginning of the thirteenth century; T. D. Hardy (Descriptive Catalogue ..., i, p. 160) from the thirteenth century; Suchier from the end of the twelfth century. All philologists who have dealt with the text since Suchier, not excepting the present writer (cf. Mod. Lang. Rev., xxi, p. 391), have copied his statement, and have treated the forms of this MS. as representing twelfth-century Anglo- Norman; yet both the handwriting and the vignettes undoubtedly belong to the following century. The grounds on which Ward assigned this MS. to the fourteenth century, and F. J. Tanquerey (L’£volution du Verbe en® Anglo-

$351 b

x Introduction

columns of forty-four lines each. On the fly-leaf a modern hand has written: Elenchus hutus hbrit. Vita sa Brendant rythmo Gallicano vetert etc. At the top of fo. 1 r° is written in sixteenth-century handwriting ; vita Sa brédani gallice. Something has been erased between the columns of fo. 1 r°. Thirty-six verses have large orna- mented initial letters, the first red and blue, the second and third red, then alternately blue and red, except that two blue initials succeed one another at vv. 309 and 371, and again at vv. 1493 and 1511; the red are decorated with blue and the blue with red flourishes. A wrong initial (Q instead of A) has been inserted in vv. 1067 and 1735.’ In each line, except those before which a large inittal stands, there is a space between the first and second letters. This is the completest and best MS., retaining most fully the archaisms of the original, though showing in its spellings abundant traces of Anglo-Norman developments of the thirteenth century.* Six lines only are missing : 477-8, 849-50, 1243-4. The diplomatic transcription published by H. Suchier® is on the whole accurate, and renders it unnecessary to enter into details here concerning the word-division and the scribe’s corrections. As regards the expansion of contractions a few explanatory notes are needed.‘ The sign £ represents either Jer (e.g. pernans 37) or par (e.g. parfuns 1127) ; hence agceut 335 may be transcribed either aferceut or aparceut (cf. v. 1141). The sign ? 1s used at the end of a word for -us, e.g. plus 26, precius 1077, and (through Latin influence) also for -es in the first person plural forms sumes 521, 535, 542, 719, fumes 522, asemblames 731, amames 732;°* it is not used in this text for initial or internal con, and Suchier’s reading tacoznes = Jacco(n)ines in v. 839 is erroneous.

francais, Paris, 1915, p. 67, ef passim) to the year 1167, or 1160 (tbrd., p. 426, note), are unknown to me. There is no evidence other than palaeographical.

1 Cf. note to v. 1735. 2 Cf. Appendix IT.

8 Op. at., pp. 553-88. The following are the principal corrections (adopting, for this note only, Suchier’s numbering of the lines) : 2 sa/drat without accent, 215 seignent, 233 bestrer, 577 ad] as, 597 purcusent, 837 tacoines, 883 deuolat) flotet, 981 uent) nent, 1047 cOmo*t, 1138 fldmans, 1174 seif without accent, 1175 Antre) Atnxe, 1230 q"5, 1237 ses] les, 1238 les] ses, 1334 Que altre, 1385 atte] acce, 1409 beisire, 1410 quire or giure, 1470 lusdi, 1498 terce, 1508 add ert after est, 1542 cors, 1579 peisuns, 1599 stu or sus, 1676 Fatters, 1709 Lai nur or Lat mir, 1741 ni ad, 1742 nerechrie, 1747 replenis, 1790 requais corrected to requets, 1834 plusur.

‘ The contractions in the MSS. of the Brendan were inaccurately classified by M. Wien, Das Verhdlinis der Handschriften der Anglonorm. Brandanlegende, Diss. Halle, 1886, pp. 7 ff. 8 Cf. Appendix IT, § 6d.

The Manuscripts xi

An apostrophe may represent either re ( p’ntf = prent 145; but /ez* = lefre 11, etc.) or er ( port = porter 180, mat = merat 338, fre = lerre 483, etc.). A form such as 'f’nent 802 may probably be expanded as prenent ; pernent would have been written pent. Que is represented by ¢° (e.g. v. 15), 97; (e.g. desque 220) or 7 (e.g. guerant 363). st is represented by # in v. 1181, by 2 in v. 1349, but usually written in full. Note the contractions xfa/ = cristal 272, thit = Jesu 574, etc., st = sunt 1034, d5 = Deus 1589. It is not clear what the scribe intended by the comma in /id’e 100. The text of this MS. was also reproduced, but in an inaccurate and unsatisfactory manner, by Fr. Michel,’ whose use of modern punctuation and diacritical signs scarcely succeeded in making the text intelligible.

The use of accents in this MS. was satisfactorily dealt with by K. Lincke.* The scribe used an acute accent very frequently to separate ¢ from an adjacent m, n, or uw (e.g. seins 19, ullain 163, suiurnerent 327) ; also on the numeral .fff. (vv. gor and 998), and in a few instances on # between other letters (e.g. mairen 174, crfe 1252, sedeir 1776). In sé¢t 359, entr¢é¢ 632, and guuardé? 839 the accents probably indicate that the consecutive vowels belong to different syllables. At the outset the scribe began placing an accent on post-tonic ¢ (diuind 2, terr¢ 3, guerré 4), but he abandoned this practice after the fourth line; ss¢#¢ 76 is undoubtedly an error for xf##. In nenémout (= ne ne m'out) 1588 the accent is used to pick out a short monosyllabic word. A double accent has in most cases been placed over the words ¢4, ¢6 and i¢d, its function being to indicate that the letter ¢ has not the velar pronunciation usual before a and a, but some other sound (¢s or 4). All these uses of the accent can be paralleled from other early French, especially Anglo-Norman, MSS. The apparent accent on ua/drd¢ 2 is in reality merely the insertion sign for the 7 of refne in v.1. The accent on the ¢ of #¢ 1800 is unexplained.

B. Paris, Bibliothéque Nationale, nouv. acq. fr. 4503, fo. rg v°- 42 r°. This MS. has had a chequered history.’ In the seventeenth

1 Les Voyages Merveilleux de Saint Brandan, Paris, 1878. Cf. G. Grober, in Ltschr, f. roman. Phil, iii, p. 133.

% Die Accente im Oxforder und im Cambridger Psalter sowie in anderen alifra. Handschriften, Diss. Erlangen, 1886. Cf. W. Buhle, Das C ims Lambspringer Alexius, Oxforder Roland und Londoner Brandan, Diss. Greifswald, 1881, pp. 47 ff.

> Cf. L. Delisle, Bibtothéqgue Nationale. Catalogue des Manuscrits des fonds Libri et Barrois, Paris, 1888, Introduction.

xii Introduction

century it was in the possession of the Lesdiguiéres family. In 1716 it was bought for the Abbey of Marmoutier near Tours, becoming no. 239 in the Abbey library, and at the Revolution passed into the municipal library of Tours. Jn 1842, or shortly after, it was stolen from this library by the notorious ‘ bibliophile’ Libri, and in 1847 was sold by him to the Earl of Ashburnham, in whose library it was numbered Libri 112. Having been sold in 1884 by the Earl’s successor to the Italian government, the MS. remained for a time in the Bibliotheca Mediceo-Laurenziana at Florence, but was recovered in 1888 by the French government and placed in the Bibliothéque Nationale. The first accurate description published was that by G. Paris,’ and it has since been described also by L. Delisle,* H. Omont,® and H. Calmund.*‘

The MS. consists of seventy-four vellum folios, and contains besides the Brendan a poem on the Assumption of the Virgin Mary, by Herman de Valenciennes (fo. 1-11), the Vie de St. Alexis (fo. 11 v°- 19 v°), the Vie de Ste. Catherine translated by Sister Clemence of Barking (fo. 43-73), and a French translation of a bull of Alexander IIT in favour of the Templars (fo. 74). The handwriting dates from about 1200 (not the middle of the twelfth century, as originally stated by G. Paris). Except for vv. 1-81, which are by a different scribe, the Brendan is written in the same hand as the preceding and following poems. Except for vv. 73-81, the poem is written out like prose, but with a stop indicating the end of each verse. Each page contains one column of from twenty-three to twenty-eight lines. The heading Jnapit uita sanct} Brandanj and the first line of the last page (Ve sunt hatties sul li parent Ainz sunt) are in red ink. Twenty-two large initials are written in red, three in blue®; the first is missing.

G. Paris drew attention to the ignorance and carelessness of the principal scribe of this MS., who betrays his insular origin by his use of Anglo-Norman spellings and by the little heed that he pays to grammar or metre. The text of the Brendan is, however, superior to that of the A/exts, and has fortunately not been touched by the

1 La Vie de Saint Alexis, Paris, 1872, p. 3. Cf. Romania, xiii, p. qo1

3 Op. at., p. 116 (no. Ixx).

® Bibliotheque Nationale. Catalogue Général des Manuscnits francais. Nouvelles acquisitions francaises, ii, Paris, 1900, p. 199.

* Prolegomena su einer kritischen Ausgabe des dltesten frans. Brendanlebens,

Diss. Bonn, 19032, p. 5. 5 So Calmund ; the colours cannot be distinguished in my facsimile.

The Manuscripts xili

medieval ‘ corrector’ who did so much harm to the preceding text. Although much inferior to A, B forms a very valuable check upon that MS., to which it is fairly closely allied.' But the principal grievance remains, namely that the scribe, losing patience with his task from time to time, has omitted a great many words, verses, couplets, and longer passages. No fewer than 169 verses of the Srendan are missing, the great majority of these belonging to the last 300 lines of the poem: wv. 12, 221-2, 395-6, 423-4, 6or—4, 617-18, 651-2, 661-2, 665-6, 669-70, 679-88, 7143-4) 843-4, 1§24, 1526-40, 1545-6, 1549-52, 1555-60, 1563-6, 1577-80, 1583-90, 1603-10, 1615-24, 1627-8, 1631-48, 1653-8, 1665-8, 1715-18, 1741-52, 1755-60, 1763-4, 1771-2, 1775-90. ‘Two lines, of the feeblest kind, have been added, following vv. 1813 and 1814 respec- tively. Vv. 1325-6 have been transposed.

Both scribes employed the usual contractions, though the second used them very sparingly and nearly always for the sake of fitting a word into the end of a line. Zsé, usually written in full, appears as éinv. 1593. Que is written 7 in v. 529, sometimes also g;. Note deurt = devrunt 62, di = dunt 174, dé = dune 453, uer*t = verrunt 1106, brand’ = Brandans (?) 1255, st’ = sunt 1496. In v. 335 sapcut probably represents s'aperguz; cf. v. 1141. In order to cancel a letter, the scribes placed a dot either above or below it.

The acute accent was freely used by both scribes. It is commonly but not invariably placed on #, not only when next to m, #, and x, but also in other positions (e.g. ofe 100, saphfre 1071). It is used regularly when two identical vowels come together (e.g. uéér 55, embleé 311, udéit 313, séélus 647, abddte 888, luurs 1685), and less regularly when v (written 2) is followed by u (e.g. aus 119, akims 650, sauured 705, uunt 1629); but in entrée 251, branche 498, and hkdan 1320 only the first vowel bears an accent. It is placed on short monosyllabic words: é 82, etc., d 145, etc., # 95, 234, etc., ¢o or 6 87, 93, etc., 6¢ 243, etc., cd 1116, 16 1243, 266 1277, /é (def. art.) 1134, ut (< habuit) 278, fd 292, fad 1835, dt 1702; and by analogy with these, also on the prefix a- (dasma 1054, dclose 1111, dtendisse 1554), On #6 102, tced 1507, audc 181, t/éc 413, etc. In the following instances it marks the tonic vowel: a4é 152, sathdn 310, clarté 1697, enarté 1698, oréd 1818. Further, the accent is freely used on certain diphthongs, in a manner which is difficult to explain, but which lends

3 B alone has preserved the correct reading in a number of instances, e.g. vv. 716, 798, 1172 (but cent for oent), 1348, 1350 and 1590.

XIV Introduction

little support to the theory! that its function was to discriminate ascending and descending diphthongs. Thus the first scribe (down to v. 81) wrote déu 19, {rt 38, 46, sf 58, lik 63, orguil 67, léf 69, Sui 70, fét 71.2 The second scribe used it, not for #, ¢#, s, or ut, but for o¢ (cber 142, trotuent 261, 294, 645, troéue g60, bées 1150, hocm 1521), ue (erées 637, cumtel 1051, suéns 1814), ou (erréut 322, bussent 657, tues 1013, poues 1014, pourent 1175, ofrent 1200, quidéue 1283) and eu (leés<lupos 1288, fea or feu 466, 758, 759). That the accent in these instances bore any relation to either the diphthong or the tonic syllable is difficult to believe, seeing that we find it also where the two vowels belong to different syllables, sometimes even on the unstressed vowel : pardis 49, uele 103, 1804, meus <*movuistt 776, cles 941, sbe 1234, celée 1281, pot past part. of paistve 1582, rué 1662. Some of these examples, as well as ue/t 325, uent 740, Suggest that the scribe was inclined to place an accent on # (vocalic or consonantal), as on #, almost as a matter of routine. The accent on ¢a, ¢o, igo is placed over the following vowel as often as over the consonant, and appears to have no connexion with the pronunciation of the consonant.

The text of this MS. has not been printed, but it has been repeatedly copied, and has been utilized by those who have dealt with the poem from Hammer and Brekke onwards.

C. Oxford, Bodleian Library, Rawl. D 913 (formerly Rawl. Misc. 1370), fo. 85. H.Suchier® was the first to draw attention to this fragment. Particulars of the illegible passages were added by R. Birkenhoff. A brief account of the fragment and of the volume in which it is preserved (a collection of fragments, mostly in French and English, bound together about 1861) is given in the official catalogue of Bodleian MSS.°

It consists of a single vellum folio, which has been torn away from some object to which it formerly adhered, with the result that the writing on the middle and lower portion of one surface (recto) has been removed or rendered illegible. The use to which it was put

1 Put forward by A. Brachet, in Revue critique @histotre et de liti¢rature, 1870, ii, pp. 254 ff.

2 es&ésl 68 presumably stands for escuil.

3 Op. at., pp. 564-6.

4 Ucher Metrum und Reim der alifrans. Brandanlegende, Marburg, 1884, p. 6.

5 Catalogs Codicum Manuscriptorum Bibliothecae Bodleianae partis quintae fasciculus quartus, Oxford, 1898, cols. 136 and 141.

The Manuscripts XV

(perhaps as part of the cover of some other MS.) and the circum- stances in which it was rescued are unknown. The writing is in two columns on each side. Col. 1 r° contains the last thirty-two lines of a rhyming Anglo-Norman exhortation to repentance.’ Col. 2r° and cols. r and 2 v° contain wv. 1-310 of the Brendan, two verses, however, having dropped out between the middle of v. 269 and the middle of v. 271. Each line of writing contains a couplet, and each column fifty-one or fifty-two lines. A large initial D in red ink marks the beginning of the poem. There is a space between the first and second letter of each couplet except the first two. The handwriting is of the first half of the thirteenth century.? The passage has been written by two scribes, the second hand beginning with v. 165. From Vv. 35 to v. r04 only a word here and there is legible, usually only the last word of a couplet; vv. 27, 29, 258-9 are also partially illegible.

It is regrettable that this MS. should have been reduced to its present condition, as its text was a very fair one and did not belong (so far as can be judged) to either of the other groups of MSS: In several instances C alone gives the correct reading.‘ The spellings betray the Anglo-Norman origin of the copyists. Accents are used solely on the letter % Both scribes made lavish use of contractions, few of which require special explanation. The sign g represents either Jer (e.g. pernanz 37, perdent 213) or par (e.g. parlerent 112, paret 276). st appears as + (v. 18) and as é (vv. 242, 281, 286). Que is represented by ¢° or 7 (both in v.15). Note cf = cs 14, nf = nul 236, 254, 302, a's = Deus 242, st = sunt 255, 261, panet = prennent 302, 307. In v. 186 Jon? (= Lat. bonus) has been interpreted as dons, and in v. 191 sum? as sumes.

Suchier gave the variants of this MS. from a collation made by E. Stengel. Closer study has made it possible to introduce a number of corrections and additions, the more important of which will be found in the variants of the present edition or in Appendix IT.

D. York, Dean and Chapter Library, 16 K. 12, part 1, fo. 23- 361°. The existence of this MS. was mentioned by T. D. Hardy,® but so briefly that Suchier* was led to believe the York and

1 Printed by E. Stengel in his glossary to the oldest French monuments, Ausgaben und Abhandlungen, i, Marburg, 1882, p. 173.

* Teste Dr. H. H. E. Craster. * Cf. ch. iv, § 6.

$ e.g. vv. 177, 246, 282, 292.

® OP. at., p. 554.

xvi Introduction

Cottonian MSS. to be identical. Fuller particulars of it were given by J. Vising.’

The volume, which is mainly in quarto, consists of thirty-six folios, size 203X137 mm. Fo. 1-22r° contain a collection of fables by Marie de France and an unknown author.? Each pagé contains two columns of thirty-four verses. The text of the Brendan was written by a single hand ; with the beginning of a new quire at fo. 25 the writing changes slightly, but gradually reverts to normal. From the same point onwards the ink used is a decidedly deeper black than in the previous quire. A note signed H. P. (i.e. Hermann Pauly) on the fly-leaf of the MS. states: ‘This volume is of the writing of the latter part of the XIIIth century ’; but the wnting has been assigned with more justice by Vising and Warnke to the first half of the century. The heading Jncpit utta sci Brendané is written in red. The poem begins with a large initial D, ornamented in red and green. Forty-two other initials are enlarged and written in red, but devoid of ornament with the exception of Q in v. 103, the tail of which is extended in a scroll to the O of v. 123. In addition, small black initials D and Z in wv. 825 and 1043 respectively are encircled by large initials done in red. Except in vv. 1, 781, 851, and 1809 the large initials do not extend beyond the space occupied by one line. In v. 18 a supernumerary g before grvané is writtenin red. An acute accent is used in a few instances only, to distinguish # from a neigh- bouring m, #, or u (e.g. deufne 2, maint 115, enemf 992, deduift 1644), and to correct # to # in suf 1277 (originally written sun).

The scribe was somewhat unmethodical in his use of contractions. Est is usually represented by + (vv. 128, 378, etc.), but by @ in v. 1350. For gue we find ¢ (v. 23), 9; (v. 192), g2 (v. 866) and ¢: (v. 1830), with various minor modifications of these signs. Both ser (pernes 399, superbe 525) and par ( part 233, parfunt 369) appear as fp. The same signs are used for ve and er, thus pmeraine = premeraine (cf. £ for Lat. prae) 952, but ube = verbe 532, conisu = conversum 720, Cf. also mnoc = Mernoc 85, 98. Similarly oind'e = oindre 177, but A’mite = hermite 1547. It is not clear whether p? 907 should be expanded as puis or pus (cf. 4? = ous 329, p19 = plus 908, etc.). Note angles) = angele(s) 521, 1732, nl’e = mule 1518. Apart from these instances, the interpretation of the contractions offers no difficulty.

1 Etude sur le dialecte anglo-normand du XIP siecle, Diss. Upsala, 1882,

pp. 18f.; cf. Calmund, Prolegomena .. ., p. 6. 3 Cf. K. Warnke, Die Fabeln der Mane de France, Halle, 1898, p. iv.

ee ee

ee eee, lO ee

The Manuscripts xvii

From the point of view of textual correctness this is the least satisfactory MS. The Anglo-Norman scribe did his work hastily and carelessly, and made a large number of gross grammatical, metrical, and other blunders. His fondness for replacing final ¢ (or @) by s (e.g. abes acc. sg. 13, verfes acc. sg. 14, Keues = level 317, duis ind. pr. 3 sg. of dutre 694) has resulted in constant confusion of number, case, and form, and produces a very disagreeable impression. His capricious attempts to expand the feminine lines to a normal length by adding a syllable’ have played havoc with the text; he has altered feminine rimes to masculine (vv. 559-60, 839-40, 887-8, 1099-1100, 1467-8, 1625-6, 1679-80), changed the sense (e.g. VV. 103, 439, 1657), or substituted lines presumably of his own composition (e.g. vv. 17-18, 239, 902). The large number of verses with syllables too many (e.g. 452, 588, 863) or too few (e.g. 320, 760-1, 1134) shows how defective was his own metrical sense. Words which he did not understand became hopelessly corrupted (e.g. a haloecs 173, daumunde 881, cestirez 1228) or were avoided by means of drastic alterations in the text (e.g. v. 286, avoiding repetition of mestiers; vv. 461-2, avoiding raps; vv. 1321-2). Other passages of which the meaning has been destroyed by gratuitous and often childish alterations are vv. 14, 188, 942, 1332, 1360, 1454, 1532, 1565, 1755-6, 1782 and 1784." Twelve lines are missing: vv. 1023- 4, 1179-80, 1395-6, 1717-18, 1723-4, 1745-6. To these might be added wv. 809, 1473, and 1581, but compensatory lines have been introduced after vv. 810, 1474, and 1582 respectively. Nine lines have been added: two after 188, two after 408, one after 549, two after 584, two after 1198. Vv. 1563-4 have been placed after 1554. In three couplets the lines have been transposed : 499—500, 1693-4, 1751-2. In spite of these many defects, D has great value as the Anglo-Norman representative of the second main group of MSS., and is the basis of the critical text when A and B both fail us (e.g. in wv. 477-8, 652, 1549).

E. Paris, Bibliothéque de l’Arsenal, 3516 (formerly BLF 283 fo.), fo. g6r° col. c-100v°, This text of the Brendan has been men-

1 Cf. Appendix I.

% The possibility that some of these corruptions were due, not to the scribe of D, but to a predecessor, is discussed in ch. iv, § 5; in most cases it is not possible to assign the responsibility with certainty.

8 D alone has preserved the authentic reading in vv. 478, 1ogo (but Cist for Cif) and 1576.

3351 Cc

XVIli Introduction

tioned, with more or less detail, by A. Jubinal,' H. Suchier,® Fr. Michel® and many later writers. A diplomatic transcription, with brief introductory remarks and palaeographical notes, was published by Th. Auracher.‘ Descriptions of the whole MS. have been given by Monmerqué et Michel,*§ by Le Roux de Lincy ‘ and in the official catalogue.”

The MS. is a large volume of 356 vellum folios, containing sixty- three different items of very unequal length. A vandalistic hand has removed most of the miniatures, and in some instances whole folios ; owing to these mutilations the numbering of the folios has varied. The first three folios, which were doubtless added on completion of the volume, contain a French calendar, health rules for each month,® a table of dominical letters, and an index to the whole volume. The table is for the century 1268-1367, and as such calculations were of little value in the Middle Ages save for the future, one may conclude that the volume was completed in the year 1267 or 1268. The thirteenth-century handwriting agrees with this date. The titles are done in red, the initials in red and blue. The Brendan is preceded by the Vie de St. Julien and followed by the Vie de St. Gregotre. It is written in columns of verse, each page containing four columns of about fifty lines each. Miniatures have been cut out from the top right-hand quarter of fo, 96 r° and the bottom right- hand corner of fo. roo v°; in consequence of this, vv. 73-92, 123-42, 1503-17, and 1554-70 are missing. The original prologue of eighteen lines has been replaced by a new prologue of four lines. Twelve verses have been omitted by the scribe: 168, 269-70, 373-4, 1143-4, 1256, 1582, 1644, 1783-4. Verse 1099 is also missing, but a compensatory line has been added after v.r100. Fourteen other verses have been added: two after v. 276, two after 278, two after 466, two after

1 La legende latine de S. Brandaines, Paris, 1836, p. vi.

3 Op. cit., p. 553>

S Les Voyages Merveillenx..., p. xxiit; cf. Le Roman de la Violette, p. xlvii note,

“In Zéschr. f. roman. Phil., ii, pp. 438-57.

8 Lai dIgnaures, Paris, 1832, pp. 35 ff.

® In his list of MSS. containing the Roman des Scpt Sages, following the Essat sur les fables tndiennes by A. Loiseleur-Deslongchamps, Paris, 1838, Pp. XXxix,

7 Catalogue Geéencral des Manuscrits des Bibliotheques Publiques en France. Paris. Bibliotheque de 0 Arsenal, iii, Paris, 1887, p. 395.

8 Published by W. Foerster in Zéschr. f. roman, Phil,, i, p. 97.

The Manuscripts xix

656, and six after 1840. Verses 657-8, 1257-8, and 1767-8 are transposed.

Auracher’s edition makes it unnecessary to enter into minute palaeographical details.! As regards contractions, we may note that identical signs are used for er and re ( Zp = gerpi 27, aps = apres 1529), for ev, re,and ter (alum’ = alumer 326, fe = tere 440, p’nez = prenes 296," ap’sse = apresse 1006, nag’ = nagier 216, volen?’s = volentters 965), for ur and or (escripture 23, aportoit 1571), and for us, uis and os (vend? = vendus 1300, p? = puis 705, n? = nos 740). Jesu is represented by jh’u 574, thu 1243, he 1247, and sAS 1251.

The text furnished by E differs considerably from that of the other MSS. Anglo-Norman rimes and forms have been avoided, the peculiar word-order of the original has been made normal, the vocabu- lary, grammar, phraseology, and metre have been modernized. In some cases these alterations have been made with very fair skill, in others (e.g. vv. 318, 332, 888, 1654, 1718) they have ruined the sense. In particular the abnormal feminine lines have been systemati- cally lengthened by the addition of a syllable, and this has sometimes entailed considerable alterations (e.g. in vv. 148, 178, 1481-2). Nearly all scholars who have dealt with this version have rightly judged it to be a modernized form (Fr. Michel: a rzfacimen/o) of the original twelfth-century poem.* The process of rewriting the poem in continental French of the thirteenth century has altered its aspect ;

1 The following corrections must be made in Auracher’s transcription (adopt- ing his numbering of the lines) : 12 sera] sara, 22 b°ndans, 110 saint} saut, 399 hee| he, 410 & | de, 548 aremusers, 564 serott, 567 noel, 607 eus| eust, 649 It) le, 721 donc, 815 oitaues, 854 enseillj| enfaillj, 859 Qis, 866 de) des, 874 Car des] Gardes, 88a fst, 883 bratst corrected from brant, 885 fist, 923 avinerti, 933 lotr} lor, 1060 motrece, 1065 lorcors, 1174 pelles| pelfes, 120a platd*, 1238 cruels, 1282 Q', 1296 horibles, 1297 sullés, 1298 pullés, 1321 A} Al, 1361 tant) caut, 1419 saint) saut, 1456 se] le, 1486 pessost] peuoit, 1523 sotorer] sororer, 1527 “nj Je| wnsyle, 1543 sotorneront corrected from sosornerent, 1549 lois, 1572 stgle is followed by part of a letter (perhaps an incomplete ¢), 1592 fast] fats, 1607 tacincles, 1627 heurt| heut, 1677 samal, 1728 retenemt. The editor was also inconsistent in his transcription of v and #, of J (given as J or J) and of the different types of s; and his separation of words frequently disagrees with that of the MS.

2 Cf. prenent 68, prenes 459, &c., written in full.

® For proof of this see ch. iii, § 2. The exception is F. J. Tanquerey, who (in his L’£volution du Verbe en Anglo-francais, pp. 5, 16a, 310, &c.) frequently treats E as representing the original form of the poem, and A as the remanie- wernt, but gives no justification for this attitude.

XX Introduction

but the changes, except in the feminine lines, are mostly superficial and can easily be discounted. Even the metrical alterations are often slight, consisting merely in the addition of a monosyllable, the substitution of a dissyllable for a monosyllable, or the transposition of words. The text which the reviser had before him seems to have been a tolerably good one, of Anglo-Norman origin. E therefore gives valuable aid in the construction of a critical text.’ It also offers considerable linguistic interest, in that it shows what words and constructions were regarded by the Picard reviser as obsolete or otherwise unacceptable ; thus cis/e 94, ceste 224, beitrer 233, n’enteins 332, ne frat faile 367, umeit 803, suduines 816, flaistre 1109, debardat 1278, agretes 1498, and many other words and expressions have been abandoned, while in vv. 344, 1007-8, 1468, 1679-80, etc. the word- order has been modernized.

The modernizing of forms which has been carried out in E requires no special explanation ; o# for e#, eu for free tonic 9 (in some instances), the vocalization of / before consonants, the dropping of final un- supported 4, the dropping of final ” in jorn, etc., analogical sien for suen, and analogical -off for -ouf in the imperfect indicative of -er verbs are the most conspicuous features. Closed a is represented by o or ou, not by w, and the ending of the rst pers. pl. is usually -ons or -on. A few words must be said, however, about the dialect. The Brendan of E, like a great many other French texts of the thirteenth century, exhibits numerous Picardisms. Final z is in nearly all instances reduced to s. Glide consonants are absent from volra 72, avenra 196, venredi 1405, etc.; hence we find also the Picard preterite 3 pl. forms prisent (for pristrent) 37, fisent 450 and misent 596. The Picard fondness for metathesis of ve appears in merkerdt for mercredt 1375, bertece for bretesche 1678, and the future 3 pl. rekerront for recrerront 1105. The treatment of palatal con- sonants is that of Picard : ¢ remains unaltered before Lat. a (caste/ 267, blance 388, aes 491, cose 505, trancans 944, etc.); before ¢ and # it frequently appears as ch (chierge 326, rechotvre 358, cherf 390, chaine < cena 831, chessent 1336, etc.), and even when written ¢ was doubt- less pronounced like ch (cf. Herc < tertium 592, 1091). The lowering of ¢ to a before vocalized 7 (in aus, ax <illos 254, 1024, solaus 1761, paraus 1762) is dialectal, though not exclusively Picard. Jn/er for enfer 65, fu <focum 758, etc.,and waumir for vomir 1427 are Picard forms.

1 E alone has preserved the original reading in vv. 187 (probably), 78a, 786 and 918,

The Manuscripts xxi

The Picard fem. nom. sg. form of the definite article appears in 4s onde 1045. The possessive vos, vo for vostre, levelled on the accus. pl., is found in wv. 878, 882, and 997. Dialectal verb-forms are sara for savra 26, aras for avrai 1436, cremir for cremetr 928, and the imperfect subjunctives osaise 1243, criaise 1244, and fuisse T400.

All the forms just mentioned might be attributed to the scribe. Is it possible to determine the dialect of the reviser himself?! A study of his metre and rimes makes it clear that he sought to obliterate Anglo-Normanisms. On the Continent it was not usual to rime 9, # (from V. Lat. 9) with # (from V. Lat. «); hence the reviser replaced poir : otir 361 by paor : socor, uns : cumpaignuns 643 by raisons : conpaignons, made similar changes in wv. 925-6, 1179-80, and 1705-6, entirely rewrote vv. 599-600 in order to avoid the rime purcusent. usent, and in wv. 1123-4 substituted crusts : pieus for cru{i)s : pu(i)s (one of his least successful emendations). In the latter part of the text he sometimes left such nmes untouched (vv. 1213-14, 1499-1500, 1525-6, 1685-6). The contracted imperfect subjunctive forms ous? 917, doust 1708 (both monosyllabic), sousum 765, ousum 766, pouse 1579, and the preterite 2 pl. ousées 1121 (all dissyllabic) were replaced by the standard dissyllabic and trisyllabic forms respectively, with other alterations to suit the metre when necessary ; oussent 657 and poust 1658 were evaded, other parts of the verbs being substituted. The contracted futures ¢ruoras etc. 246, 412, 433, 586, 1772 and /frai etc. 426, 553, 878, 883, 924, 1044 were lengthened by a syllable or replaced by the present tense. The rimes portout: out 1139 and afortout : out 1571 (combining imperfect and preterite), though not exclusively Anglo-Norman, were no longer permissible in Central or Eastern French of the middle of the thirteenth century ; they were therefore replaced by imperfects in -oit. In vv. 1497-8 the reviser rimed vores (0f<¢) with sores (t= au + yod) contrary to the practice of the author. It is clear, then, that he wrote in continental French, not in Anglo-Norman or even continental Norman. Finally, it is possible to point out a few forms introduced by the reviser, guaranteed by the metre, which defi- nitely belong to northern French dialects, viz. the 1 pl. forms estommes (of doubtful meaning) 522, avomes 541, 764, conversonmes 720,and vivomes 763, and the futures prenderés 368, esteras 425,

} The reasons for regarding E as a copy of the revision, not the actual work of the reviser, are stated in ch. iv, § 5.

XXil Introduction

esterés 863 (all three trisyllabic, with intercalated ¢). We may there- fore conclude that the reviser, as well as the scribe, was probably Picard.'

II. AUTHOR AND DATE

The oldest MS. (B) of the Voyage of St. Brendan dates from about 1200, and the sole MS. of the Latin prose translation ? is assigned to the same period. An investigation of the language of the poem ® leads to the conclusion that it was most probably composed in England. . For further evidence as to date and author we have to rely upon the prologue (vv. 1-18), in which the author mentions his name and the destination of his work. There can be no reasonable doubt as to the authenticity of this prologue. It is given by all four Anglo-Norman MSS., including the fragment C which perhaps does not belong to either of the principal groups.‘ That the Picard remanieur, whose work is reproduced in E, should have sub- stituted a different prologue (four colourless lines in which the reciter makes known his subject) is not surprising: the original dedication would have been of little interest to a continental audience of the thirteenth century. The Latin translators were naturally at liberty to adapt the work to their particular purposes. The author of the prose version preferred a pious introductory paragraph, involved in style but hackneyed in ideas. The author of the metrical version substituted a lengthy new dedication, filled out with moralizings and personal explanations. In versification and language vv. 1-18 of the Brendan agree perfectly with the rest of the poem.

The prologue may be rendered as follows: ‘My lady Aal/iz the queen, through whom divine law will prevail, earthly law grow stronger and so much warfare cease, on account of the prowess of Henri the king and through the wisdom which will be in thee, the apostoiles danz Benedeiz greets thee a thousand and a thousand times. What thou didst command, that he has undertaken ; to the best of his ability he has put into writing in the Romance tongue, as thy command was, the story of the good abbot Saint Brendan. But do

1 He altered the wording of vv. 203-4 and 659-60 in order to avoid the rime Brandan(s) : main(s), though he might have used the form Brandain as in v. 164. The reason for replacing descent: tent 1011 by avolant: tent, combining an and en in rime contrary to both Norman and Picard practice, is not clear.

3 Cf. ch. vi. 3 Cf. ch. viii, § 6. 4 Cf. ch. iv, § 6.

Author and Date XXxili

thou preserve him from being derided, since he says what he knows and does what he can. Such a servant should not be blamed ; but when a man can, and will not, it is right that he should greatly suffer for it.’

Since attention was first drawn to the poem by Francisque Michel and Abbé de la Rue,' there has never been any doubt that the personages mentioned in vv. r and 5 must be identified with Henry I, King of England (1100-35) and his second wife Adeliza,* daughter of Duke Godfrey of Louvain. No other royal pair having these names 1s available in the twelfth or late eleventh century. From the use of the future tense in vv. 2-6 we may conclude that the dedica- tion to Queen Adeliza was written shortly after her arrival in England ; we may therefore date the poem 1121 or very soon after.®

Adeliza of Louvain is one of the least known of English queens, and her most zealous biographers have found it difficult to compile a connected account of her life.‘ At the time of her marriage with Henry I she was quite young (puel/a). All authorities agree as to her exceptional beauty, which the chronicler Henry of Huntingdon celebrated in charming elegiacs :

Quid diadema tibi, pulcherrima? quid tibi1 gemmae ? Pallet gemma tibi, nec diadema nitet .. .

Ornamenta cave, nec quicquam luminis inde Accipis; illa micant lumine clara tuo.®

The Anglo-Norman author Philippe de Thatin celebrated not only her beauty but also (like the author of the Brendan) her prudence and exemplary conduct:

une geme Ki mult est bele feme E est curteise e sage, De bones murs e large.®

She accompanied the king and court in their ceaseless wanderings

1 Cf. p. ix.

% The MSS. have 4aliz, Aelizs, Alie. Latin documents offer the forms Adeliza, -ida, -ina, Adheliza, Atheledis, -da, Aledis, Athalica, Alisia. The Anglo-Saxon Chronicle has Adels.

3 The marriage took place at Windsor on 29 Jan. 1120 (old style; rrar, new style); on the following day Adeliza was consecrated and crowned as queen. Cf. Eng. Hist. Rev., xxxiv, p. 515.

* Hannah Lawrance, Historical Memoirs of the Queens of England, i, London, 1838, ch. iv; Agnes Strickland, Lives of the Queens of England, i, London, 1840, pp. 212 ff.; cf. Dict. of Nat. Biog.

§ Rolls Series, 74, p. 243.

XXIV Introduction

and figures fairly frequently as a witness to documents,’ but seems to have taken no part in politics. After the death of Henry I she lived at Arundel, and appears to have made a marriage of inclination with William of Albini, to whom she bore seven children. Ultimately she retired to the abbey of Affligam in Flanders, where she died in 1151. In view of her importance as a patroness of literature, we may justly regret that her figure is so shadowy. Not only were the Brendan and the Bestiaire dedicated to her; but even when she had ceased to be queen, a poet named David wrote at her behest a ‘chancon ’ (now lost), presumably a chanson de geste, relating the life of her late husband King Henry.’ It is therefore certain that she made a practice of encouraging vernacular literature, and we may not unreasonably regard her as the central figure of a literary circle— one of the earliest French literary circles of which there is definite record. She had doubtless learnt to appreciate vernacular literature during her girlhood in Brabant’; yet in the characteristic Anglo- Norman manner she appears to have encouraged the production of instructive and edifying works rather than works of a frivolous nature. If the Bestiaire is to be regarded as a typical product of her patronage, we cannot praise her judgement; yet she had shown sounder taste, as regards both the subject and the author selected, in ordering the composition of a French poem on the legend of St. Brendan.

The view that Adeliza of Louvain was the patroness of the author of the Brendan, though probable in itself and generally accepted, might nevertheless be impugned. C, which is perhaps entitled to an independent position in the scheme of MSS.,‘ presents in v. r the isolated reading: DiamJe Mahatt la reine ; this may conceivably be the original line, altered in the common ancestor of A, B, and D. Among a number of queens of the period who bore the name Mahalt or Mahelt (i.e. Matilda) we find one who was also wife of a King Henry, viz. Edith (or Eadgyth) Matilda, the first wife of Henry 1°; the daughter of King Malcolm III of Scots, she was

1 Cf. W. Farrer, An Outline Itinerary of King Henry the First, part a, in Eng. Hist. Rev., xxxiv, pp. 505 ff.

3 Gaimar, Estone des Engleis, vv. 6484 ff.

§ Cf. H. Pirenne, Histosre de Belgique, i, Brussels, 1902, p. 317.

¢ See ch. iv, § 6.

® Matilda of Flanders, wife of William the Conqueror, and Matilda of Boulogne, wife of King Stephen of England, were not contemporaries of a King Henry. The Empress Matilda, daughter of Henry | and rival of King

Author and Date XXV

married to Henry I in rroo, and lived until rrr8. Like her suc- cessor Adeliza, she played but a minor part in public life; but she is recorded to have possessed, as a result of her early convent life, a scholarship unusual in those days among lay-folk and especially among women, to have corresponded with Lanfranc and the learned Bishop Hildebert of Le Mans, and to have been a generous protectress of musical clerks, scholars, and poets.! Although we have no other record of her encouraging French poetry, it is by no means impossible that she may have commissioned Benedeit to write his poem on St. Brendan. If she was his protrectress, it would follow that the Brendan was written in the year rroo or shortly after.

Between Matilda and Adeliza the choice is difficult. That the poem was first dedicated to Matilda in 1100, and re-dedicated to Adeliza in 1121, is highly improbable. In favour of Adeliza we may note that French was her native tongue, and that she is known to have patronized other Anglo-Norman poets. Moreover, the evidence on which C has been assigned to an independent position is de- cidedly slender. On the whole, therefore, we are justified in con- tinuing to regard Adeliza of Louvain as our poet’s protectress. The scribe of C, or a predecessor, has substituted Maha/t for Aastiz owing to some misreading or historical error.’

Another difficulty, perhaps not very grave, is created by v. 4 of the prologue. To what circumstances is the author alluding in the words E remandrat tante guerre? During the first nineteen years of his reign Henry I had indeed waged continual warfare—against his elder brother Duke Robert, against the French, and against rebellious barons in England and Normandy. But the years rr20—-2 were singularly peaceful, being disturbed only by an expedition against the Welsh in the summer of rr2r. Can our author have written his prologue so late as 1123-5, when Henry was again engaged in suppressing a revolt of Norman barons? More probably he was

Stephen, was married to a ‘Henry’, viz. the Emperor Henry V; but the Emperor is not likely to have been spoken of as rei, and was already dead when she became a claimant to the throne of England.

1 Cf. Dict. of Nat. Biog.

% An exactly similar error seems to have been committed by the chronicler John of Hexham (fl. 1180), who referred to the Empress Matilda as ‘ Aais' and ‘ Adela’; cf. Symeonis Monachi Opera Omnia, edited by T. Arnold (Rolls Series, 75), ii, pp. 30a and go9. Both Matilda and Adeliza were common royal names in the twelfth century.

3351 d

XXVi Introduction

merely referring in a general way to the stormy character of the previous years. It might be suggested that vv. 2-4 would have been more appropriate to the year rroo than to rr2z. In r1roo Henry’s accession put an end to William Rufus’s scheme for the conquest of France; while his Charter, and his marriage with Matilda, a princess of English descent, were public evidence of his intention to govern with justice and moderation.’ Yet v. 5 indicates that peace is to come, not through abandonment of warfare, but through victory.

Can the author of the Brendan be identified ?_ Verse 8, in which he names himself, appears perfectly explicit, but on closer examination is seen to teem with difficulties. One point is indisputable, namely that Benedeiz is the actual name of the author of the French poem, not of a source which the poct followed or claimed to follow; no other interpretation is consistent with the context. The trisyllabic form, reduced in BCD to Beneiz,? is required by the metre, the medial break falling between afostorles and dans; while the ending -deiz (A -diz) is required by the rime. The title dans (Lat. dominus, domnus) could refer either to a nobleman or to an ecclesiastic*; here, and in combination with the preceding word, it can only refer to an ecclesiastic. This view is confirmed by the subject and character of the poem, by the author’s fondness for adding religious and moral reflexions, and by his familiarity with Latin vocabulary, construction, and versification. The title dom(s)nus, originally restricted to prelates of the highest rank, had long been extended to abbots and monks, and was commonly applied in the Middle Ages (as still to-day) to monks of the Benedictine, Carthusian, and Cistercian orders. The word therefore throws little light on the author’s status. But the chief puzzle lies in the word afostosles (Lat. apostolicus). The only recognized use of this word in OF is as a substantive, meaning ‘pope’, referring sometimes (e.g. afostorie 1038) to the first pope, the apostle St. Peter. But to give it this interpretation here is out of the question; moreover, there was no pope named Benedict between Benedict X (1058-9) and Benedict XI (1303-4). Can

1 Cf. H. W. C. Davis, England under the Normans and Angevins, chs, iii and iv.

* Either a contracted form, or a more popular development omitting counter- final ¢.

3 Cf. Du Cange, s.v. domnus.

4 Cf. Smith and Cheetham, Dictionary of Christian Antiquitics; New Eng. Dict., s.v. dom.

Author and Date XXVII

apostosfes here have preserved an archaic sense of Lat. afostoltcus, vz.‘bishop’? In the earlier Middle Ages the Latin word could be applied to bishops other than the pope.' By the ninth century it had become appropriated to the pope, but relics of the former use are found as late as the eleventh and even the twelfth century,” and it is conceivable that some prelate of the early twelfth century may still have used the title. Apart, however, from the improbability of this explanation, there is no trace of any English bishop or other Church dignitary living in r12x named Benedict ; the name, in fact, is sur- pnsingly rare in all English records of the period. We are therefore reduced to a third interpretation, namely, to take aposfoiles as an adjective. Du Cange quotes the expressions afostolicus ordo ‘monasticus ordo’ and afostolica vita ‘monastica vita’; may not afostoiles signify merely ‘ belonging to a monastery ’, reinforcing dans which conveys the same meaning, or perhaps indicating that dans is an ecclesiastical and not a lay title? Unfortunately no similar instance of this use of OF afostotle has been met with. Provision- ally, however, and until some better explanation is forthcoming, we may regard the author of the poem as an Anglo-Norman monk named Benedeit.

No one has hitherto ventured to identify the author of the Brendan with any historical personage. He might be expected, as a protégé of the queen and presumably a frequenter of the court, to figure as a witness to charters or otherwise to be mentioned in official records ; but this expectation remains unfulfilled. The numerous official documents of the Norman period collected in the Regesta Regum Anglo-Normannorum,! or quoted in W. Farrer’s [tinerary of Henry I,‘ mention but one Benedict—an archdeacon of Rouen and royal chaplain, occasionally appearing as a witness to charters down to 1106; this personage, besides being too early for our purpose, lived on the Continent and not in England. After all, there is no definite

1 Cf. Du Cange; Smith and Cheetham, Dictionary of Christian Antiquities.

* At the Council of Reims in 1049 it was solemnly declared that the Pope was Primate of the whole Church and ‘ Afoséolicus’; and the Archbishop of Santiago de Compostella was excommunicated for venturing to assume ‘ cu/men apostolic: nominis’, This Archbishop was, however, no ordinary bishop, but had doubtless assumed the title in order to claim apostolic dignity for his see, which possessed the supposed relics of the Apostle St. James. The twelfth-century example is from the works of Rupert of Deutz.

* Edited by H. W.C, Davis; vol. i (down to 1100) alone has appeared.

* In Eng. Hist. Rev., xxxiv (1919).

XXVili Introduction

evidence proving that the author of the Brendan frequented the court, though vv. 14-18, emphasizing the author's readiness to do service and his liability to unwarranted attacks, can easily be inter- preted as polemic against court rivals.

English literary annals of the period mention only one Benedict, namely Benedict of Gloucester, author of a Latin life of St. Dubricius." A plausible case could be made out for identifying the author of the Brendan with Benedict of Gloucester. The latter was a Benedictine in the Abbey of St. Peter at Gloucester *; he was therefore entitled to the appellation dominus or dans, and doubly entitled (firstly as a monk, and secondly as one enjoying the patronage of St. Peter) to the epithet apostolicus or apostoiles. The Abbey of St. Peter was in a flourishing condition, both materially and intellectually, during the century following the Conquest ; it possessed a valuable library, and produced several distinguished scholars and writers.» Documents of 1121 or shortly after show Henry I confirming its rights or adding to its privileges. Like the author of the Brendan, Benedict of Gloucester was interested in Celtic saints; in his prologue he explains that, after prolonged study of the lives of saints, he has picked out St. Dubricius as the most eminent of all the British saints of his time. He flourished after 1120, as he describes the translation of the bones of St. Dubricius from Bardsey to Llandaff, an event which took place in that year.

But these coincidences may be illusory, and a number of difficulties must be overcome before the identification can be accepted. In the first place, Benedict of Gloucester has certainly been ante-dated. The Dicthonary of National Biography, following Wharton and Tanner,’ describes him as flourishing in 1120; but this date is based on the assumption that Benedict was personally present at the trans-

1 Edited, with omissions, by T. Wharton, Anglia Sacra, ii, London, 1691, p. 654; see also p. xxvi of the same volume, and the Dict. of Nat. Biog. Dubricius was the reputed founder of the bishopric of Llandaff; according to the tenth-century Annales Cambniae he died in 612, but twelfth-century accounts place him earlier.

2‘ Ego Benedtictus habitu coenobsi Apostoli Petri Claudiocestriae monachus adornatus (Wharton: adhortatus)’.

8 Cf. Victoria History of the County of Gloucester, ii, London, 1907, pp. 53 ff.

* Farrer, of. ctt., pp. 515 and 528,

® ‘Cum in gestis sanctorum patrum diutius operam studio dedissem, sanctumque Dubniaum omnium sanctorum tunc temporis in Britannia degentium meritis et vita praeslantissimum ... aegre tuli ipsius gesta nesari...’

Author and Date xxix

lation of St. Dubricius’ bones. A comparison of Benedict’s account of this ceremony with his principal source, the Vita S. Dubrias ascribed to Geoffrey of Llandaff,’ shows that he has copied the description almost word for word ; a few phrases have been omitted or added, but in no case is an addition based on a personal remini- scence. Moreover, he has combined the original life of St. Dubricius with certain passages from Geoffrey of Monmouth’s Historia regum Britanniae, in which the saint, described as Archbishop of Caerleon, crowns Arthur king, and harangues the Christian army before the battle of Badon Hill? He cannot therefore have compiled his work until after the first publication of Geoffrey's Astoria (1136-8), and perhaps did not do so till considerably later. Sir T. D. Hardy may therefore have been justified in assigning Benedict’s Vita S. Dubriat to the second half of the twelfth century.* It is still possible that Benedict may have written the Anglo-Norman poem in his early days, the Vita S. Dubriat much later in life. Against this hypothesis we can only urge the entire absence of any resemblance or point of contact between the two works. Even if we assume that he was hampered by using Latin, the avowed plagiarism,‘ the commonplace- ness, and the lack of artistic sense shown by Benedict of Gloucester make it incredible that he can have been the author of the Brendan.

III. VERSIFICATION §1. THE METRE

The Voyage of St. Brendan is written in the standard OF metre for marrative poems other than chansons de geste, viz. octosyllabic lines miming in pairs. With the exception of the tenth-century Passton and the Life of St. Leger, it is probably the earliest extant French poem written in this metre’; yet the technical perfection

1 The text of the Book of Lian Dév, ed. J. G. Evans, Oxford, 1893, pp. 78 ff.

2 Chapters r and 4 of book ix ; edition of San Marte, Halle, 1854.

3 Descriptive Catalogue. .., i, p. 42. Yet in vol. ii of the same work, p. 178, following Cave (Seriptorum ecclesiasticorum historia Iiteraria, ii, Oxford, 1743, p- 213), he assigns Benedict of Gloucester’s work to 1130.

* ‘Quocirea quae in autenticis patrum digestis super eodem paire repert, com- pulare et coadunare stilo non erubut.’

8 The earliest French version of Marbode’s Lapidary and the two lapidaries in octosyllabic couplets by Philippe de Thaiin are roughly contemporary with

the Brendan, but cannot be accurately dated; cf. P. Studer and J. Evans,

XXX Introduction

attained by the author indicates that he was using a well-established form. The development cf the OF octosyllabic line from the Latin iambic dimeter has been traced by G. Melchior,’ in whose disserta- tion particulars are given of earlier literature on the subject. The earliest French poems in this metre were all written by authors familiar with Latin.

§ 2. THE FEMININE LINES

The Brendan consists of 1840 lines, of which 1242 have masculine rimes and 598 feminine rimes. The masculine lines are normal ; it is true that the MSS. frequently have lines with more or fewer than eight syllables, but in every instance, provided allowance is made for certain peculiarities in the counting of syllables, a comparison of them leads us back to a regular octosyllabic line. The feminine lines, on the other hand, do not show this uniformity ; some of the MSS. present them more or less consistently as lines of eight syllables (i.e. stressed on the seventh), others as lines of nine syllables (i.e. stressed on the eighth). Thus, to take a typical example, in vv. 145-6 A, B, and C read (with slight variations in spelling) :

I 2 3 4 5 6 7 8 Dunc | prent | cun|gé | a | ses | frelres

I 2 3 4 5 6 7 8 As | quels | il | ert | mult | dulz (B bons) | pelres; whereas D reads :

I a 3 4 5 6 7 8 9 Dunc | prent | con|gié | de | tuz | ses | frelres

I 2 3 4 5 6 7 8 9 As | quels | il | es|teit | mult | duz | pelres; and E: 1 2 3 4 5 6 7 8 9 Dont | prant | con|gié | a | tos | ses | frejres 1 2 3 4 5 6 7 8 9 As | quels | il | ot | es|té | dous | pelres.

Anglo-Norman Lapidaries, Paris, 1924, pp. 23f., 200 ff., 260 ff. The Provencal Chanson de Sainte Foi d Agen and the Franco-Provencal Alexander poem by Alberic, both earlier than the Brendan, are in octosyllabic verse, but the lines are grouped in tirades; similarly Gormont et Isembart, which in its extant form is later than the Brendan.

1 Der Achtsilbler in der altfrans, Dichtung mit Ausschluss der Lyntk, Diss. Leipzig, 1909; cf. A. T, Baker, in Mod. Lang. Rev., xi, p. 435.

Versification xxxi

Itis obvious that little progress can be made with the classification of the MSS. or the establishment of the text until we have decided which type of line was that of the original and which has been tampered with. In a case of this sort, the numbers and the relative ages of the MSS. count for little; it can be argued, for instance,’ that A, B, and C, which have the shorter lines, are all derived fom a MS. with altered text, and that D and E represent the primitive text. It might be suggested that the author’s sense of the sylabic principle in versification had been blunted by contact with English *; that he used both types of feminine line, like the first French translator of Marbode’s Lapidary and several later Anglo- Norman writers; or that at least he shows the beginnings of the laxity rhich characterizes most Anglo-Norman versifiers.* Yet none of these arguments is consistent with the regularity of the masculine lines, or with the virtuosity that our author shows in his rimes and other technical details. A priori it is improbable that a poem originally composed in normal octosyllabic lines should have had the feminine lines systematically reduced by a syllable; whereas the converse is natural and highly probable, the unusual type of line being likely to jar upon scribes, readers, and audience equally. But the priority of the shorter lines is not so obvious as to make proof unnecessary ; the view that the longer lines are those of the original poem is apparently still held at the present day.‘ F. Diez,’ who was the first to draw attention to the peculiar form of the Brendan, declared without hesitation that the abnormal lines were the work of the author, the normalized lines due to scribes. The same view was upheld by H. Suchier,* who considered the feminine lines to have lost their initial unaccented syllable in imitation of English accen- tuated verse, and defended in detail by J. Vising,’” who, however, rejected Suchier’s explanation. The opposite view found a defender in A. Boucherie,* who compiled a list of some twenty instances in which for reasons of syntax the original line must be regarded as

1 A. Boucherie, in Revue des langues romanes, xxiii (1883), pp. 180-4.

7H. Suchier, Ueber die Matthaeus Paris sugeschriebene Vie de Seint Auban, Halle, 1876, p. 24.

* Cf. J. Vising, Anglo-Norman Language and Literature, London, 1933, Pp. 79 ff.

* Tanquerey, op. cit.; cf. p. xix, note 3.

8 Aliromanische Sprachdenkmale berichtigt und erklart, Bonn, 1846, p. 100.

* Op. at.; cf. E. Koschwitz, in Zéschr. f. roman. Phil., ii, p. 341.

1 Etude sur le dialecte anglo-normand . . .. pp. 45-52- 8 Op. cit.

XXXIi Introduction

having nine syllables ; but his syntactical arguments were based on the practice of modern rather than medieval French, and subsequent writers ' have demonstrated the correctness of Vising’s view.

If the eight- and nine-syllable versions of a feminine line are care- fully compared, it will be seen that in some instances the two versions are equally acceptable (e.g. vv. 37-8, 54, 75-6, 134, 207, 225-6), but that in others the longer line is ungrammatical (e.g. vv. 654 E, 1005-6 D, 1271 E, 1454 E, 1789 DE), or illogical (e.g. vv. 103 DE, 121-2 E, 224 E, 351 D, 645 D), or meaningless (e.g. vv. 888 E, 974 E, 990 E), or evidently modernized (e.g. vv. 601-2 E, 815-16 E, 1099-1100 DE, 1679-80 DE, 1718 E), and that hardly any instance can be found in which the longer line is superior to the shorter. In E, which regularly employs the nine-syllable feminine lines, the ninth syllable has sometimes been produced by the introduction of a Picard feature (e.g. vv. 522, 720, 763-4, 1045); such lines cannot be attributed to the Anglo-Norman author. Furthermore, in a large number of instances the extra syllable destroys the medial break (e.g. vv. 37, 53, 84 D, 99 E, 103 E, 147 D), which is almost always observed in the masculine and the eight-syllable feminine lines.” We know that the subdivision of octosyllabic lines into hemistichs 1s a sign of antiquity, and its absence a sign of modernity.* It would be extremely difficult for a reviser to introduce a regular break into lines which do not possess one, whereas it is easy to destroy an existing break. D and E show far greater irregularity as regards the medial break in the feminine than in the masculine lines ; this can only be due to the fact that the former have undergone alteration.‘ For all these reasons the observance of the break must be credited to the author, non-observance to a reviser. Finally, in a great many of the nine-syllable feminine lines the extra syllable is harmless in itself, but unnecessary ; this fact alone constitutes a strong argument against attributing the extra syllable to the author, whose style (as exemplified in the masculine lines) is characterized by remarkable concision, pithiness, and avoidance of padding. In the whole poem, if we omit those lines where a hiatus is optional (e.g.

1 Birkenhoff, Ueber Metrum und Reim..., pp. 7 ff.; W. Hammer, in Zéschr. J. roman. Phil., ix, pp. 77 ff. ; Calmund, Prolegomena .. ., pp. 27 ff.

2 Cf. Calmund, Prolegomena..., pp. 49 f.

* Cf. § 3.

4 E, however, in which the text has been modernized and partly re- written, often has no break even in masculine lines, e.g. vv. 275, 426, 474, 766, 1526.

Versification XxxXiii

533, 1340), and ignore the unpronounced ¢ in guarKe) 328 A,’ /(e)res 553 BD, truole)rat 586 ABD * and 4verat 605 A,’ there are only ten feminine lines which contain nine syllables in all the MSS. In several of these lines the MSS. disagree among themselves, and it is easy to see that scribes have added syllables independently, in two or more different ways. Thus in v. 800 we find the two readings :

AB Cume se lur fust destinee, DE Cum s’ele lur fust destinee,

from which we may safely reconstruct the reading of the archetype * as Cum se lur fust destinee. In v. 1316 three different readings are offered :

A Mes granz peines n'i ai dunc mie,

B ” ” » naijo , ”» 9

DE De mes granz peines n’ai jo mie, pointing to the conclusion that A and B have independently added a syllable to the original line

Mes granz peines n’ai dunc mie,

while the common ancestor of D and E ‘ introduced a different con- struction for the sake of obtaining nine syllables. In vv. 615 and 1553 the circumstances are similar. In v. 1666 the scribe of A obviously repeated gue through inadvertence. There remain the following five instances, in which the MSS. show some measure of agreement among themselves : 327 ABE Treis jurs enters i sujurnerent. D Treis jurz al chastel surjurnerent. 588 ABDE As (D Jesque as, E Les) uitaves de Pentecuste. 866 ABDE Desque (8 D.as, D Jesque as, £ Les) uitaves de Pentecoste. 1573 ABE Tuzdis tres jurs en la semaine. D ” » » dey ” ° 1598 ABDE Mis ethez (8 eded, DE eez) est cent e quarante. In these lines it is most probable that the monosyllables #, de, de, en (or de), and ¢ respectively are supernumerary, and that they were introduced either in the archetype or independently by different scribes ; they are certainly the obvious syllables to add, though in no

1 Cf. Appendix II, § 6 f. 2 Cf. ch. viii, § 3, xxxi.

% The term ‘archetype’ is here used to designate the MS. from which A, B, D, and E all descend; cf. ch. iv, § 5.

‘ Cf. ch. iv, § 2,

3351 e

XXXIV Introduction

case indispensable.’ This view receives support from v. 292, where ABDE all have .

Qui (A Que) forment fud e bone e bele, but the fragment C alone has preserved the unexpanded form of the line :

Ki forment fut bone e bele. We may therefore conclude that the original poem contained lines of eight syllables only, and that the feminine lines which appear with nine syllables in the MSS. have been altered by revisers or scribes.

The Brendan is not unique in this respect. Lines in which a final unstressed syllable is counted, instead of being supernumerary, are found in a small number of other texts, not all of Anglo-Norman origin. Unfortunately, in Anglo-Norman texts it is often impossible, owing to the early degeneration of versification, to say with certainty how many syllables a line was intended by its author to contain. Elie de Wincestre, in his L’afattement Caton? (an Anglo-Norman version of the Disticha Catonis, dated about 1140 by Vising *) systematically followed the same principle as the author of the Brendan, applying it to lines of six and ten as well as eight syllables. His example probably influenced another Anglo-Norman translator of the Disticha, viz. the monk Everart, whose version (of the end of the twelfth century)‘ is in lines of six syllables, grouped in six-line strophes ; some of his feminine lines, it is true, contain seven syllables in their present state, but may have been altered by the scribes. The existence of such poems as these appears to have created considerable uncertainty as to the correct form of feminine lines, and various Anglo-Norman poenis are found in which the authors employed both types concurrently. Thus in the earliest French version of Marbode’s Lapidary,° although the majority of the feminine lines are of the normal type, many can be read as having either eight or nine syllables (e.g. vv. 105, 209, 262), others vary in the MSS. (e.g. vv. 106, 138, 845-6), while vv. 653-4 undoubtedly contain eight syllables

1 See the notes to vv. 588 and 1598.

* Edited by E. Stengel, in Ausgaben und Abhandlungen, xivii, Marburg, 1886, pp. 106-156.

3 Anglo-Norman Language and Literature, p. 48.

4 Edited by E. Stengel, parallel with Elie’s poem; also by F. J. Furnivall, The Minor Poems of the Vernon MS., part ii (Early English Text Society), rgor, Pp. 553 ff.

5 P. Studer and J. Evans, op. at., pp. 19 ff.

Versification XXXV

only. ‘The same uncertainty prevails in a life of Thomas Becket,’ in an early thirteenth-century version of St. Patrick’s Purgatory,’ in a thirteenth-century Zvangile de l’Enfance, and in the poem of the Brut cycle entitled De grauatz jaians.‘ Continental texts (all in octosyllabic couplets) in which masculine and feminine lines have the same number of syllables are few but important : Terramagnino of Pisa’s Doctrina de cort® ; the Breviari d’ Amor by Matfre Ermen- gaut of Béziers*; and the first recension of Le FPelerinage de Vie Humaine by Guillaume de Deguileville.’ The wide distribution, as regards both time and place, of the examples of this metrical system is remarkable ; but the three continental texts have not necessarily any connexion with those written in England. On the whole the choice of this form was unfortunate, especially for longer poems. A succession of octosyllabic lines, some stressed on the eighth and some on the seventh syllable, but not alternating regularly, produces an uneven and consequently displeasing effect. By giving the stressed and unstressed final syllables equal metrical value, such a system carries the syllabic principle of French versification to an extreme, and neglects the principle of stress.”

1 Fragments d'une Vie de Saint Thomas de Cantorbery, edited by P. Meyer (Soc. d. anc, textes fr.), Paris, 1885; cf. Introduction, p. xxxiv.

2 Edited by J. Vising, Giteborg, 1916; cf. pp. 8 ff. Out of 186 feminine lines in this text, 131 are a syllable short according to the ordinary rules.

3 E. Gast, Die beiden Redaktionen des Evangile de 1 Enfanee, Diss. Greifswald, 1909. In the first 24 strophes of the Anglo-Norman version eight-syllable feminine lines are used consistently; from strophe 25 onwards there is irregularity, culminating towards the end in complete metrical confusion.

* Published by A. Jubinal, Nomveaxn Recueil de Contes, Dits, Fabliaux .. ., ii, Paris, 1842, p. 354, and by Fr. Michel, Gesta Regum Britannsae, for the Cambrian Archacological Association, 186a, p. 199; not yet critically studied. To judge from the confusion of verb-endings shown by the rimes, the poem was not written before the thirteenth century. Feminine lines with only eight syllables predominate.

5 A versification of Raimon Vidal’s Razos de trobar, written in Provencal by a Pisan in the third quarter of the thirteenth century; cf. P. Meyer, in Romamia, viii, pp. 181-210. It should be noted that Terramagnino’s command of the language and metre was extremely poor, and that he did not even understand the distinction between masculine and feminine rimes.

* An encyclopaedic compilation in Provencal, begun in 1288; edited by G. Azais for the Société Archéologique of Béziers, 2 vols., 1862.

7 Written 1330-2; edited by J. J. Stiirzinger for the Roxburghe Club, London, 1893; cf. letter by P. Meyer, printed in Furnivall’s 7mal-Forewords to sy ‘ Parallel-Text Edition of Chaucer's Minor Poems’, London, 1871, pp. 100-2.

® Cf. F. Diez, Altromanische Sprachdenkmale, Bonn, 1846, p. 100.

* Cf. E. Stengel, in Ausgaben und Abhandlungen, x\vii, p. 109.

XXXVI Introduction

The reasons for giving the same number of syllables to feminine as to masculine lines probably varied according to circumstances. Terramagnino of Pisa and the later Anglo-Norman writers may have neglected the distinction between masculine and feminine endings owing to their imperfect understanding of Provencal or French pro- nunciation. Conscious imitation of a literary model will account for the adoption of this practice by a writer like Everart. So popular a work as the Anglo-Norman Brendan may well have exerted a direct influence on later authors, and it may be noted that the poem of St. Patrick’s Purgatory belongs to the same cycle of eschatological literature. Other authors, such as Guillaume de Deguileville, may have adopted an abnormal metre either to avoid monotony (as Diez suggested) or in the hope of attracting attention. Matfre Ermengaut’s treatment of the octosyllabic line was doubtless due to his familiarity with Provencal lyric poetry, in which the rule prevailed that masculine and feminine octosyllables grouped together should always contain the same number of syllables... None of the Anglo-Norman writers in question is likely to have been influenced by the courtly lyric ; more probable in their case—for all of them were clerks familiar with Latin—is the influence of medieval Latin verse? The Latin octosyllabic line, much used in the Middle Ages, naturally did not admit of any variation in the number of syllables. In the case of the author of the Brendan, it is highly improbable that his choice of metre was due to ignorance, accident, or slavish imitation of a model. In spite of the modest tone of the prologue, he was something of a virtuoso ; the strict regularity of his counting of syllables, his obser- vance of the medial break, the accuracy and richness of his rimes, all show the hand of a man who prided himself on his technique. His choice was certainly deliberate, and due in all probability to imitation of medieval Latin verse.? ‘This view is confirmed by a study of his rimes and of the internal construction of his lines; in each his practice shows marked peculiarities, only intelligible on the assump- tion that he was following Latin models.‘

1 e.g. nos. 2, 7, 13, and 38 of Bertran de Born’s poems, edited by A. Stimming, Halle, 1913. Masculine lines of seven syllables were, however, combined with feminine lines of eight syllables. Cf. Azais’ Introduction to the Breviari, pp. cxi ff.

2 Cf. H. Suchier, Reimpredigt, Halle, 1879, p. xlix.

3 So G. Melchior, of. cit., p. 24; but Melchior’s attempt to read the lines of the Brendan as stressed verse conflicts with the principles of French accentu- ation. 4 Cf. §§ 3 and 6.

Versification XXXVil

The MSS. of the Brendan, as of most of the above-mentioned poems written in a similar metre, reveal a strong tendency on the part of scribes to make the feminine lines of normal length by adding a syllable. As a rule this has been done by some simple expedient, such as introducing hiatus, adding a conjunction, adverb, article or pronoun, repeating a preposition, substituting a longer form or Word of the same meaning, replacing one tense by another, adding a prefix, or altering the word-order. But frequently the syllable could not be conveniently added without some more drastic change. In D and E not only are new constructions and words of different meaning substituted for those of the original, but in many instances whole lines and couplets have been entirely rewritten. The result has been a grievous corruption of the author’s style and metrical system, and sometimes a complete distortion of his meaning. An examination of each of the MSS. from this point of view! has thrown much light on their relative reliability, and has made it possible to determine with some precision the original text of the lines in question. In A and C the feminine lines are not expanded to any appreciable extent ; in B they are expanded sporadically (18-25 per cent.), in D in the majority of instances (71-2 per cent.), and in E almost uniformly (98-8 per cent.).

§ 3. THE INTERNAL CONSTRUCTION OF THE LINE

An analysis of the internal construction of the octosyllabic line in this text produces somewhat remarkable results, Every line can be assigned to one of the following types:

(1) Fourth syllable stressed and terminating a word ; e.g. 2 3 4 5 6 7 8

Masc.: v.9 Que | co|man|das, | ¢o | ad | en|pris. 914 lines. rt 2 3 4 § 678 Fem.: v.2 Par | qui | valjdrat | lei | dijvijne. 414 4,

) For details, see Appendix I. Cf. Birkenhoff, Ueber Metrum und Ram .. +; pp. 7-17; Calmund, Prolegomena . . ., pp. 27-44.

XXXVili Introduction

(3) Fourth syllable unstressed, terminating a word ; e.g.

t 2 3 4 5 6 7 8

Masc.: v.§ Por | les | ar|mes | Henjri | lu | rei. 314 lines. 1 2 3 6 7 8

Fem.: v. 156 Fors | qua|tor|ze | de | lur | frelres. 155 55

(4) Fourth syllable an unstressed word ; e.g.

I a 3 4 5 67 8

1840 ,,

In type 4, the unstressed fourth syllable may consist of : a definite article: vv. 23, 133, 398, 952, 1064, 1693, 1739, 1742 ; an indefinite article :

gog Vers eals veint uns marins serpenz; & possessive adjective: vv. 174, 397, 695, 696, 956, 1106, 1273, 1277, 1426, e.g. 397 Dunt poiim sa feste faire ; a personal pronoun: vv. 223, 352, 768, 988, 1428, e.g. 223 Li_abes dunc les amonestet ; a preposition: vv. 44, 127, 183, 225, 725, 925, 1192, 1317, 1441,

1734, ©-§.- 183 Ne plus que a quarante dis ;

a conjunction: vv. 76, 752, 1452, 1616, 1737 ; a negative particle: vv. 155, 1392, e.g. 155 Que mener ne volt lur peres.

Similar lines will be found in the tenth-century /asston, vv. 108, 263, in Gormont et Isembart, vv. 19, 50, etc.

The following types of line, found in other early French poems written in octosyllabic verse, do NOT occur in the Brendan:

(5) Fourth syllable stressed, followed by an unstressed vowel which undergoes elision ; e.g.

1 2 3 4 5 6 7 8 Mystere d’ Adam, v. 232 Tu | es | trop | ten|dre_e | il | trop | dur.

1 Edited by P. Studer, Manchester, 1918.

Versification XXX1X (6) Fourth syllable stressed, followed by a supernumerary unstressed syllable ; e.g. 2 2 3 4 - 5 6 7 8 Mystere d Adam, v. 140 N’es | tu | en | gloilre? | ne | poez | mofrir. First French version of Marbode’s Lapidary,! v. 194

12 34- 5 6 7 8 U|me | relfreijde | ki ] a | trop | chialt.? (7) Fourth syllable not stressed nor terminating a word ; e.g.

1 2 #3 #456 7 8 Gormont et Isembart, v. 543 De | ceo | fist | Lolo|wis | que | pruz.

I 2 3 4 § 6 7 8 Mystere d Adam, v. 183 Deus | t’a | fait | gar|dein | de | son | ort.

In the Brendan there is seldom any doubt as to the type to which a line should be referred; but the following lines perhaps require comment : V. 40 appears in Bartsch’s text ‘ as Cum hoem qui eret de grant sens (type 2) ; but the form eref is in none of the MSS., and it is much safer to follow AD, supported by C, and read Cum hoem qui ert de mult grant sens (type 1). V. 92 appears in Bartsch’s text ° as U nuls nen entret fors li piu (type 2); but the reading of A, supported by B, is at least equally acceptable : U nuls n’entret fors sul li piu (type 3).

V. 354, as given in ABE (D is hypermetric), appears to belong to type 7: Qu’en la (DE sa) nef receut |i peres;

but the author may have regarded the prefix ve- as an independent unstressed word, and the line has therefore been assigned to type 4.

1 P. Studer and J. Evans, Anglo-Norman Lapidaries, p. 22.

2 Similar lines occur in the tenth-century Passion (vv. 64, 247, 266) and in many Anglo-Norman poems of the twelfth and thirteenth centuries. Cf. G. Paris, in Romania, ii, p. 297; H. Suchier, Usber die... Vie de Seint Auban, Halle, 1876, p. 26; M. K. Pope, £tude sur la langue de Frere Angier, Paris, 1903, p. 72; A. T. Baker, in Mod. Lang. Rev., vii, p. 184; P. Studer, Le Mystere d Adam, pp. liiif.

3 Edited by A. Bayot (Class. fr. du moyen dge), Paris, 19a1.

« Bartsch-Horning, La langue et la Iitt. fy., col. 72, line 5.

® Op. at., col. 73, line go.

xl Introduction V. 1052, as given in AD, 1s a syllable short : Sachez murir nus estout (D estuet) ; B and E have added the syllable gue before murir, thus bringing the line under type 7, but it is probable that their additions were made

independently. The syllable was doubtless missing in the archetype, the original line being

Sachez murir [que] nus estuet' (type 1).

V. 1248, which Calmund discussed without reaching a conclusion,’ is to be read : Ja’st ta mercit itant bone (type 1), and not, with B: Ja est ta mercit tant bone (type 7). The form st¢ant is in ADE, and to alter the word-order would be unjustifiable. V. 1659 might be read:

Dunc dist li ostes: Ne targez (type 2), but more probably a syllable was missing in the archetype : Dunc dist l’ostes: Ne [vus] targez (type 3). V. 1716, as it stands in A, appears to be an example of type 6: Ne me merveille si poiir unt ;

but the correct verb-form would be mervei/, and DE offer an alterna- tive reading V’est mervetlle, which has been adopted in the critical text.

In other instances (e.g. vv. 6, 420, 802, 1090, 1760) where the reading to be adopted is debatable, a line agreeing with type 1, 3, or 4 is furnished by one or more MSS., or is readily reconstructed from their divergent readings.

Two tendencies, sometimes mutually conflicting, governed the construction of the octosyllabic line in the earliest French and Provencal poetry; (a) the tendency to place an ictus on the fourth syllable, and (4) the tendency to divide the line into two equal portions (or hemistichs). As regards (a), the practice of the Brendan, apart from its avoidance of types 2 and 5, agrees on the whole with that of earlier and contemporary texts (notably the Clermont poems,°

1 For the word-order, cf. ch. viii, § 5, Xiv. 2 Prolegomena..., pp. 46f. 3 Cf. G. Paris, in Romania, i, pp. 292 ff.; ii, pp. 295 ff.

Versification xli

the Chanson de Sainte Fot d’Agen,' the Alexander fragment by Aibenic,* Gormont et Isembart, the early Lapidaries, and the A/ystere @’Adam), and of other Anglo-Norman and provincial French texts down to a considerably later date (e.g. the Roman du Mont Saint Michel, the lives of St. Osith ‘ and St. Modwenna,' and the Passion of St. Andrew). In the majority of lines (1330 out of 1840) the fourth syllable is a stressed one, while in the remainder (with forty- one exceptions) the third syllable is stressed and is followed by a syllable containing feminine e¢. But as regards (4) the Brendan is far sticter than any other text. In the whole poem there are only three lines in which the fourth syllable does not coincide with the end of aword. Lines of types 2 and 5 are frequent in all the other texts Just mentioned, type 2 also in Latin octosyllabic verse of the period ; the deliberate avoidance of them seems to be peculiar to the author of the Brendan among OF poets, and betokens an exaggerated respect for the division into hemistichs. The bold hiatuses which are tolerated between syllables 4 and 5, but not in any other position,” point to the same conclusion ; so also do the surprising liberties which the author took with the word-order,® unnecessary and inex- plicable except as a means of balancing the hemistichs.

Why did the author of the Brendan limit himself thus to so few types of line? Mere primitiveness or lack of skill is not an adequate explanation. If we follow F. Diez® and G. Paris" in deriving the structure of the French octosyllabic line from that of the Latin iambic dimeter of the early Middle Ages, we must regard 3 and 2 as the primitive types, which, however, owing to the oxytone tendency of the French language, early became outnumbered by type r. Type 5 is a mere variety of 1, type 6 has probably been influenced by epic versification, types 4 and 7 show the beginnings of freer treatment of the octosyllabic line (with no constant stress

1 Edited by A. Thomas (Class. fr. du moyen dge), Paris, 1925.

2 Cf. P. Meyer, Alexandre le Grand dans la litt. fr. du moyen dge, ii, Paris, 1886, pp. 74 f.

3 By Guillaume de S. Paier; edition by P. Redlich, Marburg, 1894.

* Cf. A. T. Baker, in Mod. Lang, Rev., vii, pp. 179 ff.

* Cf. H. Suchier, Ueber die... Vie de Seint Auban, p. 54.

* Cf. A. T. Baker, in Mod. Lang. Rev., xi, pp. 435 f.

7 Cf. § 4, @, ii.

® Cf. ch. viii, § 5, xiv.

> Altromanische Sprachdenkmale, pp. too and 109.

18 Romamia, i, pp. 292 ff.; and in the preface to A. Tobler, Le Vers francais, Paris, 1885, p. xii.

3351 f

xlii Introduction

save on the eighth syllable) as practised later by Wace and Crestien de Troyes. The avoidance by our poet of types 2 (with two excep- tions), 5, and 6 is clearly not a sign of primitiveness, but the result of deliberate choice. The explanation of this, as of so many other features of the technique of the Brendan, is probably to be found in medieval Latin poetry. In many Latin poems of the twelfth and thirteenth centuries the octosyllabic line was regularly divided in the middle, i.e. types 2 and 7 were avoided (types 5 and 6 do not occur). Sometimes the division was emphasized by the introduction of internal rime.’ A typical example among many is the celebrated satire on Rome by Gautier de Chatillon (second half of the twelfth century), beginning : Propter Sion | non tacebo sed ruinam | Romae flebo, quousque justitia rursus nobis | oriatur et ut lampas | accendatur justus in ecclesia.’

Naturally it is impossible to point to any actual model which the author of the Brendan followed ; but that he should have imitated a practice prevailing in Latin poetry is entirely consistent with what we know of his tastes and habits.

Although in this poem the fourth syllable almost invariably coin- cides with the end of a word, it does not follow that there is always a natural break in the sentence between syllables 4 and 5. In the large majority of lines the hemistichs undoubtedly consist of self- contained phrases; but the lines assigned to type 4 have no natural break after the fourth syllable, and even in those assigned to types 1

1 Cf, W. Meyer, Der Ludus de Antichristo und tiber die lateinischen Rythmen, in Gesammelte Abhandlungen sur mittellateinischen Rythmik, i, Berlin, 1905, Pp. 317.

2 Published by E, du Méril, Poésses populaires latines antericures au dousiome siecle, Paris, 1843, p. 231, and by J. A. Schmeller, Carmina Burana, Breslau, 1904, p. 16. Other examples will be found among the poems edited by W. Meyer, in Die Arundel Sammlung mittellatetnischer Lieder (Abhandlungen der kgl. Geselischaft der Wissenschaften eu Gottingen, philol.-hist. Klasse, Neue Folge, xi. 2), Berlin, 1908; notably no. 24, beginning

Licet eger | cum egrotis

et ignotus | cum ignotis, which figures also in Carmina Burana, p. 41, and in the works of Gautier de Chatillon (K. Strecker, Die Gedichte Walters von Chatillon, i, Berlin, 1925, no. 27).

Versification xliii

and 3 the medial break sometimes parts closely-linked words. We find the break separating

preposition and substantive: 644 L'abes ovoec | ses cumpaignuns 1434 Vinc ci entre | nune e midi

possessive and substantive: 765 Ainz que vostre | venir sousum

adjective and substantive : go8 Plus que pur nul; mal qu'il traient 1062 Sailent bestes | ruistes de mer 1763 N’i vient nule | nue del air

adverb and adjective : 1014 E trenchantes | fort les poes adverb and adverb: 938 Drechent forment | halt les testes subject and verb : 732 Pur lui que nus | mult amames

1058 Vers tuz vus fud | Deus bons guaranz auxiliary and participle: 484 Si cum lur out | dit cil frere

1057 Perilz avez | suffert plus granz.

See also wy. 71, 161, 222, 352, 446, 515, 572, 894, 908, 938, ror4, 1276, 1402, 1466, 1700, 1763, and 1787. In view of the difficulty of composing a narrative poem on such rigid principles, it is not sur- prising that the author should not have succeeded in every instance in adapting sense to form.

Whether this division of the octosyllabic line constitutes a ‘ caesura ’ is a controversial question into which it is unnecessary to enter deeply here. The answer depends on the meaning attributed to the term. If by ‘caesura’ we imply either (a) that the break is compulsory and its position rigidly fixed, or (4) that it must immediately follow a stressed syllable, or (c) that it must correspond to a natural pause in the sentence, or (¢@) that it permits the introduction of a supernumer- ary unstressed syllable, then there is no caesura in the Brendan. If we interpret the term more loosely as a habitual break in the line at a certain position, then there is undoubtedly a caesura in this poem. Although the division of the octosyllable was due in the first instance to the tendency to stress the third or fourth syllable, in the Brendan it is independent of the stress. ‘Tobler’s affirmation ? that the division

' See A. Tobler, Vom frans. Versbau, 5th edition, Leipzig, t9to, pp. r1of.; G. Paris, in preface to Tobler, Le vers francais ancien ef moderne, Paris, 1885, pp. xif.; F. Spenz, Die syntaktische Behandlung des achlsilbigen Verses in der Passion Christi und im Leodcgar-licde, Marburg, 1886, pp. 1-11; G. Paris, in Romania, xxv, pp. 533f.; E. Stengel, in Grundriss der roman. Phiiologie, ii, 1, Strassburg, 1902, pp. 44-6; A. Schulze, in Zéschr. f. roman. Philologie, xxx, P. 355; C. Appel, in Melanges Chabaneau, Erlangen, 1907, p. 199; G. Melchior, Der Achtsilbler in der altfrans. Dichtung, Diss. Leipzig, 1907; A. T. Baker, in Mod. Lang. Rev., xi, p. 435.

? Still upheld by Prof. Vising; cf. Literaturblatt, xlviii (1927), col. 113.

xliv Introduction

of the octosyllabic line was involuntary certainly does not apply to this poem.

The strictness of Benedeit’s system of versification has been of great assistance in the establishing of the critical text, and has some- times been a deciding factor in the choice between alternative readings. Thus in v. 26 A and E have Que plus demander, BCD Que demander plus, the former neglecting the medial break, the latter observing it ; there can be no doubt that BCD have preserved the word-order of the original. In v. 127 we must read en pensed as two words, not the participle expensed, and ascribe the line to type 4, otherwise the fourth syllable of the line would be the first of a poly- syllabic word ; the sense of the passage supports this interpretation. In v. 766 we have a choice between

AD Volt Deus qu’a vus cunrei 6usum and BE Volt Deus que cunrei ciissum. The dissyllabic pronunciation of o«swm must be accepted as that of the author; B and E independently have made it trisyllabic, and have omitted @ wus in order to do so, but in the process they have destroyed the medial break. In v. 426 we find three readings, all equally satisfactory from the point of view of meaning: A E ta feste demain 1 fras B E ta feste demain feras DE Ta feste demain i feras. Metrical considerations make it certain that A has preserved the original reading ; the common ancestor of D and E, in order to efface the dialectal form /ras, has remodelled the line, but has destroyed the medial break in so doing. Other instances will be found in vv. 6, 92, 134, 420, 476, 1248, 1319, 1374, 1436, 1462, 1659, 1688, and 1760. Thus not only the addition of a syllable to the feminine lines, but also the non-observance of the medial break, reveals the fact that a line has been tampered with.

§4. THE COUNTING OF SYLLABLES

(2) Elision and hiatus. The Brendan shows the same hesitation (or rather licence) with regard to elision as many other French texts of the first half of the twelfth century.

1 On rhythm as an aid to textual criticism, see A. T. Baker, in Mod. Lang. Rev., xi, p. 436.

Versification xlv

. Monosyllabic words.

Definite article. Masc. nom. sg. 4 normally undergoes elision (at least fifty instances), though often written in full; it creates hiatus in eight instances: / ades 137, 457, 503 (?), 1098, 1205, “i uns 783, li hom 957, & hermites 1543. Masc. nom. pl. & occurs four times before a vowel, and in each instance causes hiatus : & altre_oisel 519 A, lf afamét 797, ii otsel 856, dt altre 1418.) Masc. acc. sg. /e always undergoes elision, though often written in full. Fem. sg. /a always undergoes elision, though occasionally written as /e before a vowel, e.g. /e arbre 496 A, le unde 1232 A, le une 1235 A (cf. Appendix II, §§1 and 6 e).

Personal pronouns. Jo undergoes elision (or synizesis ?) in jo_osase 1243, but not in jo en vols 432, fo ai 1543. Me, le, se always elide their vowel, though sometimes written in full; note ¢vaist Sarere 936, jelet [amunt 1236, and Jaisez /’ict 1475, with the pronoun placed after the verb. Dat. sg. 4 undergoes elision before e# (preposition and pronoun) in vv. 311, 414, 1612°; it creates hiatus in / enverat 130, Uf est 1143, 1593.

Demonstrative. (Co undergoes elision in cest 546°; it creates hiatus in ¢o0 ad 9g, ¢0 ert 110, ¢o enprist (not enpreist with B) 114.

Relative pronoun. Quis regularly causes hiatus (seventeen in- stances) ; the apparent elision of # (or synizesis of # and é) in gui ert 6 (AC) must be corrected on metrical grounds (cf. gué eri 40, 42, n't ert 228). On the apparent instance of elision in v. 728: lcest muster que_uncore & est, see para. ili. On qui t>qui see section ¢. Que (masc. and fem. acc, sg. and pl., neut. nom. and acc. sg.) undergoes elision in eighteen instances : VV. 127, 196, 300, 312, 354, 363, 466, 548, 652, 755, 930, 1065, 1201, 1279, 1308, 1310, 1501, 1732; it creates hiatus in nine Instances: vv. 108, 345, 374, 436, 816, 919, 980, 1564, 1832. The same hesitation is shown when relative gue is used as a temporal conjunction: vv. 1775 (elision) and 1837 (non-elision). Que in v. 38 (que ert, with hiatus) is nominative, but has probably been substituted for gui by the scribe of A.

1 Similarly in Roland, in the works of Ph. de Thaiin, in the Life of St. Osith (cf. Mod. Lang, Rev., vii, p. 93), and other twelfth-century works.

3 Perhaps also in Prierent [en 115; see the note.

* Perhaps co'st, with aphaeresis, in the original ; cf. ja's¢ 1248.

xlvi Introduction

Prepositions. De always elides its vowel, though often written in full; in v. 1759 the reading «/ é3¢ de or (A) requires a cor- rection. On a followed by a-, see section ¢.

Negative particle. Ve always undergoes elision, though sometimes written in full; ne ¢ fargez 1659 A requires a correction.

Prefix. In rea/at 956 the e is naturally elided.

Conjunctions. Si or se (<sé) undergoes elision in v. 414; in v. 411 DE it makes hiatus, but the reading has been rejected. Sz (<sic) creates hiatus: vv. gt1, 1222 (s# £4), 1373, 1421, 1614. Ne (<mec) undergoes elision in vv. 760, 1246, 1260, 1336, 1747, 1762, but not in wv. 234, 742, 1502, 1677, 1719, 1790." Que (of whatever origin) normally undergoes elision, though often written in full; it creates hiatus in vv. 183, 372, 876, 1120.

ii, Words of two or more syllables, ending in unstressed e.

The treatment of such words shows remarkable consistency. Between the fourth and fifth syllables no elision is tolerated, else- where elision is compulsory. Thus we find :

En mer halté un grant piler (v. 1068),

A feste_altré en trestut l’an (v. 1319), but

Veient terre halte_e clere (v. 483),

E tu e tuit cil altre_oisel (v. 519).

The lines with medial hiatus are vv. 31, 64, 425, 443, 449, 469, 691, 769, 788, 826, 831, 973, 1068, 1162, TrQy, 1225, 1281, 1286, 1312, 1319, 1321, 1409, 1449, 1487, 1554, 1604, 1644, 1673, 1692, 1733, and 1760 (thirty-one in all).

In some of these lines the hiatus could be set aside without difficulty ; thus in vy. 1286 we might read riches with B, in v. 1312 titel with B, in v. 1692 smaragdes with BDE, and in v. 1319 /feste and a/tre could be transposed with the support of BE; but none of these corrections is necessary, and some of them conflict with the context (v. 1692) or with the author’s principles of versification (v. 1319).? Noteworthy examples of elision within the hemistich are cizge_alumer 326, Punde_arere 1232, ne change_enfern (1st pers. sg.) 1357, port? en od tei (imperative) 1607. In v. 842, where A and B have éde/e

1 Classed by G. Rydberg (Zur Geschichte des franzdsischen 2, Upsala, 1896 ff., Pp. 99) as ‘ logical hiatus’, i.e. due to an effort to bring out a thought or word more clearly,

? Cf. Calmund, Prolegomena..., p. 48.

Versification xl vii

i funt, the addition of p/us is required by the context and indi- cated by the Latin prose translation. The atonic termination -ef in verb-forms is never elided in this text.!

ii. Aphaeresis of initial e.?

This occurs in @ streis (future of estre) 863,° Ain 1072‘ and ja’st (A Jest, BE Ja est, D Jas est) 1248.5 The author may have written ¢o’s¢t in v. 546; cf. para. i. In v. 1666 the reading gu’unt, indicated by A (B is missing) has been adopted in the critical text; but D has gui est, E gui ert (both expanding a feminine line), and it is possible that the original reading was gui’s¢.6 In v. 728 we ought perhaps to read

Icest muster qui’ncore i est (ABD que uncore, £ qu’encore).’

Aphaeresis is possible also in vv. 414 and 1612, where the MSS. have /’en (= 4 en), but the original text may have had /s’” ; in v. 311, on the other hand, where en is a preposition, aphaeresis seems less likely.*

(6) Fall of internal vowel. In certain future and conditional forms an unstressed ¢ had already fallen from pronunciation in the language of the author, and though sometimes written did not count as a syllable *: /ras 426, 1044, 1603, fle)rat 367, 924, 1566, 1633, /res, Sreis 553, 878, 881, 883, truverai 433, 586, truv(e)rat 246, 412, 1772, resuscitrai 1567. The dissyllabic conditional fereit 43 A (B /rei?, —1) has been rejected in favour of faisei¢, which has better MS. support. The future form suffreiz 551 A has been lengthened to sufferets on metrical grounds and with the support of the other MSS. Apart from these instances, there is no sign that the author of the

' See ch. viii, § 3, Ixxvii.

? Cf. G. Rydberg, op. at., pp. 8a ff.

5’ Hence we may read (in accordance with the usual practice of the Anglo- Norman MSS.) se spandent 573, se spant 1174, me scorcent 1407, 1413, rather than s’espandent, &c.

* Cf. Roland, v. 3364; Ph. de Thaiin, Cumpos, v. 2059; Ls Coronemens Looi's, vv. 94, 104; Beroul, Zristan, v. 732.

5 Cf. Roland, v. 2001.

* Cf. Roland, vv. 6, 1276; Ph. de Thain, Cumpos, v. 2779; Beroul, Tristan, VV. 2192, 3518.

7 Cf. Birkenhoff, Ueber Metrum und Reim,.., p. 22; Calmund, Prolegomena..., p- 56; Roland, v. 38a; Horn, v. 1469.

® Yet cf. Ph. de Thatin, Bestatre, vv. 444, 2526, and p. xxxv of Walberg’s Introduction. In St. Leger, v. 42, fu l’namet is a possible reading, though rejected by G. Paris.

» Cf. ch. viii, § 3, xxxi.

xl viii Introduction

Brendan was prepared, like other Anglo-Norman versifiers from Gaimar onwards, to treat feminine ¢ as optional or ignore it altogether. In the following verb-forms, all spelt as in A, the e is purely graphic’: averez 227, 587, muverum 392, liverat 605, muverez 770, muveras 779, vivere 960, liverez 1292. LEntrin 443, 686, 814, 1645, 1683, and suz{e)rain, -ein (dissyllabic) 564, 1682, are derived by regular phonetic development from *integrinum and *supera- num respectively ; although displaced on the Continent by the trisy]- labic enterin (influenced by entier ?) and souverain (influenced by Latin ?), they are not to be regarded as contracted forms.?. On the addition of inorganic ¢ by the Anglo-Norman copyists see Appendix I, § 6f.

Certain words borrowed from Latin often retained (in spelling) vowels which were ignored in pronunciation ; thus in (e)spfirit 131, etc., the first # was unpronounced,’ while in ange/e 100, etc., the first ¢ was long retained to indicate that the g was fricative. Passibile 791 A for paisib/e may be a similar instance, but is more probably a scribe’s error. The post-tonic ¢ in vicforie and similar borrowed — words has no syllabic value.

(c) Syaizesis, or the contraction of two contiguous vowels into one syllable.

This occurs in the Anglo-Norman preterite and imperfect subjunc- tive forms dussent 657, sdusum 765, dtisum 766, Oustes 1121, pouse 1579, dus 1604, poust 1658, and déust 1708‘; also in aler for a aler 160, avaler for a avaler 872° and qui for gui i 600, 1338.° It is true that in cases like the last-mentioned Mall and Tobler considered that gue had been substituted for gui’; but there is no sign that the author of the Brendan replaced gui by gue in any other

1 Cf. Appendix II, § 6f..

2 Cf. W. Hammer, in Zéschr. f. roman. Phil., ix, p. 84.

5 The rime esp(#\rtt: gust 131-2 has better MS. support than espirt: uirt, given by C only; cf. the note.

* Cf. ch. viii, § 3, xl.

5 In other texts we find Javal for la aval, lamunt for la amunt, laler for la aler, a’mendement, &c.; cf. Suchier, Ueber die... Vie de Seint Auban, p. 29, and Reimpredigt, p. 107; J. Vising, Le Purgatoire de Saint Patrice, vv. 420, 440, 7O1, 715, and notes.

6 Similarly in Roland, v. 2833 ; in Guernes’ Vie de Saint Thomas (cf. Walberg’s Introduction, p. cxliii); in Beroul, Zrnistan, v. 2208; &c.

7 E. Mall, Der Computus des Philipp von Thaun, Strassburg, 1873, p. 34; A. Tobler, Vermischte Bettrdge, i?, p. 123, note; cf. T. A. Jenkins, L’Espurga- toire Seint Patriz of Marie de France, Philadelphia, 1894, p. 44.

Versification xlix

circumstance. Fe:mes 470 was probably trisyllabic in the original, but may have been dissyllabic.! On jyo_osase 1245 see (a) 1 above. No contraction has taken place in /a wu 1196, 1630, 1742, /a ¢ 1618, neis 331 or nient 985, 1482, 1642. On cheles 343 see the note.

(2) £aclisis, or the process of contraction by which an unstressed word loses its vowel (other than an initial vowel) and combines with a preceding word.

This takes place in the Brendan under certain well-defined con- ditions.

i. Preposition+ definite article. The prepositions a, de, and en make the usual combinations with & and ées, viz. al, del, el, as, des, and es. Before a noun in the singular beginning with a vowel we find the same forms, e.g. de/ or 312, al air. 975 ; as the MSS. normally divide the words in this manner, their practice has been followed in the critical text,* but in such instances the article really undergoes elision, not enclisis (cf. e /’air 255, not e/ air). Similarly when the article is feminine (e.g. a/ is/e 436, al estorce 1482, del altre part 1485) the apparent enclisis is graphic only.*

ii. Personal pronouns. The pronouns /, /, and /es undergo enclisis, but not the remainder. To write nem out for ne m’out in v. 1588 would be unjustifiable (and unsupported by the MSS.) ; cf. si me pendi 1274, ne me rendt 1305, ne me valt rien 1461, and similarly jo te cumant 1264, nuls ne se taist 376, etc. The combinations found are : (with 2) jol 649, 1479 ; fed 14 ; chil (= guil) 1206 ; sil 530, 918 ; nel 128, 435, 698, 748, 1232, 1478, 1766, 1788; quel 1264; az 260 (al guerre), 1684 (al faire) ; eissil 595 ;

(with 2s)‘ ned 1588; [wre/] 1774°;

(with 2s) quis, chis 212, 786, 898, g10, 1698 ; [¢os] (D ces) 1638 ;

1 Cf. ch. viii, § 3, xxxv.

2 Cf. T. A. Jenkins, La Chanson de Roland, Boston, &c., 1924, p. cxXvii, note.

3 Cf. A. Mussafia, in Zéschr. f. roman. Phitl., iti, p. §94-

* Rydberg, op. at., p. 495, was mistaken in asserting that enclisis of /¢ does not occur in OF. It is found in the tenth-century Passion, v. a14; in St. Leger, vv. I10, 112; and in Beroul, 7risfan, vv. 810, 3446. The example in Roland, v. 2668, has not been accepted by Stengel or Jenkins. There is no instance in the works of Ph. de Thatn. The phenomenon was of course common in Provengal.

5 L'urel semblet forment pore; urel = ure li (< hora silt). The contraction does not actually figure in any MS., but the variants clearly point to it. For enclisis after fem. ¢, cf. Alexis, v. 588; Roland, vv. 959, 1760, 3811 ; &c.

3351 g

] Introduction

sts 195, 358, 1214, 1346, 1785; mes 1332; gues 115; las (fa adv.) 11go.

In v. 1281 B and E offer the combination je/= je fa (la = l’offrande), but this is probably due to inadvertence ; the insertion of the pronoun is unnecessary, A and D having merely jo. Inv. 1788 nel probably represents not ne /a (da = melodie) but ne fe (de neut. sg., referring to v. 1787 as a whole). When a vowel follows, e.g. in sz aspirat 142, kil unt hatt 1030, nel ad menur 1240, the enclisis is apparent only. Enclisis has been avoided in é/a les 1117, st Uf ditt 1779.

(e) Words with double forms. The existence in OF of many alternative forms, identical in meaning but differing in the number of syllables, was a great convenience for metrical purposes.

In the Brendan we find many of the usual double (or triple) forms : cum and cume; orand ore; uncand unches; dunc, dunches, and idunc ; tloec and tloeches (cf. aloeces) ; grant (fem.) and grande ; tel, ited (fem.) and ¢ele; cf and tat; al and tal; cst and tast; ¢o and igo; /a and ila; tantand itant; probably also nu/ and neil (v. 411). The simul- taneous use of ere?, ere, and ert (as well as eséett) as impf. indic. 3 sg. of es¢re is worthy of note. The substantives dehart and dehaite are used without difference of meaning ; similarly ¢urment and furmente, cumpaine and cumpainte. For references, see the Glossary.

Cum(e) and fors could be followed by nouns in either the nominative or the accusative case. Masculine nouns and adjectives ending in e could take or omit nominative s. These and many other options facilitated the versifier’s task.

§5. THe Coupler

The Brendan was used by P. Meyer’ to illustrate the primitive form of the couplet (or pair of lines riming together). In early French texts the couplet of six- or eight-syllable lines conformed to rules remarkably similar to those laid down later for the Alexandrine by Malherbe. A couplet could consist of a single sentence, or of two or more sentences, or a sentence could be extended over two, three, or more couplets; but if the couplet contained more than one sen- tence, the end of the couplet must coincide with the end of a sen- tence ; and if a sentence was extended over two or more couplets, it must end with the second line of a couplet. Enjambment, i.e.

1 Le Couplet de deux vers, in Romania, xxiii, pp. 1 ff.; cf. G. Melchior, of. at., pp. 17 ff.

o

Versification li

beginning a sentence within a couplet and carrying it over into the next, or carrying a sentence into a new couplet and terminating it before the end of that couplet, was very rare in the oldest texts, and only became frequent after the middle of the twelfth century. In fact the treatment of the couplet, according to P. Meyer’s conclusions, furnishes useful presumptive evidence as to the age of a text. The strictness of the rules that governed the couplet in early texts is ex- plained by the fact that the couplets of six- and eight-syllable lines were developed from Latin lines of twelve and sixteen syllables respectively, with internal rime.

On closer inspection, however, it appears that the Brendan pre- serves the cohesion of the couplet less consistently than P. Meyer believed. Even in the passage which he quoted (vv. 1-38)? it is dificult to make his rules apply to vv. 13-18. In the rest of the poem a considerable number of enjambments occur, Thus we find (2) Couplets in which the first line ends a sentence, and the second forms a complete sentence by itself, e. g.

Vv. 395-8 ‘Il nus est douz e prest amis, Qui prestement nus ad tramis Dunt poiim sa feste faire. Pensez de la nef sus traire.’ Similarly vv. 107-10 (?), 469-72, 834-6, 1001-4, 1577-80, 1811-14. This is but a slight irregularity, and is known to occur early, e. g. in the first French version of Marbode’s Lapidary.

(4) Couplets in which the first verse forms a sentence by itself, and the second begins a new sentence, e. g.

Vv. 153-6 Puis les baiset Brandan e vait. Plurent trestuit par grant dehait Que mener ne volt lur peres Fors quatorze de lur freres.

Similarly vv. 187-90, 243-6, 1061-4, 1769-72, 1773-6. (c) Couplets in which each line belongs to a different sentence, e. g. Vv. 1039-44 Cum I’abes fist le servise,

Si cum la lei est asise, Chantout mult halt, a voiz clere. Dunc li dient tuit li frere:

* Beal pere chers, chante plus bas, U si go nun, perir nus fras...’

Similarly vv. 13-14, 439-40 (?), 833-4, 889-90, 1089-90, I111-12,

1 Op. ait., p. 13.

lii Introduction

1573-4(?). P. Meyer’ pointed out the occurrence of such couplets in Gaimar’s Estorte des Engleis.

The majority of the couplets in the Brendan undoubtedly conform to the normal practice of the period; but the numerous excep- tions clearly show a tendency towards greater freedom. In this, as in so many other respects, the author appears to have been in advance of his age.* He permits himself bold enjambments within the couplet :

Vv. 1525-6 Puis le baiset; ses cumpaignuns Dist qu’amenget: ne failet uns.

Vv. 1531-2 Cil li firent. Puis les menet A sun estre, lur enseignet.— or combines a couplet with a following hemistich :

Vv. 431-3 E puis irez en altre liu, U jo en vois e la vus siu, Mult pres d’ici; la vus truveral... See also vv. 177-8, 1141-3, 1267-8, 1271-2. The following pas- sage, if correctly punctuated, is broken up as much as any passage of Crestien de Troyes or Raoul de Houdenc : Vv. 335-7 Cil s'aparceut que l’abes sout Del larecin; cument il l’out Cunuit; atuz confés se rent... It would therefore be quite unjustifiable to condemn any reading in this text on the ground that it infringed the rules of the couplet, or otherwise to use the treatment of the couplet as an aid to textual criticism.

§6. THE RIMES

The author of the Brendan chose his rimes with care, though not all would have been permissible in standard French of the period. In cerne : verne 873, fort : gort 167, etc., he appears to rime ¢ with ¢ and g with 9 ; but in reality ¢ and 9 in such words had become open through the influence of the following y+consonant. Similarly no distinction is made between ¢ and ¢ when followed by /+consonant (e.g. eals : otseals 579). The combination in rime of « (< Lat. #)

1 Op. at., p. 13.

2 In Anglo-Norman texts of a much later period the couplet rule was still observed; cf. G. Paris, in Romanta, xxv, p. 533; A. T. Baker, in Mod. Lang. Rev., vii, pp. 176f.

Versification liii

and #, o (<V. Lat. 0), due probably to the fact that # had not become modified in the author’s dialect, is frequent in this text. The nasal diphthongs ain and ein are combined, but in feminine rimes only (e. g. chacines : semaines 869, meindres : graindres 1007). The nmes solet/: fedeil 581 and : pei? 1761 show the dialectal develop- ment of ¢ to ef before -/mouillé. The combination mien : soen 751 seems to be a licence, but can be paralleled from other texts’; the diphthong was in both words an ascending one. The adjective rustes 41 (: syustes) was evidently unfamiliar in this form to the copyists, three of whom substituted the more usual ruéstes (accompanied in E and perhaps in D by the substitution of vistes for justes) ; but either form is defensible. As regards the treatment of consonants, we may note the dialectal form doche 1217, riming with roche ; the confusion of final #s and after a vowel (feisun : cisum 835); and the combi- nation in rimes of /and /mouillé (e.g. peril : costil 429, travatls : calg 1177), a and #-mouillé (e. g. plein : desdeign 235, feignent : pet- nent 215) respectively. In England the sounds /-mouillé and 2-mouillé were difficult to maintain, and early became depalatalized. Apart from these instances the author of the Brendan did not, like many OF writers, tolerate dissimilarity of consonants in the feminine rimes. The phonetic problems involved in the above-mentioned rimes will be dealt with more fully in chapter viii. Other rimes treated as inaccurate by previous writers have been justified in the present edition.

In a few instances, for the sake of rime, the author availed himself of double forms existing in the language. He used en? (: defent) 295 by the side of en (: sen) 1096; ere (: frere) 442, 992 by the side of ert (: pert) 1648, 1676; Brandain (: vilain) 164, (: main) 203, 659 by the side of Brandan (: pan) 481, (: an) 825°; possibly also parais (: pats) 1822 by the side of parei's (: ques ?) 546, (: poesteis) 1651, (: regueys) 1795.5 Note also afpostérie (: glorie) 1038 by the side of afostoile 8 ; but the latter is not in rime.

The predilection of the author of the Brendan for rich and more particularly double or ‘leonine’* rimes was pointed out long ago by I. Freymond,’ and was discussed in somewhat greater detail by

1 Cf. ch. viii, § 2, xl. 3 Ibid., xi. * Ibid., liii.

* The term is used here in the sense adopted by Tobler ( Versbas®, pp. 134 £.).

S Uber den reichen Reim bei altfrans, Dichtern lis sum Anfang des XIV Jahrhunderts, in Ztschr. f. roman, Phil., vi, pp. 1-36 and 177-915; especially p- 180.

liv lntroduction

Birkenhoff.! So far as we know, the author of the Brendan was the first French poet who enriched his rimes intentionally, and for several decades he seems to have found no imitators. The examples of rich rime found in Gaimar, Beneeit de Sainte More, Wace, Marie de France, etc. are probably accidental ; Crestien de Troyes was the first courtly poet to use such rimes freely. The proportion of rich rimes in the Brendan, though not of course so high as in the works of Gautier de Coincy, Rustebeuf, Baudouin de Condé and other authors of the thirteenth and fourteenth centuries, is remarkably high for so early a period. At the beginning of the twelfth century the process of dropping from pronunciation consonants following the tonic vowel had not gone very far; consequently there were few rimes based on homophony of the vowels alone, and it was quite unnecessary to extend the homophony to the preceding consonant. As Freymond pointed out,’ rich rimes are specially frequent in didactic poetry, and the use of them (especially when leonine) in the earlier period is a sign that the author was imitating medieval Latin verse. This confirms the opinion already put forward, on account of the peculiar metrical form of the Brendan, that the author of the work derived his system of versification from Latin.

The rich rimes in this text are unevenly distributed. Here and there extensive passages will be found to be almost devoid of them, whilst other passages present long series of rich and leonine rimes with hardly any ordinary rimes among them, e.g. vv. 739-60, 1273- 1312, 1793-1806. In feminine verses rich rimes (according to the modern definition) are much less frequent than in masculine verses, being less needed (for in feminine rime-words there already exists homophony of the final syllable) and more difficult to devise. It must however be remembered that certain medieval theorists® re- garded a// feminine rimes as leonine, and that such a view would be specially applicable to the Brendan, in which the same syllabic value is given to feminine as to masculine endings. According to the authors of Zas Leys d’ Amors,* O.Prov. natura : noyridura was a

1 Ueber Metrum und Ream..., pp. a3ff.; cf. Calmund, Prolegomena.. ., PP. 75-7:

2 Op. cit., pp. 179 f.; cf. pp. 13 f. 3 Cf. Freymond, pp. 4 ff.

‘ Edited by Gatien-Arnoult, in Monumens de la littérature romane, i-iii, Toulouse, 1841-3. The passage defining rims leontsmes is in vol. i, pp. 160 ff. ; it is missing trom the recent edition by J. Anglade (Toulouse, 1919), owing to a lacuna in the MS. followed.

Versification lv

leonine rime like gwerriers : derriers ; hence in the Brendan rimes such as eve : creire 147 would be on the same footing as pend? : menct 249. The present investigation is based, however, on the more usual definitions: the rime is regarded as rich when the homophony extends to the consonant preceding the tonic vowel, as leonine when it extends to (or beyond) the vowel preceding the tonic syllable." ‘Broken’ rimes (e. g. fast unt : estunt 403, rendt : en di 1305, as tu: Das fu 1327) are placed on the same footing as unbroken. Rimes extending to the consonant preceding the tonic vowel, but not to the whole syllable (e.g. prent : rent 73, etsil : cil 561, mescreire : lur etre 983, dammes : flammes 1131) are counted as rich. Those in which voiced and unvoiced consonants are combined (e. g. orguil : escuil 67, tutt : duit 265 *) can hardly be classed as rich, though the author may have intended to reinforce the rime in some of these instances. Rimes showing identity of the penultimate vowel, but not of the consonant that follows it,° are in this text so few that they are almost certainly accidental ; they are not included in the following computa- tion. Complete lists of the rich and leonine rimes will be found in ch. viii, § 1.

Rich rimes of the ordinary type are not very numerous in this poem. They may be classified under two headings :

Entire, 1. e. including the whole tonic syllable. Masc.: 30 instances, e.g. guarnir : venir 299, tut el: tel 963, aprés : aprés 1781. Fem.: 10 instances, e. g. fedes : taceledes 493.

Incomplete, i.e. not including the whole tonic syllable. Masc.: 18 instances, e.g. estrett : dreit 169, Pair : esclair 499 eistd : cil (¢ = ts) 561. Fem.: 12 instances, e. g. /amme : flamme 1147. Total: 48 masc.-+ 22 fem. = 70.

Leonine rimes are much more plentiful. They may be classified as follows :

Extending to the vowel preceding the tonic syllable.‘

1 So Tobler.

? So vv. 169, 309, 541, 693, 759, 783, 1335, 1455. 1495, 16ar, 1811 (all masc.).

3 Called ‘double rimes’ by Tobler (Versbau®, p. 163), ‘assonance-rimes’ by Stengel (in Grober’s Grundriss, ii, 1, p. 66). Possible examples are sculante: curante 177, § sunt: 8 sent 717, suti : fst il 1077, amestistes : les listes 1693.

* Some half-a-dozen disputable instances have been included, notably veabies : deables 341, les tistes : amestistes 677, travailes : apareiles Bag.

lvi Introduction

Masc.: 158 instances, e.g. vetheir : setheir 55, dais : pars 617, malfez . esthalfez 1137.

Fem.: 12 instances, e.g. susurnerent : turnerent 327, arbalesie : galeste 1155.

Extending to the whole syllable preceding the tonic syllable.’

Masc.: 8 instances, viz. exvetat: aveiat 139, temples : sustentez 201, mesters : mesters 285, savum : nus avum 743, manddl : cumandét 831, 881, saveir : repos aveir 1439, aelicius : precius 1743.

Fem.: r instance, viz. perisetent : guariseient 1465.

Extending to the second vowel before the tonic syllable.

Masc. only: 5 instances, viz. heritét : deseritét 51, tenablement : veablement 50, deseritél : veritét 535, prestement : vestement 959, verablement : creablement 1219.

Extending to the whole of the second syllable before the tonic syllable.

Masc. only: 1 instance, viz. avistuns : divisiuns 1783.

Total: 172 masc.+13 fem. = 185.

Thus out of a total number of g20 rimes, 255 (27:7 per cent.) ? are rich, and of these 185 (20-1 per cent.) are leonine. Were we to adopt the point of view of Zas Leys d’Amors, and class all the femi- nine rimes as leonine, the number of rich rimes would be 519 (56-4 per cent.), and of leonine rimes 471 (51-2 per cent.). If we take the masculine rimes alone, the rich rimes number 220 out of 621 (35-4 per cent.), the leonine mmes 172 (27-7 per cent.).

The striking feature in these figures is not the number of rich rimes, but the number of leonine rimes. The high proportion of the latter, the inclusion among them of broken rimes,‘ and the tendency to group them together, all indicate that the author of the Brendan introduced them of set purpose, most probably in imitation of medi- eval Latin verse. This point being established, a further analysis of the rich or leonine rimes, based on the grammatical or etymological or semantic relationship of the words rimed together,‘ would serve no

1 The following rimes might be transferred from the first to the second class : truvum :ruvum 401, 747, prees i crees 1203, plentet: volentet 1769, truvat: ruvat 1829; but the homophony does not extend to the whole of the last two syllables.

2 Freymond, of. ot., p. 25, computed the proportion of rich rimes to be a7 per cent., and Birkenhoff (p. 23) agreed with him. Note that Freymond counted feminine rimes of the ledes : faceledes type as leonine.

3 Nearly always a sign of artificiality; cf. Freymond, p. 181.

4 Freymond, pp. r9 and 25; Birkenhoff, p. 27.

Versification vii

useful purpose. It is obvious that, in order to obtain so many leo- nine rimes, the author had sometimes to combine two similar verb- forms or two words derived from the same stem ; but on the whole his choice of rime-words shows surprising freedom and ingenuity.

The leonine rimes have sometimes been of considerable assistance in determining the choice of reading. Other things being equal, the reading which offers a leonine rime is the more likely to be that of the orginal. Thus in wv. 61-2 the rime saveir : aveir furnished by BDE is preferable to vetheir : aveir furnished by A; similarly deoum : recevum 757 with A, rather than chantum : recevum with DE, and a/at : devalat 885 with B, rather than ala: devolat with A. In v. 1610 the unique word sururer, given by A and E, is guaranteed by the rime with demurer, against surjurner offered by D. In v. 798, where all four MSS. present entirely different readings, that of B has been preferred, not only because of its intrinsic superiority, but also because it makes the leonine rime afameé? : amét.'

The correctness of the rimes has sometimes been obscured by the erratic spellings of the Anglo-Norman scribes, including those of A (adopted, so far as possible, in the critical text). Thus vé/s: amis 45, raps: dras 461, nefs:reis 1681 are good rimes, in spite of the reintroduction of the stem-consonants by the scribes ; vee : otde 99, Dedithes : niies 213, nots : podes 941, in spite of the graphic preser- vation of the intervocalic dental ; also, for various reasons, /eé : fat 69, meinge : prenge 119, dulceur : seignur 691, men: son 751, Judus : fius 1291, estot : poet 1637, vere : materie 1679, and cardunt : fusun 1745. Again seez : veres 359, esmaez : atez 365 and pluius : annuus 975 are leonine, in spite of the suppression of intervocalic #; so also are pre(s)ez : creez 1203, ple(!)ners : suveners 1569, and plentét : volun- ted (corr. -entet) 1769.

In the few instances where a word has been rimed with itself, either it has two entirely different meanings (mes¢ers 285, en 369), or the addition of the preceding syllable produces a normal leonine rime (gui es: a es 1265, as tu: vas tu 1327, t est: n't est 1381). When two consecutive couplets have the same final stressed syllable, they are differentiated as leonine rimes by the preceding syllable (enceiat : aveiat, trat : aspirat 139-42, verrez : crerrez, crendres : pren- drez 475-8, and similarly vv. 755-8, 1285-8, and 1373-6); the only exception is seisante : nonante, cinquante : quarante 1595-8.

1 Cf. also the choice between quesis and queis 545, between aasmat and aesmat 1054, and the difficult passage 477-8. 3361 h

lviii Introduction

The practice, common in certain OF texts,’ of repeating a sentence in a different order for the sake of obtaining another rime—a temp- ting convenience to the versifier, but liable to degenerate into mere padding—was not followed to any appreciable extent by the author of the Brendan; in vv. 62-3 and 1333-5 the repetition could hardly have been avoided.

IV. CLASSIFICATION OF THE MANUSCRIPTS?

Four MSS., ABDE, contain the complete text. Each presents many errors not found in the others; in each a number of lines are missing, but the omissions seldom overlap; none can therefore be derived from any of the others. There is ample evidence to prove that A and B are related on the one hand, D and E on the other.

§1. THE Group AB

In the following instances, A and B have the same corrupt reading, while the reading offered by D and E is irreproachable : °

43 A fereit, B freit (-1); DE faiseit. Either tense will give satis- factory meaning, but our author would have used the contracted conditional form /vert* ; moreover, the imperfect has the independent support of C and L.°

97 AB de/ isle; D cel isle, E cts slles. ZIsle must be the subject of the sentence.

1 e.g. Le Mystere d’'Adam, especially vv. 315 ff.

2Cf. M. Wien, Das Verhdlinis der Handschriften der anglonormannischen Brandanligende, Diss. Halle, 1886; H. Calmund, Prolegomena..., pp. 12 ff.

5 The attacks made in recent years (e.g. by J. Bédicr in his introduction to Le Lai de Ombre, Paris, 1913, and by Dom Henri Quentin in Essats de critique textuelle (ecdotique), Paris, 1926, pp. 36 ff.) on the use of ‘common errors’ as a means of classifying MSS. do not materially affect the present chapter, because a close study of the metre, rimcs, phonology, morphology, and style of the Brendan, and a comparison of its text with the Latin original (see ch. v) and the Latin prose translation (see ch. vi), furnish reliable criteria for detecting corrupt readings, i, e. readings which cannot have been so written by the author. MSS. which share a long series of palpably corrupt readings must inevitably descend from a common ancestor other than the original.

4 Cf. ch. viii, § 3, xxxi.

5 L is the Latin prose translation; cf. ch. vi. N is the Navigatio Sanch Brendani; cf. ch. v.

Classification of the Manuscripts lix

254 AB entrer; DE munter. Entrer is not a suitable word to use with reference to cliffs, but was substituted for munter probably on account of entrethe 251.

384 A de le nager se tart(+ 2), B den. sen t.(+1); DE del n. tart.

500 AB & clair (!); DE Pesclair (: air), Cf. 1764.

556 AB£n sunarbre; DE Ensuml arbre. Cf. Lsuper arborem. In 886 A alone commits the same error.

609 AB daistt sen sunt (en from 610); DE 34. se sunt.

870 AB set semaines; D oef s., E .vitt.s. From Easter Day to the octave of Pentecost is eight weeks ; cf. 866, the words dous meis in 590 and 1630, and octo sepitmanas in L.

1020 AB far mer (repeated from 1019) ; DE mult cler. 1172 AB cent (due to misreading of oén¢); D otent, E oent.

In other cases A and B have the same syllable added or omitted,

while D and E are metrically correct :

246 AB ¢o que (+1); DE que.

724 AB ces/(—1); DE icest.

969 AB omit ¢ (—1). 1098 AB lades (—1); DE & abes. 1383 AB cume (+1); DE cum. 1397 AB Dunc (—1); DE Dunches. 1521 AB £ist(—1); D LZissit, E iss¢ (with alteration of word-order).

Elsewhere A and B have a common error, while the correct reading is given by either D or E: 108 AB gu’i/(—1); D que i (cf. C), E alters. 353 AB Cust fud un(s) (—1); D Last fut un, E Ce fu li uns. 572 AB omit Halt (—1), correctly given by E ; D alters. 588 AB As ustaves; D Jesque as u.(+1), E Les octaves. The correct reading, indicated by D, is Desque uitaves ; cf. N and L usgue ad octavas, and 866. 752 AB Pur le super (+1); D alters, E Por soper. 1239 AB Zurment grant (—1); D Tormente g., E Torment a g. 1570 AB omit erv# (—1), correctly given by D; in E the line is missing. 1576 AB pleiner past (—1); D mult pl. p., E plener repast. 1696 AB eod(+1); De, E of. Mention must also be made of instances in which AB and DE present different readings, both erroneous : 303 AB cum lur plout(—1); DE cum il lur plot, Both readings

Ix Introduction

are presumably erroneous for cume lur plout; the introduction of #7 as subject of an impersonal verb is contrary to the author’s practice.’

411 ABE sé lur falt nule rien(—1); DE £ se tl lurfautn.r. The original reading was probably £ sé /ur falt neiile rien.

800 AB Cume se (+1); DE Cum sele (+1). Both erroneous for Cum se.

842 AB dele (—1); DE mult dele. Both are probably erroneous for plus dele, cf. L pulcrius.

924 AB ferat; DE en ferat. The latter, though apparently hyper- metric, represents the original reading en frat; in A and B en has been omitted.

It is probable, however, that in some of these lines (vv. 303. 411, 842) the archetype was already corrupt. The same may be said of 1390, where AB have ze in error for se, but D and E have made independent alterations.

§2. THE Group DE

In many instances, D and E have a corrupt reading in common, while the reading of A and B is satisfactory :

51 ABest; DE fut. The present tense is required, cf. 1654.

147 AB £ di(s)t lur ad de seon eire; for de, D has tut, E frestot (+1). The use of ‘vf is inadmissible; the voyage is still only a project, and Brendan cannot disclose all its vicissitudes to his followers.

249 AB Qui; DE Xarz, connecting the line erroneously with 248.

295 AB Portes ; D Pernez, E Prenés, influenced by prendre 296.

311 AB Mist len talent; DE M. i 4. The preposition en is necessary, cf. OF avotr en talent.

365 AB nen esma(ijez; D ne vus maez(!), E ne vos esmaics (+1).

401 AB ruvum (: truvum); D pernum, E prendon, destroying the leonine rime.

528 AB De(s) vertuz Deu nus doust (A dut) paistre; DE omit Dex and replace doust by devett, Deu is necessary to the sense ; cf. L.

539 DE omit sé before cume; but sé is necessary for metrical reasons,

1 Cf. ch, viii, § §, xii.

645

761-

848

985

1072

1118

1190 1245 1316

1682 1705

1712

1807-

In

Classification of the Manuscripts xi

AB funtaine ; DE funtaines. The singular 1s essential, cf. dudle (: frudle).

2 AB esteint: peint; D esprent: peigni, E esprant: paint, destroying the rime.

AB Mais dunc lur nef rechargerent (: targerent); D Tus ala mer se vepeirerent, EE ad’ilueques Sen repatrierent, The introduction of the reflexive pronoun by D and E makes the line hypermetric ; while the use of refairier destroys the leonine rime.

A nent (—1), B nient; DE vient, through confusion with vient 986.

AB Riches estreit hi'n (B ki en, +1) fust sire; DE R. esteit ki en fu s.(+1). The change of tenses ruins the sense of an excellent line.

AB bien les enseignet; D tres bien le seignet, Ef. b. les saine. The use of seigner ‘to make the sign of the cross over’ anticipates 1123. -

A Zas,B La les(+1); DE Zes. Za is necessary to the sense.

DE omit veis, which must accompany de majestét; cf. L.

AB Mes grans peines n'ai (A n't at, +1, B n'ai so, +1) dunc mie; DE De mes g. p. n'ai jo mie, destroying the medial break.

A Fatters, B Fattres; DE Meistres, an unsuitable word.

AB desus estait; DE estett desus. A present tense is required, and ester is preferable to esére.

DE add gut, repeated from 1711.

8 AB alat: portat; D pris: pareis, E pris: paradis, anticipating the mme-words of 1809-10.

the following instances, although the reading of DE is not

manifestly incorrect, it is inferior to that of AB: 26 AB ne savrat; DE n’en s. The unnecessary en is probably

due to 25.

151-3 AB Cumandel... baiset; DE Comandat... baisat. The

present tense is used in the preceding and following lines.

207-8 AB levet... s(e)ignet; DE levat...seignat. Same as 151-3.

445 AB Puis que unt (A out) tut fait; D P. que unt f, EP. guwil ont f., both omitting ¢¢. D introduces an unusual hiatus, E adds a superfluous pronoun.

886 A dum, B dunce (= dunt); DE sé, probably influenced by 885.

I xii Introduction

920 AB Pur grant turment plus ne stuvett; D Plus g. t. pas ne stuvett, E Plus g. t. n't estovoitt. 1120 Aa enjern, Ben ¢.; DE pres d’e. The use of pres is probably based on the knowledge of what follows. 1178 AB freis (pl.); DE fret. Cf. calz (pl.) immediately follow- ing. In other cases D and E have a common error, and the correct reading is given by either A or B, but not both:

426 A E fa feste demain i fras,B E taf. d., feras; DE Jaf. d. Jeras. Cf. p. xliv.

612 A tendre, B prendre; DE tenir. Tenir sun curs is ‘to con- tinue a voyage already begun’; here ¢endre is the verb required.'

688 A Stant cler[e]s sunt les haspes (: jaspes), B missing ; D Lestees erent tutes les capes, E Event listees bien les chapes. The lines of D and E are hypermetric and have no medial break, while the rime jaspes: capes is imperfect.

716 A de leu (for lu = lui), B de lus; DE del liu, repeated from 715. Cf. L esse ejus.

757 A devum, B levum (: recevum); D Pye E cantons, destroy- ing the leonine rime. Devum is supported by L.

988 A /ur (= Ja lur), B da (i.e. fa part); DE Zé, referring wrongly to peisun 986.

1054 A asinat (representing asma?), B aasma; DE aesmat, destroying the leonine rime.

1584 A [Que] n’ot mester de beivre rien, B missing; DE replace betvre by nule. The point is that the fish brought by the otter served as drink as well as food; cf. N and L.

1688 A d’tsselites, B de par eslites; DE pieres eslites. The latter reading, besides being hypermetric, breaks the connexion between éssedites (or eslites) and grisolites 1687.

1779 A divisit (for -sef), B missing; DE & devise, with elision of unstressed -e¢ before a vowel, contrary to the author's practice.

1781 A Vait al avant e cist apres, B missing ; for ast, DE have av, The use of a/ for distance and as¢ for nearness was normal in twelfth-century French.’

? See the note to this line.

2 Cf. K. Nyrop, Grammaire historique de la langue fr., ii, § 566. In 609, where ABD have cil ¢ ai, the question of relative distance does not arise.

Classification of the Manuscripts xiii

1789 A Lur nature ne poet prendre, B missing ; DE substitute fo(:)ent for poet, adding a syllable, but leaving nafure in the singular.

Lines in which both AB and DE present erroneous but different readings were mentioned in the preceding section. On lines in which both have the same erroneous reading, see § 5.

In nearly a hundred instances D and E employed the same device (not adopted by A or B) for adding a syllable to the feminine lines.’ The alterations of this kind in wv. 1316, 1688, and (1789)* have already been dealt with, on account of the common errors to which they have given rise. The remainder may be briefly summarized as follows :

Introduction of hiatus: 53, 521, 715, 1039.

Addition of a monosyllable : article, 54, 225, 488, (666,) 709, 726, 780, 905, 973, 1001, 1027, 1419, (1646) ; conjunction, 233, 352, (604,) 989, 1302, 1441, 1487; preposition, 593, (685,) 779, 1085, 1086, 1691, 1692, (1747); adjective, 145, 984, 1326; adverb, 156, 214, 342, 622, 1164, 1465, 1730; pronoun subject, 397, 725, 800, 1214, 1448; pronoun object, 1116, 1294, (1627,) 1750; substan- tive, 207.

Substitution of a longer form of the word: n’] nen 254, 420; S | set 936 ; cist | tcést, etc. 334, 791, 1037, 1349, 1544, (1790); Msit] trestut 573, 710; sur] desur 554, 840, 874, 884; unc] unches 1684; entrin | enter(r\in (686,) 1683 ; fort | forte 628 ; fait} fatte® 1241 ; addition of prefixes in revirent 838, detraire 1205.

Change of tense : (221,) 240, 251, 569, 645, 672 (D has inadvertently dropped 2/), 1255, 1481; erent | esfetent 389.

1 Cf. ch. iii, § 2. * Parentheses placed round verse numbers in this paragraph indicate that the line is missing in B. 3 Presumably for fastes, agreeing with the direct object following (enxpeintes). * Those lines have not been included, in which (a) D and E probably intended to introduce hiatus, but apparent hiatus already exists in A and B: 988, 1118, 1248, 1293, 1340; (6) D and E have added the same syllable, but without necessarily increas- ing the total number of syllables : 528 (devest for doust, but Deu dropped), 1118 (tres added), 1a7o (¢ added after rore) ;

Ixiv Introduction

There is no doubt that in many, perhaps even the large majority of these lines, the alteration is a sufficiently obvious one to have suggested itself to two scribes or revisers independently. The addition of an article, the substitution of sa for ct, of estetent for erent, etc., by scribes who made a point of expanding the feminine lines, can be explained without postulating a common source; especially as we find that the scribes of A and B have frequently hit upon the same alteration as D and E.!_ But the coincidences between D and E are so numerous that they must to some extent have been derived from acommon source. The changes in vv. 899, rr1r7, 1205, 1214, 1241, 1316, 1614, 1688, 1713, and 1789 cannot all have been introduced independently.

§ 3. AB acainst DE

The passages in which one of these groups can be shown to have a corrupt reading have now been dealt with. ‘There remains a large number of instances in which the reading of AB differs from that of DE, but neither can be definitely condemned as corrupt.

Often there is nothing to choose between the two readings: 19 AB (seinz) Deu, D (s.) Auem, E (bons) hom*?; 95 AB U, DE La; 143 AB e certement, DE certeinement; 167 AB £E suz, DE Desuz; 193 AB a tet, DE od tet; 198 A celet, B ceille, DE celat; 216 AB de nager, DE del n. (cf. 384) ; 220 AB Desque, DE. Zant gue; 234 AB devrunt, DE detvent; 237 AB dur vitaile, DE la v.; 296 AB go, D jol, E sel; 297 AB pur set, DE par set; 354 AB la nef, DE sa n.; 648 AB Di(s)t lur Pabes, DE Dist fi abes; 650 AB Dia, DE De at (so 779, 1792); 650 AB parlé od gent, DE p. a. g.; 831 AB dur mandet, DE le m.; 946 AB cez, DE dur; 957 AB mais desperer, DE desesperer ; 992 AB vos, DE nus; 996 AB Quelque semblant qu'il nus mustrast, g¥#’ not in DE; 1074 AB As funs, DE Alfunz; 1103 AB s’en feignent, DE se f.; 1114 AB erent (impf.), DE serené (fut. ?); 1227 AB ert, DE fu; 1237-8 AB desus: dejus, DE dejus: desus; 1241 AB les

(c) Dand E have added the same syllable, but in a different manner: 38, 895, 1156;

(@) A or B (usually the latter, sometimes both) have made the same alteration as DE: 259, 260, 292, 333, 441, 442, 494, 495, 553, 588, 674, 719, 839 (?), 865, 866, 893, 949, 1070, 1072, 1269(?), 1350, 1420, 1453, 1598, 1714, 1740.

1 Vv. 292, 588, 866, and 1598 may already have had an extra syllable in the archetype; cf. p. Ixxviii.

* In vv. 19, 115, 143, 193, 198, 216, 220, 2397, 287, 296, and 297 the reading of AB has the support of C; cf. § 6.

Classification of the Manuscripts Ixv

empentes, DE ses e. (ses from 1242?); 1293 AB as gabs, DE a gas; 1302 AB ne, DE n’en; 1342 AB sett, DE est; 1345 ¢ in AB, not in DE ; 1369 AB dost, DE fut; 1430 AB pur pot, DE a@ f.; 1472 AB e grant peril, DE e od p.; 1683 AB ert entrins, DE enterins ert (+1); 1726 AB aun, DE nums; 1823 AB & parent, DE st p.

The same remark applies to the following lines, which support the grouping AB against DE, in spite of minor divergences in the MSS. of one group: 71 DE introduce ée/, 287 st, 452 wus (cf. 365); 115 DE omit en, 1032 Ja; 548 AB de mer fe han, D grant ahan, E tcest hahan; 811 A curent (for cureit), B currett, DE curut; 897 AB Trestout (B -tuit) curent, D Mult c. tost, E Molt tost ¢.; 1145 AB Jetant flammes, D E jetat flambes (+1), E Geta grant flame (+1); 1694 AB Forment lutsent, D Mult resplendent, E M. resplendoient (+1); 1797 A cent mil fant (—1), Bc. m. itant, DE cent milije tans.

Often, however, although both readings make good sense, it is clear that the reading of AB is more archaic or more unusual than that of DE. When we find AB di, DE jur 592; AB nepurtant, D nepur- guant, E nonporquant 1103; AB oc, D ¢o, E ce 27, 243, 979, 1467; AB cum ‘when’, DE guant 228 (influenced by 227?), 241, 309; AB achant ne prent, 1D tute ne fent, E parmi ne fent 1018; AB de smaragde, D d’esmareldes, E a’esmeralde 1083 (cf. 1692); A un nud fed, Bun fed (—1), DE un povre 1456, etc., we can only conclude that the text of DE has passed through the hands of a reviser who has made sporadic attempts to get rid of uncommon and old-fashioned words and expressions. In a similar manner antiquated or dialectal verb-forms are often eliminated in DE; thus murget 61 is replaced by muret D, mutre E,*? munto(u)t 496 by muntaé (cf. 1585), estret[e|nt 620 A by #7 serunt, respunent 965 by respundent, s'aparcout 1141 by se parcut D, Sapercut E, oustes (B ustes) 1121 by edstes (cf. 917, 1579, 1604, 1658). ret 46 (represented by ert A, tert i/ B) is replaced by esteit in DE. The pronoun #/ has been introduced in DE as personal subject in 21, 966, 1297, and as subject of an impersonal verb in 303, 411, 1399. In116and gg9 the tonic possessive, used adjectivally, is replaced by the atonic form. The word-order has been rejuvenated by DE in 378, 402, 1261, 1462, and 1780. We may therefore con- clude that, other things being equal, the group AB is more likely to represent the original than the group DE.

2 L has parentes gus, but this is hardly a sufficient reason for preferring the reading of DE. > Cf. 1764 A tolget, DE torllet.

335% i

Ixvi Introduction

The MSS. frequently fall into the same two groups even in such superficial matters as spelling, e.g. AB divin, DE devin 479, 922, 1649 ; AB enfermetét, 1) enfertes, E enferté 7411; AB averez, DE avrez 587; AB muveres, DE muvres 770; AB veient (<véntt), DE vient 934; AB lusdt, D lunsdt, E lundt 1476; AB grisolites, DE orts- 1687 ; and the insertion or omission of nominative s, e.g. AB Deus, DE Deu 737; AB hu, DE Uile)us 773; AB uns draguas, DE un dragun 1020; AB chaiti( f)s, DE ch)aittf 1361; AB ostét, DE ostes 1379; AB uns glavies : savies, DE un glaive : saive 1713-14; cf. DE que for masc. nom. gus 1400. Many of these coincidences may of course be accidental, and they are mentioned merely as corroborative evidence.

§ 4. OTHER PossiBLE GROUPINGS

It would be difficult to defend any other grouping of the four MSS.

AD. The errors common to A and D are few in number and of no critical value. The coincidences in 114 (omission of m//), 242 (n’ for nen), 366 (nen for n’), 518 (omission of £), 551 (addition of ¢), 957 (4 omitted before om), 962 (tante for tantes), 1198 (dur for 7 in a feminine line), 1563 (repetition of en, BE missing), and 1682 (omission of em) are insignificant ; moreover the two MSS. disagree in other details of each of these lines, and often in the number of syllables. The dropping or adding of fem. ¢ in /arcin 1278, dimaine 1311, munte 1355, and the substitution of gue for gui 1273, 1336, | 1670, are slips to be expected of Anglo-Norman copyists. Other , errors involving metrical irregularities, common to A and D, are £ #/ for Ci? 585 (probably due to misreading of a capital letter), the omission of ¢ 643, the addition of /es pron. 695, the omission of ore _ after Uncore 1059 (haplography in the case of A), the omission of | mult 1173, the omission of en in Ens en enfern 1308 (haplography), Puis del jurn for Pose del 7. 1377 (due perhaps to the uncommonness of this use of Jose), and the omission of # 1828. In rog2 the same syllable is lacking in A and D, but the archetype was probably ' corrupt, cf. p. xl. In 1518 BE read correctly Ains que trovet nule rien att, but AD add » after ¢vovet, which the scribes presumably interpreted as indic. pres. 3 sg. ; in no instance does the poet elide un- | stressed -e¢. The readings morz for mors ‘bites’ 1025, a/ for e/ 1496, | cunduit for cunduis! 1557 (BE missing), 4 for s# 1825 are unimportant | coincidences ; while (¢)urent for event 562 is due to a fairly natura! |

1 But cf. v. 422, where AD have enfertet, BE enfermete.

Classification of the Manuscripts Ixvii

misinterpretation. In 1796, where AD have mu/t, BE fant, either reading is acceptable. Although some of these common errors are noteworthy, all can have arisen independently. The fact that A and D have sometimes preserved the authentic reading, corrupted in B and E independently (e.g. in 40, 274, 357, 418, 1328, 1592), creates a false impression that they are related.

AE. There are few errors common to A and E. The alteration of word-order in 26 (destroying the medial break), the substitution of nom. gue for gut (?) 38, eslist for esltt 107, grant for gransz 611, chatines for chaeines 869, are all trivial points. In 792, 1003 (cf. 1004), and 1206 the expansion of a feminine line in E coincides accidentally with a mistake of A. The following common errors are more stnking, but insufficient to prove relationship :

177 BD O:ndre (supported by C and L), AE /uindre. See the note to this line. 281-2 BD s'est mis: s'est assis (supported by C); AE se(s) mist: Sasist, 737 AE omit mult. 1323 Bore (+1), D or de, AE or (—1, probably for ore). 1359 B Far le lundi, D P. les lundis, AE P. lund? (~1). In spite of the many alterations introduced in E, it happens not infrequently that A and E, being the most carefully executed MSS., have preserved the original reading, corrupted in B and D (e.g. in 486, 872, 1447, 1532, 1700); allowance must of course be made for the regular addition of a syllable to the feminine lines in E.

BD. Between B and D there are many minor points of contact, due mainly to similarity of habits on the part of the copyists. Many of the details in which they agree against A and E are purely super- ficial, e.g. their preference for the forms prametre 121, 412, surjurner 327, 771, Segret 1094, aclose 1111, flambes 1211, detries 1236, etc. ; their omission or addition of flexional s in Deu 242, menez 250, un mes 405, chanuz: tendus 827-8, jugement 1254, etc.; their omission of mult 723, en 1319, fas 1793; their introduction of the definite article 528, 890, 1122, # 847, & pron. 1323, 1599; their substitution of nent plus for ne plus 183, a haltes vois for a halte v. 189, nen for w 226, guele for quel 343, es dur for es vus 356, aporte for portet ‘brings’ 409, tcel for cel 536, 1521, nus for uns 568, apres for puis 589, plus que for plus de 706, par yur for le urn 749, ad for ¢ 752 (cf. at 751), t¢0 for ¢o 762, al for en, ef 880, 1459, od for a 1164, ad

Ixviii Introduction

for est 1334 (influenced by 1333), /’ades for Brandans 1479, gu'il for gue t{ 1837; and their changes in word-order 74, 1229, 1544. They have employed similar methods in expanding the feminine lines 99, 615, 1223, 1315, 1421, and r59r. In the following instances B and D have an incorrect or altered reading in common, when either A or E is missing or divergent: 161 A (C) fant quant, E ¢. que, B ¢. cume(+1), D ttant cum (+1); 477 BD add & (+1), A missing; 526 A mal Peslavat (for s’eslevat), E si avala, BD mar se leva; 866 BD add as, E alters ; 983 A puint, E de rien (+1), BD nient; 1016 A gue nel en port, E Que...ne Tenport, B gu il ne p.(—1), D Ril nen p.(—1); 1044 A murer, E perir, BD peril; 1163 A £, E alters, BD A; 1272 A Jesu, E J. Crist (+1), B Celui, D Lcelut (+1); 1435 A Hui, E Ci, B Ore (+1),D Or.

None of the common readings just mentioned can safely be used for the purpose of classification ; in fact the lines quoted frequently contain divergences which make it clear that the agreement is acci- dental (e.g. 409, 528, 589, 1425, 1793). The following coincidences are more striking :

53 AE crett (<credit); BD cret ( <credo). 61 AE voldreit; BD voldrat (cf. 72, 80, etc.). 187 A Entrent(—1), E Ens entrent; BD Entrerent. The original reading may have been either Znz entrent or Entrerent. 606 AE Cum il voldrent (E voloient, +1) plus recetvre; BD Zant cum tl (D ail en, +1) voldrunt r. 782 Ade leu, Ede lui; BD del liu. The scribes misread /ui as liu ; cf. 716, where B has de /us correctly, DE del liu. 812 A dormant geisent, E a. gisotent; BD dorment e¢ gisent. 987 AE /a; BD &. The pronoun should refer to part, not to peisun; cf. 988 B /a, DE &. 1385 AE &t fus st prent; Bui feus esprent, Dif. sesprent. Cf. 761, 1157. 1394 AE tut un yurne une nuit; BD tute 7. ¢ tute n. 1707 AE surplantes ; BD purplantes. r80or A O or venis, E Or v.(—1); BD Or ev. 1834 A gwen lui virent, E alters; B gue tl v., D guilv.(—1). The words ef /ui are necessary to the sense. The changes in 606, 1385, 1394,and 1707 consist in the substitution

of a common for a less common word or phrase ; those in 812 and 1834 are simplifications of construction ; those in 53, 782, 987, and

Classification of the Manuscripts Ixix

1801 are the result of fairly natural misinterpretations ; all can have suggested themselves to different scribes independently. We cannot therefore regard this series of common readings as sufficient to over- rule the classification set up in the preceding sections. B and D are so negligently written and contain so many errors, that it would be surprising if their errors did not sometimes coincide. Rarely (e.g. in v. 1766) do they agree in furnishing the correct reading when A and E fail us.!

BE. Readings in which B and E agree against A and D are fairly numerous, but of little significance. Variants such as de a for @’sct 33°, 770, 1335, Quel...que for Quelgue...que 360(B —r), #’ for nen 541 (B —1), ous for nus 993 (cf. 992 AB vus, DE nus), Messe for Afesses 1092, gue for dative gui 1484, a for en 1548, fant for mult 1796, pur for de 1806, are trivial ; moreover in many of these instances either reading is acceptable, and BE may have preserved the wording of the original text. Equally negligible are the omission of mu/¢ 40, en 1051, sé 1230, and the introduction of ¢ 552, 1791, and mus 641, all producing metrical irregularities in B alone; the frequent inser- tion of the definite article*; and the substitution of one tense for another, e.g. estuet: puet (supported by C) for estout: pout 179-80, met for mist 181 (so C), dit for dist 1475, 1829, vint for vient 1722. The alterations of the word-order in 201, 203, 481, 586, 1296, and 1319 are the result of a common tendency to modernize and normal- ize. In 766 B and E have both omitted a ous in order to make Soussum, seuson trisyllabic. The same obvious syllable has been added to feminine lines in 237, 439, 494, 1031, 1040, 1218, 1264, 1488, and 1596 ; in 177 both B and E have added fxs? by anticipa- tion from the following line. Not infrequently B and E have made identical alterations, but in different circumstances, thus revealing their independence, e.g. in 357 (a detvre for ie b.), 520 (semble for sembles), 940 (omission of gut), 1007-8 (fut for fust), and 1017 (introduction of de). The following coincidences are more noteworthy :

102 AD (C) gu’a Brandan dist; BE que Brandan(s) d.

236 A(C) par, D alters ; BE sens (cf. 235).

478 A missing, D plus vus prendres (: crendres); B plus vus peneres (+1), E plus penerds.

3 In vv. 316 and 503 the reading common to BD may conceivably be that of the original. 2 Cf. ch. viii, § 5, ii.

Ixx Introduction

str AD & vols (: cols); B la voiz, E sa vols. 767 AD solt (: volt); BE sot. 902 A curt, D alters; BE crut. See the note to this line. 927 AD Deus; BE Deu. The omission of flexional s is here probably due to a misinterpretation. 1009 AD ne crement; BE nel c. (1 = le peril). 1052 BE add gue. The line probably lacked a syllable in the arche- type; cf. p. xl. 1061 AD Chantat; BE Chantes (influenced by Clames 1060). 1324 AD quant ‘since’; BE cum. 1369 AD ferlies (< ferro-ligatus) ; BE ferm lies. 1374 AD Que; BE Cume. 1659 A ne é targez, D n't ¢.(—1); B ne target (—1), E ne sargies. This line was probably corrupt in the archetype.

The majority of these common mistakes are the result of fairly natural misinterpretations, or of misreadings easily explained palaeo- graphically, and do not necessarily indicate any relationship between Band E. Others are probably due to corruptions in the archetype. The coincidence in 478 is striking, but both B and E may have found penres for prendres in the MSS. which they were copying. In a few instances (e.g. erent 562, pelfizs 1227, pose 1377) Band E have pre- served an indisputably correct reading against A and D.

ABD, ABE, ADE. There is little possibility of grouping ABD against E, ABE against D, or ADE against B, although a few com- mon errors might be alleged in support of each of these combina- tions. Thus ABD have minor errors in common, while E gives a satisfactory reading, in 1130 (ABD /uneirs, -1, E fonoires) and 1652 (ABD poéstis, AD -1, B adds pleonastic en, E poésteis); ABE all omit mult, correctly given by D, in 15761; and ADE present com- mon errors, while B gives the correct reading, in 557 (AD Quant vint le jurn al declinant, E li sors, B deljur), 802 (ADE omit a), 1350 (ADE add es, expanding a feminine line ; cf. 1349), and 1685 (ADE grant, B grans).? But these errors, nearly all trivial, cancel one another out; the very fact that a few instances, and a few only, can

1 It is possible, however, that a syllable was missing in the archetype, and that D alone has made a satisfactory emendatiun. Note also 652, where D alone has preserved the authentic word-order, but B is missing. ;

* A peculiar case is 798, where ABDE all present different readings; see the note.

Classification of the Manuscripts Ixxi

be found to support each grouping shows that the agreement is in each case accidental. ‘

BDE. More can be said in favour of grouping BDE against A. In a number of instances BDE present a common reading which is unsatisfactory, while the reading of A has every appearance of authenticity :

134 A Sur les treis jurs la semaine ; BDE replace Sur by Su/. Cf. the note to this verse. 220 BDE omit ‘vs (but otherwise diverge). 433 BDE omit 4, in order to make ¢ruverat trisyllabic. 568 A quies (: gracies) ; BDE guidez (destroying the leonine rime). 712 A al resirent; BDE se refirent (cf. Quant sunt refait 707). 1043 A chante; BDE chantez. The singular is required, cf. /ras

1044. 1205 A veit; BDE vit (cf. 1202). The present tense is used before and after. 1346 A da mer de sal; BDE de mer le sal. See the note to this verse.

1739 A Gardins est la praierie ; BDE replace es? by e.

In 259, 495, 674, 719, 893, and 949, BDE have employed the same devices for expanding feminine lines. Elsewhere, although either reading might stand, that of BDE is more hackneyed or otherwise inferior :

55 Avoldret plupf. ind. ; BDE voldreit. A historic tense is required (cf. 47), and it is improbable that the author would have used the two conditionals vo/dreit and devrett in close suc- cession in so heavy a manner.

357 A portet ‘brings’; BDE aporte. Cf. 1575. 471 A Sur les greinurs ‘greater than the greatest’; B tut & g., D tut des g., E tos it g. 769 A Des prep.; B Del, D De, E De da. 781 Amist; BDEdist. Metre un jurn is more idiomatic than dive un surn. 1813 A Puts; BDE Puis gu’. The reading preserved by A breaks the unity of the couplet, and therefore invited alteration by scribes. 182r A Ja; BDE Za.

But the appearance of unity presented in these instances by BDE

is deceptive. The scribes were dealing with a difficult text—difficult

Ixxii Introduction

not only on account of its antiquity, its exceptional subject, and its unusual forms, but also because the author had chosen to write in an idiomatic and sometimes contorted style. While the scribe of A copied the poem with very fair fidelity, the scribes of B, D, and E (or their predecessors) did not scruple to replace exceptional words, forms, and turns of expression by those current in their own day. Their frequent agreement merely proves that different scribes mis- interpreted or modernized the same words or passages in the same manner.

A. The archaic character of A, compared with the other MSS., has already been illustrated by several of the examples in the preced- ing paragraph. Further illustration is easy to supply. Very fre- quently A has retained archaic or uncommon words and forms, which the other MSS. have replaced or avoided, e.g.: r A Donna, BD Dame; 10 A Secund prep., BD Sulunc (so 807); 46 A Quer, BDE Kar (cf. 97, 231, 253); 55 A nepurtant, BDE nepurguant (so §77; in 629 AD have nepurtant) ; 194 A(C) donz, BE dans (cf. 454); 400 A at, BDE jur (so 751); 443 A entrin, BDE enter(r)in (cf. 686, 814) ; 529 A pur oc gue, BDE pur ¢o que; 620 A nael, BD noel, E noel (cf.879); 801 A dens adv., BDE ens; 811 A sumnes ‘sleep’, BE sumei/s, D alters ; 845 A ds paschur, B jur pascal, DE jur de pasches ; 1223 A tla, B da, D tl da, E tluec; 1688 A tsselites, BDE estes. The sparing use of the articles, a characteristic of the author’s style,’ is a strik- ing feature in A, but the other scribes have frequently added the article which appeared to them to be missing, e.g. in 25, 522, 552, 577, 710, 751, 780, 819, 893, 943, 953, 1070, 1081, 1139. Similarly BDE have added a personal pronoun subject in 1447, and BE have added 7? (D defective) as subject of an impersonal verb in 1486. The free use of the tonic forms of personal and possessive pronouns? is another archaic feature most conspicuous in A, e.g. 49 A lus dative (before the verb), BDE & (so 60, 1451, 1838; cf. 1447, 1541); 144 A seon allelment, BDE sun a.; 737 A lui acc., BE &, D nus; 997 Aas vos suspeis, BDE a v0(z) s. (cf. 999, 1092); 1435 A met repos ‘1 rest’, BDE me +. (cf. 520). Only A has preserved the neuter pronoun ¢/ 128 (B ¢o, CD #/), and the phrase / g/us 760 in which 4 (</Uo) indicates difference of measure; cf. also the retention by A of the neuter form ce/ 60, 1457. In 491-2 A treats arbre as feminine, while the other MSS. give it the more usual mas-

1 Cf. ch. viii, § 5, ii. 2 Ibid., xi and xiii.

Classification of the Manuscripis Ixxiii

culine gender. The peculiar verb-forms found in A (cf. ch. viii, § 3) have constantly been altered in the other MSS.; in addition to voldret 55 and guies 568 (cf. quien? 1220) we may mention dour 158 (BDE aut), estout (pret. of estuveir) 179, 530 (replaced by estuet or estut in the other MSS.), detent 232 (BCD dervent, E alters), fras 426 (BDE feras, cf. 553, 878, 883), cundiient 639 (BDE cundusent), strett (for stéveiz) 863 (B estrez, DE esteres); nearly all these forms are guaranteed by metre or rime as having belonged to the original. In 788 BDE by adding ~’ have obliterated the rare use of nud without a negative particle, and in 1426 they have added the negative particle after Ne guart quant (see the notes to these lines) ; again in 1336 they have added a second »’ (< nec) to the phrase esteit ne tvern. In 358 BDE have altered the rare construction of over with an accusa- tive of the person, and in 771 and 866 they have modified the prepo- siional use of desgue. In 839 A alone has preserved the name Ja\s\coines, which the author derived from the Vavigatio. On the liberties which the author took with the word-order, usually followed in A, but often modified in the other MSS., see ch. vili, § 5, xiv. The lines in which A alone offers a satisfactory text (e.g. 617-18, 688, 1717-18, 1757 ff.) are numerous. All readings of A are there- fore entitled to our careful consideration, even when they conflict with those offered by B and DE, or by B and either D or E. But A is naturally not infallible. Its peculiar forms, when not guaranteed by nme or metre, are often to be attributed solely to the late Anglo- Norman scribe. In 169-70 A’s rimes estreits : dre{i|s are opposed by C as well as BDE (all estreit : drest), In 420 A alone has modified the word-order. In 456 the form /uigneit given by A, though making excellent sense, 1s difficult to uphold against BDE /usert (: muvest).* In 519 BDE a/ has been preferred to A 4, in 970 BDE fenes to A lasses, in 1414 BE m’esforcent to A me forcent (D me folent), in 1669 B plein (DE plain) to A dreit,on the ground that they are more graphic or less usual. In the following instances a choice is more dificult: 335 A aparceut, BDE s’apercut; 370 A dist, BDE dit; 1269 A Cil is respunt, BDE Cil (4) respundit; 1272 A que, BDE qui acc.; 1400 A fant fort, BDE si f.

1 Cf. ch. viii, § 3, xxix.

Ixxiv Introduction

§ 5. THe Common ANceEsTors oF AB (a), DE (8), AND af (x)

The common ancestors of our two pairs of MSS. can be recon- structed with a fair degree of precision. We will designate them a and £ respectively.

a. The text of a can usually be determined without difficulty from a comparison of A and B, checked by the other MSS. When A or B is missing, certainty cannot always be attained ; but as a rule the reading of the surviving MS. agrees so closely with that of the 8 group that both can be regarded as reproducing the archetype. The text thus reconstituted is an extremely good one, without a line missing, and preserving in the main the metrical and linguistic pecu- liarities of the original. The errors common to A and B (cf. § 1) may be ascribed to a, although in a few instances it is possible that A and B introduced the same error independently. A certain number of lines in a were metrically irregular. A syllable was probably missing in 108 (g'sd for gue 11), 303 (cum for cume), 353 (Cist for Tcist), 411 (mule for netile), 572 (Halt), 590 (4 for ict), 718 (8), 724 (cest for icest), 842 (plus), 933 (cest for ceste?), 969 (), 1052 (gue after murir?), 1098 (l'abes for i abes), 1130 (tuneirs for tuneires), 1239 Turment for Turmente), 1397 (Dunc for Dunches), 1436 (Eneveis for Eneveies), 1521 (ist for Lissit), 1570 (ert), 1576 (mult), 1652 (poésts for potste’s) and 1659 (vus). A supernumerary syllable had been introduced in the following lines: masc., 246 (¢o), 384 (se), 741 (enfermettt for enfertét), 752 (le), 1383 (cume for cum), 1696 (od), and perhaps 1563 (en); fem., 292 (e), 327 (#), 588 (de), 800 (Cume for Cum), 866 (de), 1553 (gue), 1573 (en) and 1598 (e). The word-order had probably been altered in 1374, possibly also in 652 (missing in B).

The following lines contained errors: 43 (/eret?), 97 (del isle), 254 (entrer), 465 (Z or £ de for De), 500 (/ clair), 556 (En sun arbre), §88 (As for Desgue), 609 (s’en), 870 (set), 924 (en omitted), 1020 (far mer), 102X (¢ omitted), 1024 (des or de for dels), 1172 (cent), 1173 (e omitted), 1197 (¢s¢ omitted ?), 1390 (ne for se), 1402 (dunt omitted ?), 1476 (/usdt), 1735 (Quant) ; possibly also 478 (not in A), 812 (gisent for giseient) 860 (/a for Pan), 918, 934 (gui or guil veient for qui vient), 1557 (cundust), and other lines missing in B in which the reading of A is unsatisfactory." Whether A and B respectively were

1 Vv. 1549, 1564, 1610, 1620, 1716, 1760, 1764.

cent ee

Classification of the Manuscripts Ixxv

copied direct from a can hardly be determined. They exhibit no divergence which, in view of the habits of the scribes, compels us to postulate the existence of intermediate MSS. ; but at the same time the existence of one or more intermediate MSS. is possible in either case. In v. 798, where the reading of B probably represents the original, A has an entirely different line (modelled on 800), although there is no apparent reason for the change ; but it would be rash to assume, on the strength of this single instance, that A was copied from a different MS., in which 798 was missing.

B. The text of 8 can likewise be determined from a comparison of D and E, checked by the other MSS., but with less completeness and precision. The greater divergence between the members of this group, their systematic but mostly independent expansion of the feminine lines, the thorough modernization of E and its transposition into another dialect, often make certainty impossible. In the passages where E is missing, the reading of D is often unsatisfactory or mani- festly rewritten,’ and there is then no means of reconstructing 8 ; similarly in a few passages where D is missing.?, When reconstituted as far as possible, the text of 8 appears tolerably good, but distinctly inferior to that of a. It seems certain that vv. 630, 798, 1304, and 1473,° each represented in D and E by entirely different lines, neither agreeing with the reading of a, were missing in 8; also v. 1582, which in E is missing, and in D has been replaced by a stop-gap line. As shown in § 2, the process of lengthening the feminine lines had not been carried very far in 8; it is true that D and E employed the same device in nearly a hundred instances, but a large propor- tion of these changes may well have been made independently. Of the other alterations and errors common to D and E, mentioned in §§ 2, 3, and 4, the majority may reasonably be attributed to B. A number of masculine lines were metrically irregular in 8. A syllable was probably missing in 185 (eng for enens), 204 (sus), 265 (sen), 271 (enz), 283 (sul), 318 (ve-), 368 (cel for icel), 427 (en), 445 (tut), 539 (si), 636 (/ur), 737 (mult), 802 (a), 1052 (gue after murir ?), 1058 (dons), 1172 (e), 1173 (mult), 1174 (par fair, -2), 1397 (4), 1436 (Eneveis for Eneveies), 1545 (mult for grant ¢), 1599 (? for fer), 1659 (vus), 1798 (ad ?), 1799 (A), and 1801 (O). There were probably

1 e.g. vv. 14, 17-18, 88-90, 92, 192, 138-9, 1143-4, I§11, 1556, 1560, 1565-6.

2 ¢.g. vv. 1199-80, 1396, 1718. ? The line compensating for 1473 is placed in both D and E after 1474.

Ixxvi Introduction

supernumerary syllables in 365 (vus), 863 (estrez or esterez for streiz), 878 (feres for freiz), and 881 (do.). The word-order had been altered in 378, 402, and 1780. The following lines, in addition to those mentioned in previous sections as containing alterations common to D and E, were probably corrupt in 8: 476 (Zn dui plus miele crerrez ?), 691 (e od for e or o@), ggo (asses for a cez?), 1038 (De for A?), 1454 (Que for Quel?), 1735 (Quant). In 178, 220, 461, 548, 556, 607, 617-18, 701, 803, 839, 886, 1266, 1321-2, 1520, 1654, and 1757-8 the reading of 8B seems to have been partially unintelligible to the copyists ; while 688, 848, and 1807-8 had been entirely rewritten.

8, ¢«. It is not likely that either Dor E is a direct copy of 8B. The constant rewriting of whole lines, a characteristic of D, may usually be attributed either to the scribe of D or to a predecessor ; but in some instances the reading of D can hardly be explained unless we assume that the scribe was copying a MS. other than B: 235 (sans vent missing, D added par mer to complete the line), 454 (evidently a stop-gap line), 590 (do.), 809 (missing, but a compensa- tory line added after 810), 1162 (a feeble stop-gap line), 1565 (unin- telligible, but desevra¢ may represent desevrerat, inspired by frat serure- ment 1566), 1581 (similar to 809). We may conclude that between Band D there was at least one intermediate MS. (5), in which 454, 590, 809, 1162, 1581, and perhaps others of the lines missing or rewritten in D were wanting.’ The Picard reviser, whose work is preserved in E, seems to have had before him a text close to 8, but introduced a large number of changes. The Anglo-Norman origin of his source can be inferred from the violent measures often adopted in order to get rid of non-continental forms (e. g. in 599-600, 643, 1180, 1657-8, 1706, 1708), possibly also from certain misreadings (e. g. s2/ oretor for nul lorethur? 745). He appears to have misread weit as vent in 1181 (this entailed the rewriting of 1182), er (< heredem) as air in 1654, and uirent (<viderunt) as orrentin1834. A good many lines have been entirely rewritten, sometimes for no apparent reason.’ E can hardly be the actual work of the reviser, but may well be a copy of the revision (e). Vv. 168, 1256, 1582, and 1644 have been

1 The divergence of D in 188, 275-6, 407-10, 902, 1040, 1414, and 1632 may be due to either the last scribe or a predecessor. Cf. p. xvi.

2 e.g. 108, 166, 256, 659-60, 749, 1137, 1222, 1421, 1529. The alterations in 204, 276, 304, 1084, 1390, and 1592 are the result of attempts to improve the rimes; cf. p. xxi.

Classification of the Manuscripts —_\xxvii

dropped out entirely, leaving the couplets incomplete.’ Sixteen lines are a syllable short, according to the metrical system of the reviser *; in some of these instances the reviser probably neglected to add a syllable to the feminine lines, but in others (420, 461, 666, 855, 1147, 1226, 1323) it seems clear that the syllable was inadvertently dropped by acopyist. The mistakes made in 577 (as for a or a/), 623 (Dont for Vont), ro1r (vint avolant, mming a” with en contrary to both Anglo- Norman and Picard practice), 1127 (/a/s for vats and de for des), 1142 (/ for guz’), and 1309 (Ced for C'est) are those of a copyist rather than of the reviser; while 318, 461, 974,and 1762 are unintelligible in E as they stand.° It is unlikely that the reviser, who was endea- vouring to make the text intelligible and metrically regular, would have tolerated all these defects in his own work.

x. A comparison of a and £ enables us to reconstruct their com- mon original (the archetype of the four MSS. which contain the whole text ; here designated x) with surprising accuracy, except for a few minor details. The defects of a, not very numerous, are in nearly all cases corrected or explained by 8. Where 8 fails us, owing to the absence of one MS. and the divergence of the other, or occasionally the divergence of both MSS., the text offered by a is almost uniformly satisfactory, and may be accepted as representing the archetype. Hesitation is possible only in a few lines where a and £ offer different readings, both equally acceptable (cf. § 3), or where B agrees with B against A, without obviously modernizing.‘ In such instances the choice of reading has necessarily been somewhat arbitrary ; else- where it has been almost mechanical. The text thus reconstructed, complete (so far as we can tell) in every line, is a most satisfactory one. It cannot however be the original itself; for though it pre- serves faithfully the metrical and other peculiarities of the original, it contains a small number of defects for which the author (in view of the minute care which he clearly bestowed upon his work) cannot be held responsible. A syllable is missing in 303 (cum for cume), 411 (nule for neiile), 590 (c# for ict), 842 (plus), 1052 (gue), 1436 (Eneveis

1 Four couplets also are missing, but these may have been omitted by the reviser.

2 The defective feminine lines are mentioned in Appendix I; the masculine lines are 461, 737, 1226, 1323, 1801.

3 Note also 466, the first words of which have been inadvertently written at

the end of the preceding line. Vv. 934, 1012, and 1630 contain corrections which show the scribe at work,

* e. g. in 335, 902, 970, 1269.

Ixxviii Introduction

for Eneveies), 1659 (vus), and perhaps 1652 (poéstis for poédstez's). There is a supernumerary syllable in 292 (e), 588 (de), 866 (de), 1598 (ec), and perhaps 1696 (o¢). In 364 and1151 & has been substituted for /a, and perhaps in 1390 ne for se. On the reading Quant for Avant in 1735, see the note to this line. It is of course possible that in some of these instances the error was introduced indepen- dently ina and 8. The existence of intermediate MSS. between x and a, x and £ respectively cannot be demonstrated.

§ 6. THE PosiTIon or C

To assign to the fragment C a position in the scheme of MSS. is not easy, comparatively few of the passages on which the classifica- tion is based occurring within the first 310 lines of the poem. The conclusions that can be drawn from a comparison of C with the other MSS. are mainly negative.

It is certain that C does not belong tothe Bgroup. Inno instance does it share a mistake common to D and E: 26 C ne, DE n’en; 147 C de, D tut, E trestot; 151-3 B Cummande ... baise, DE Comandat ... baisat; 207 C hevet, DE vat; 249 C Kt, DE Kar; 295 C fortes, DE pernes. When a and £ present different readings, either of which is acceptable, C always agrees with the former: 19 C Deu, DE huem; 21 C gue, DE gu’tl; 116 C des suens, DE ses; 143 Ce certement, DE certeinement; 167 C E sur (for sus), DE Desus ; 193 C a, DE od; 198 C ceilet, DE celat; 216 C de, DE del; 220 C Desque, DE Zant que; 228 C Cum, DE Quant (so 241, 309); 234 C devrunt, DE deivent; 237 C dur, DE la; 265 C eissent sen, D ¢. fors, E st tssent ; 273 C a, DE de; 283 C suf, not in DE; 296 C ¢o, DE jol; 297 C pur, DE par. The feminine lines which D and E ex- panded in the same manner are unexpanded in C, viz. 145, 156, 207, 214, 221%, 225, 233, 240, 251, 254; in v. 134 C and D have added de, but E is missing. The coincidences between C and either D or E are accidental: 5 CD Far; 22C De miels entent, E Entends miex ; 92 CD omit suf; 108 CD gue il uit (-1); 116 CD sett for fer; 131 CD a/ for el; 163 CE a la roche; 182 CD avett for avetent ; 192 CD tei (+1) for 2; 204 CE dregat; 241 CE avint; 298 CD mentez (influenced by fries 297). In 42, C supports the reading justes against D (?) and E visées.

At the same time, C cannot be assigned to the a group. It is true that in v. 303 ABC have cum (-1) for cume (DE cum if); but this

Classification of the Manuscripts Ixxix

corruption probably goes back to x, and in any case is insufficient to prove relationship. In the other passages where A and B have a common error, C presents the correct reading: 43 A /erett, B freit, C feseit; 246 AB add ¢o(+ 1), C does not; 254 ABentrer,C munter. If these instances are few, and if C agrees as a rule with AB against DE, the reason is that a and C preserved the original readings far more faithfully than £.'

Can C have been derived from x? The evidence available is slender and contradictory. On the one hand cum for cume 303 might be used to support such a view ; but on the other hand C has preserved the original feminine line 292, to which x probably added a syllable. No other passage in vv. 1-310 appears to have been corrupt in x. It cannot be demonstrated that the readings of C in vy. 1 (Mahal!) and 18 (puis) are better than those of x (Aaiiz and mult respectively). The rimes of C in wv. 131-2 (espirt : virt) have been regarded as those of the original; but see the note to these lines. The ‘common error’ cum for cume 303 is too trivial to allow us to postulate a common ancestor (other than the original) for x and C. We must tentatively regard C as derived independently from the original, probably (in view of the many individual errors which it contains) with several copies intervening.

C is of great utility, in spite of its shortness and imperfections. Its independence of both a and £ having been established, the uncer- tainty as to its ultimate position is of no practical consequence. It covers the longest of the passages missing in E (vv. 1-18, 73-92, 123-42); and when in these passages the readings of a and D differ,’ its support of a shows clearly that D is at fault. The effect is similar when D and E offer divergent readings, neither agreeing with a.” When A or B agrees with D or E against the other pair, C always supports the reading already preferred on other grounds: 26 BCD demander plus, AE p. d.; 102 ACD gue a Brandan, BE que Brandan(s); 107 BCD eshit, AE eshst; 181 BCE met, AD mist. The effect of this independent testimony of C is to increase our respect for the a group, whose reading we are encouraged to adopt throughout the text unless it can be definitely shown to be wrong. When the MSS. of the a group disagree, C more often supports A:

1 Wien (of. at., pp. 29 ff.) and Calmund (Prolegomena . . ., p. 26) assigned C to the a group.

2 Vv. 14, 17-18, 82, 84, 88, 89, 90, 138, 142.

3 Vv. 115, 166, 172, 204, 255-6, 275, 286, 287.

Ixxx Introduction

37 pernans, 161 tant quant, 185 entres enent, 220 fuz, 228 Cum vens n't ert, 236 par, 260 # not added, 282 puis ; sometimes it supports B : 23 /a scripture (preserving the medial break), 276 Dun entaillét sunt ii parct. C shares a faulty or altered reading with A in wv. 6 (addition of Z), 224 (cesset for cestet), 232 (aler for haler), and 254 (omission of ¢vs); with B in wv. 134 (Sud for Sur), 188 (da for ja), and perhaps 280(Zatrerent ens for Entrent enens). The vocabulary and forms of C are often as archaic as those of A, e. g. dons 8, 194, aloekes 173, eret 222, bettrer 233, alceur 279.

§ 7. CONCLUSION

The results arrived at in the preceding sections may be expressed schematically as follows :

O |

|

L represents the Latin prose translation (see ch. vi), R the Latin rimed translation (see ch. vii), A and p the lost MSS. of the Anglo- Norman poem on which they are respectively based ; « is the work of the Picard reviser of the poem. A critical text can be obtained only from the x family. The text of x has been reconstructed in the manner indicated in § 5; the few faulty readings that it contains have been eliminated as far as possible, on the ground that they cannot (for reasons already stated) be due to the author. In a small number of instances an alteration or choice of reading has been based on C, L, or R.

A B R

1 Cf. ch, viii, § 2, xxi.

—

Disitioed Bal Google

(eoS1-Sibi aa ‘y ‘sy) of 6 0} *X “@ NVISVdSHA NOLLOD ‘SAW “LING ‘SW

nm a ' TAQ , | Smqurp mgps aun) anBey

Te ate “

ON ‘o or’ How)

=,

Digitized by Google

———7~

Digitized by Google

Relation to the Navigatio S. Brendani \xxxi

V. Tue RELATION OF THE ANGLO-NORMAN POEM TO THE NAVIGATIO SANCTI BRENDANI

The author of the poem gives no hint as to his source, merely stating, in wy. 9-13 of the prologue, that he has put into writing in the Romance (i.e. French) tongue the story of the good abbot Brendan. That the source was not French is clear from the wording ofv. 11, In favour of a Celtic source there is no evidence whatever ; the personal knowledge of Brendan’s homeland which wv. 163-4 seem to imply is illusory. It cannot be doubted that the source was a Latin one, in all probability the Mavigatio Sancti Brendant2 This celebrated prose work was read, copied, and translated through. out Europe in the Middle Ages, far surpassing in popularity all other versions of the legend. The Anglo-Norman poem is by no means a close translation from the Latin, and its dependence on the Vatigatio is not so obvious as in the case of most other renderings ; but there is ample internal evidence to prove that the avigatio was its source.

A detailed discussion of the origin and formation of the Vavigazéio, or of its historical and literary interest, would be out of place here.‘ It will suffice to say that, according to the view now generally held, the Navigatio is a Christianized smram, or prose story of an adventurous sea-voyage. The émramisa literary genre which flourished in Ireland between the sixth and twelfth centuries; a few imrama are still extant, including one which has striking affinities with the Brendan

1 See the note to v. 164.

2-H. Suchier, in Boehmer’s Romanische Studien, i, p. 555, seems to have been the first to state this point explicitly. F. Novati (La ‘ Navigatio Sanctt Brendani’ in antico venestano, Bergamo, 1892, ch. iii of Introduction) suggested that the source of the Anglo-Norman poem differed from the vulgate version of the Navigatio; but the present chapter aims at showing that all the divergences are the result of alterations by the Anglo-Norman author. On the Latin prose version in which C, Plummer believed that he had found the source of the poem, see ch. vi.

* On other medieval versions see H. Zimmer, in Z¢schr. f. deutsches Alterthum, xxxiti, pp. 129 ff.; A. Schulze, in Zéschr. f. roman. Phil., xxx, pp. 257 ff.; C. Plummer, in Zétschr. f. celt. Phil., v, pp. 124 ff. Zimmer believed in the existence of another Latin version, now lost, presumed to be the source of some of the German versions. Similarly F. Novati (op. crt., ch. ii) postulated an altered form of the Navigatio as the source of the Italian versions.

‘ See G. Schirmer, Zur Brendanus-Legende, Habilitationsschrift, Leipzig, 1888 ; and the above-mentioned works of Zimmer and Schulze.

3351 l

Ixxxii Introduction

story, namely the Jmram Maelduin.' As to why the saintly Brendan of Clonfert (d. 576) became the hero of such a legend no convincing explanation has yet been put forward. The conversion of these romantic, often poetic, but entirely pagan stories into an edifying tale of saintly endeavour, suitable for monastic reading, is one of the most remarkable fours de force in the history of literature. The known facts regarding Brendan’s career and the formation of his legend have been conveniently summarized by C. Wahlund, ? whose work should also be consulted for its valuable bibliographical information. The date of composition of the Vavigatio is unknown, but cannot have been later than the tenth century, and may have been as early as the ninth ; the oldest extant MSS. date from the tenth century, and in them the text has already become corrupted. *

The vogue enjoyed by the Vavigato in the Middle Ages is attested by the existence to-day of not less than 80 MSS,, scattered all over Europe. * Unfortunately this abundance of material has its drawbacks. The self-sacrificing scholar who will collate and classify the whole of these MSS. has not yet been found. A partial attempt at classifica- tion was made by C. Steinweg, on the basis of 31 MSS. wholly or partly collated; but his argumentation and conclusions are open to serious objections, ® and the whole task needs to be taken in hand afresh. For the present we must depend on the editions, based on one or several MSS., published by A. Jubinal,* C. Schréder,”

1 Edited, with English translation, by Whitley Stokes in Revue celtique, ix, pp. 447 ff., and x, pp. 50 ff.; with French translation, by F. Lot, in d’Arbois de Jubainville’s Cours de littérature celtique, v, 1892, pp. 455 f.; see also G. Schirmer, op. cit., pp. 42 ff.; H. Zimmer, of. at., pp. 150 ff.; C. Boser, in Romania, xxii, pp. 578 ff.

3 Die altfranz. Prosaiibersetzung von Brendans Meerfahrt, Upsala, 1g00, Introduction.

3G. Schirmer, op. cit., p. 68, ascribed the Navigatio to the ninth century ; C. Steinweg (see below), p. 48, to the end of the ninth or beginning of the tenth. H. Zimmer, of. cit., argued for the middle of the eleventh century as the date of composition; but this is much too late, cf. F. Novati, op. at., ch. i; C. Boser, in Romania, xxii, p. 589; C. Plummer, Vitae Sanctorum Hiberniae, i, Oxford, 1910, p. xli, note.

4 74 MSS. were listed by C. Steinweg, Die handschriftlichen gestaltungen des lateinischen Navigatio Brendani, in Roman. Forschungen, vii, Erlangen, 1891, pp. 2ff.; seven others were mentioned by C. Wahlund, of. aiZ., p. xxx.

6 Cf. C. Boser, in Romania, xxii, pp. 578 ff.

6 La légende latine de S. Brandaines, Paris, 1836, pp. 1-53 ; edition based on Paris MSS. Cf. Schroder, p. xviii; Steinweg, pp. 29 f.

7 Sanct Brandan, Erlangen, 1871, pp. 3-36; based on two MSS., at Leipzig and Wolfenbiittel respectively.

Relation to the Navigatio S. Brendani \xxxiii

P. F. Moran,' the Benedictines of Monte Cassino,? the Bollandists,° and C. Wahlund,* and on those early MSS. which happen to be available.’ For the purpose of this chapter, the texts of Schréder and Wahlund have been constantly utilized, while others have been consulted in special cases. Owing to the freedom with which the Anglo-Norman author has retold the story, the lack of a critical edition of the Vavigatio does not seriously hamper our investigation of its relation to the poem.

A comparison of the two texts 1s facilitated by the character of the story. The poem, like the Vavigatio, consists of a series of episodes, each usually self-contained and but loosely connected with the remainder. Although the order of the episodes is not wholly hap- hazard,* the unity of the tale depends mainly on the fact that the same party of voyagers undergoes each adventure. The Anglo-Norman text has therefore been divided up in this edition into sections corresponding as nearly as possible to those of the Vavigatio; complete correspondence is impossible, owing to the alterations introduced by the poet. The sections of the poem have purposely not been numbered, the many different numberings of the sections of the Navigatio being already sufficiently confusing.” Calmund ® has sum- marized in parallel columns the subject-matter of the Vavigatio and the poem respectively, indicating the discrepancies between the two. Our present plan is to take each episode of the poem in turn, examine its relation to the Vavigatio, and discuss the divergences.

(N = Navigatio Sancti Brendani; sections as in Wahlund’s edition. P = the Anglo-Norman poem.)

Vv. 1-18. The prologue or dedication in P naturally has no counterpart in N.

} Acta Sancti Brendani, Dublin, 1872, pp. 85-131. Cf. Steinweg, p. 23.

2 Bibliotheca Casinensis, iii, Monte Cassino, 1877, pp. 411 ff.

3 Acta Sanctorum Hiberniae ex Codice Salmanticensi. .. edita opera C. de Smedt et J.de Backer, Edinburgh and London, 1888, cols, 113 ff. Cf. Steinweg, Pp. 29-32.

* Op. at., pp. 2-100 (‘Kompromiss-Text’) and pp. 102-200 (text of MS. Bibl. Nat., lat. 15076).

® In Oxford, for instance, the Bodleian MS. Laud Misc. 410, fol. 40b-68 a. The tenth-century MS. at the British Museum (Add. 36736, fol. 168 b-aqo b) has also been consulted.

® Cf. pp. cii ff.

7 For a concordance, see the footnotes to Wahlund’s second text of the Nazvigato (pp. 103-200).

* Prolegomena ..., pp. 152-207.

Ixxxiv Introduction

Vv. 19-38. Saint Brendan: his ancestry, country, status and character; = N §1. The poet has merely amplified the bald statements of N, but omitting the names of persons and places (all somewhat outlandish, and of no interest to an Anglo-Norman audience).

Vv. 39-70. His desire to see the Other World. Brendan beseeches God that before dying he may see paradise and hell. This passage corresponds to nothing in N, and appears to have been invented by the poet in order to account more satisfactorily for Brendan’s surprising resolve. In N Brendan’s desire is due (so far as we can tell) to curiosity, awakened by a chance meeting with Barintus ; whereas in P it is the result of his pious reflections and enterprising spirit! On the whole the innovation, though not specially ingenious, seems an improvement.

Vv. 71-102. The narrative of Barintus; = N § 2, but greatly altered and abridged. In N Barintus visits Brendan—whether purposely, is not stated; in P Brendan goes to Barintus for information and advice ; in both Brendan clearly knows beforehand that Barintus has been on a marvellous voyage across the ocean. The poet has wisely ignored the incomprehensible melancholy and tears of Barintus on his meeting with Brendan. The only personal. information about Barintus vouchsafed by N is the name of his uncle ; P, on the other hand, describes him as a hermit living in the woods with three hundred monks. The Anglo-Norman author may well be held responsible for these alterations ; the hermit in the wood, to whom all classes resorted for absolution and advice, was a stock figure in medieval literature.* Barintus’ description of his voyage in search of his godson Mernoc,’ and of his stay on the insula deliciosa is briefly (and rather obscurely) summarized in P. His visit to the Zerra repromissionis sanctorum, described in detail in N, has been omitted altogether in P ; although the mention in vv. 95-6 of the scent of the flowers of paradise shows that the poet was familiar with the whole episode. This important omission 1 This point is insisted on in vv. qo-1: Crem hoem qui ert de mult grant sens,

De grans cunseils e de rustes.

2 Cf. Ogrin in Beroul’s 7nistan ; branch viii of the Roman de Renart; Le chevalier au barisel; &e., &c.

5 Contrary to the statement of Steinweg (p. 9), Mernoc appears to be the original form; cf. C. Wahlund, Brendans Meerfahrt, pp. \vii and 231. Some

MSS. of the Navigatio have corrupted Mernoc atque procurator into Mernocatus procurator.

Relation to the Navigatio S. Brendani \xxxv

is advantageous from a literary point of view, the description of paradise being reserved to form an effective climax at the end of the story. To introduce, as N does, a full description of the voyagers’ goal at the very outset results not merely in repetition but also in an inartistic anti-climax. The presence on Mernoc’s island of the scent which enabled Barintus to dispense with food and drink is explained in P as due to the nearness of the island to paradise; whereas in N Mernoc and Barintus are stated to have brought it back in their clothes on returning from paradise.

Vv. 103-156. Brendan chooses fourteen companions, secures their co-operation, and prepares for departure; = N§3. In N Brendan, in order to make his first proposal to the fourteen chosen monks, shuts himself with them in an oratory; in P this detail 1s omitted, and the first speech is briefly and indirectly reported (vv. 109-10). On the other hand P adds a number of interesting details, for the most part such as would naturally suggest them- selves to a clerical writer: the monks divide up into pairs to discuss Brendan’s proposal (v. 112), Brendan exacts from them a pledge of constancy (vv. 117-22), he addresses them a second time in the chapter-house (vv. 125-34), and an angel brings him instructions for the voyage (vv. 137-44). In both N and P Brendan decrees a fast of three days a week for forty days, at the end of which he commends his community to the prior. Brendan’s farewell, indicated in N, is more fully developed in P (vv. 145-56). At the end of this section, N mentions that Brendan and his fourteen companions paid a visit of three days and nights to the isle of a holy father named Aende; the poet has wisely omitted this episode, which contributes nothing to the story.

Vv. 157-184. Preparations for the voyage: the starting-point chosen, the boat constructed and provisioned ; = N §4. In this section P follows N fairly closely, but there are two interesting discrepancies. The point of embarkation, which is more fully described in P (vv. 167-72) than in N, is designated sedes Brendant in N, & Salt Brandan in P (v. 164). Again, whereas in N the boat is described as a coracle, with a frame of exceedingly light material (naviculam levissimam costatam et columnatam ex vimine} sicut mos est in illis partibus), in P timber is mentioned (mairen

1 So Wahlund’s ‘Kompromiss-Text’, p. 10, line a5 ; his second text (p. 110, line 25) and Schroder’s edition have ex silva.

Ixxxvi Introduction

174, fust sapin 175). These discrepancies are discussed in the notes,

Vv. 185-202. The three intruding monks; = N§5. There is one slight discrepancy : in N, when the monks appear, Brendan is standing alone on the shore and blessing the harbour before embarking ; in P (v. 187) he is already on board when they appeal to him for admission.

Vv. 203-264. The first voyage; = N §6. This section fur- nishes a particularly interesting example of re-treatment by the Anglo-Norman author, who has introduced into the narrative not only more logic and coherence, but also a different ideal from that of N. In N the monks do not row so long as the wind is favourable, but when (after fifteen days) the wind fails they row until their strength gives out; whereupon Brendan exhorts them not to lose heart but to leave their sail spread and entrust them- selves to God. They therefore let themselves drift until forty days have gone by and their food comes to anend. In P, on the other hand, the monks begin, with excessive enthusiasm, by rowing’ even though the wind is favourable; when the wind fails, they become dismayed, but Brendan encourages them, and wisely advises them to row only when necessary. They therefore continue to row energetically until forty days are completed, when their provisions come to an end, their strength fails, and fear enters their hearts. There is no doubt that the version of P is superior: the lack of food furnishes a reasonable explanation of the monks’ fear and loss of strength. All through the poem the willingness, or rather anxiety, of Brendan and his monks to toil and undergo hardship for a pious purpose is their most conspicuous virtue. Vv. 241-6, which have no basis in N, emphasize the moral of this episode (viz. ‘God helps those that help themselves’) and justify the changes introduced. Apart from this main point, we may note that P has omitted certain details found in N: the monks’ attempt to obtain fresh water from the streams which descend from the island, Brendan’s stern rebuke, and his prohibi- tion of the removal of any furniture from the ship. On the other hand P has added some picturesque details in vv. 203-8 (Brendan’s prayer and blessing for his monks), 213-14 (the monks see naught save sea and sky), and 257-8 (the perils of the rocky shore).

1 Note that mager in this text always signifies ‘to row’, though nwavigare in N means either ‘te row’ or ‘to sail’. ‘To sail’ is either ceerve or sigler.

Relation to the Navigatio S. Brendani \xxxvii

Vv. 265-306. The uninhabited city; = N §7, first part. InN (end of § 6) a dog meets the voyagers as they walk along the shore, and acts as guide ; P omits this picturesque trait. The descriptions of the town and palace show numerous discrepancies. N speaks of a stronghold (0ffidum) containing a single great hall (aulam magnam); P describes a city of many palaces (and yet uses the word casfel in v. 267). P has omitted to mention the water for washing feet, the vessels, bridles, and horns hanging on the walls in the interior of the hall, the table which the voyagers find ready laid for them, with its linen and the repast of fishes and white loaves. On the other hand, P has added particulars of the splendour of the castle as seen from without (vv. 268-70), its wall of crystal (v. 272), the palaces all built of marble (vv. 273-4), the walls adomed inside with gems set in gold (vv. 275-6), and the plate of gold and silver (vv. 291-2). Apparently the poet has sought to add to the interest and marvellousness of the story, perhaps also to please his court audience, by increasing the magnificence of the descriptions and introducing in each case the most precious materials known to him. It may also be noted that P mentions explicitly (v. 278) what is left to inference in N—that no human being is visible in the city.

Vv. 307-354. The stolen goblet; = N § 7, middle part, but with important differences. The Anglo-Norman poet has made an intelligent alteration in this episode. In N (at the beginning of § 7) on arriving in the hall Brendan explicitly declares that he sees Satan tempting one of the three intruding monks to steal; the subsequent theft, committed in the face of this direct warning, becomes highly improbable. In P (vv. 297-8) Brendan merely gives the monks a general warning against disloyalty to God, and there is no serious improbability. The object stolen in N is a silver bridle (/renum argenteum),' apparently one of those pre- viously described as hanging on the wall ; in P this is replaced by a golden goblet («# hanap d'or), perhaps because the author did not regard a silver bridle as sufficiently desirable to justify the commission of the crime. The strange detail of the negro boy (infantem ethyopem) who in N plays with the bridle in front of the thievish monk, and is later seen to fly from the monk’s bosom when he repents, was passed over by the poet, who speaks merely of

1 The translation of /renum in this passage as ‘ bridle ' has been disputed ; cf. Wahlund’s note, pp. 237 f.

IXxxviil Introduction

‘a devil.’ Other details omitted in P are the casting down of the stolen article by the detected thief, the devil’s claim to have occupied his abode for seven years, and Brendan’s reply forbidding him to harm any other human being until the Day of Judgement. The quaint passage (vv. 323-6) in which the poet emphasizes Brendan’s ability to see in the dark without a candle has no counterpart in N.

Vv. 355-376. God provides all necessaries; = N § 7, last part. The passages in N and P correspond, except that N speaks of ‘a youth’ but P of ‘a messenger’, the message is somewhat ex- panded in P, and P adds reflections (vv. 371-6) on the all-providing solicitude of God.

Vv. 377-484. The Isle of Sheep; = N §8. The poet has omitted to mention the abundance of streams and fish-pools on the island, the docility of the sheep which is captured for the voyagers’ Easter dinner, and the seizure of a