The Verb in the Book of Aneirin: Studies in Syntax, Morphology and Etymology [Reprint 2011 ed.] 9783110962321, 9783484429123

The "Book of Aneirin" is a thirteenth-century manuscript collection of Welsh praise-poetry. In comparison with

180 43 69MB

German Pages 490 [492] Year 1996

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Acknowledgements
Introduction
Section I
1. Statistical Analysis, Basic Order
2. Functional and Formal Syntactic Analysis
3. Syntax and Reconstruction
Section II
Editorial Practice
V
V[S-pro]
V[S+pro]
[S-pro]V
[S+pro]V
V[O-pro]
V[O+pro]
[O-pro]V
[O+pro]V
V[S-pro][O-pro]
V[S+pro][O-pro]
[S-pro]V[O-pro]
[S+pro]V[O-pro]
V[O-pro][S-pro]
S[O+pro]V
OV[S-pro]
Enclitics in Compounds
Rel
Misc
?
Instt. 598-603
Notes
1. Verbal forms
2. Verbs
3. Absolute vs. conjunct
4. Developments of the PIE s-aorist in Welsh
5. Future in -(h)awt / -(h)awr
6. Oedd and the imperfect
7. Further forms of bot
8. The verbal particles
9. CA 1. 285 and archaic case forms
10. Gwant
11. CA 11. 1055-56/1078-79
12. CA awdl XCIII
13. Gwyr a aeth
14. CA 1. 616
15. Kyui
Bibliographical Abbreviations
Bibliography
Concordance of Lines
Index of Words
Recommend Papers

The Verb in the Book of Aneirin: Studies in Syntax, Morphology and Etymology [Reprint 2011 ed.]
 9783110962321, 9783484429123

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

BUCHREIHE DER ZEITSCHRIFT FÜR CELTISCHE PHILOLOGIE HERAUSGEGEBEN VON KARL HORST SCHMIDT

Band 12

GRAHAM R. ISAAC

THE VERB IN THE BOOK OF ANEIRIN Studies in Syntax, Morphology and Etymology

Max Niemeyer Verlag Tübingen 1996

I Emrys

Die Deutsche Bibliothek - CIP-Einheitsaufnahme [Zeitschrift für celtische Philologie / Buchreihe] Buchreihe der Zeitschrift für celtische Philologie. - Tübingen : Niemeyer. Früher Schriftenreihe Buchreihe zu: Zeitschrift für celtische Philologie NE: Zeitschrift für keltische Philologie / Buchreihe; HST Bd. 12. Isaac, Graham R.: The verb in the Book of Aneirin. - 1996 Isaac, Graham R.: The verb in the Book of Aneirin : studies in syntax, morphology, and etymology / Graham R. Isaac. - Tübingen : Niemeyer, 1996 (Buchreihe der Zeitschrift für celtische Philologie ; Bd. 12) ISBN 3-484-42912-7

ISSN 0931 -4261

© Max Niemeyer Verlag GmbH & Co. KG, Tübingen 1996 Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Gedruckt auf alterungsbeständigem Papier. Druck: Weihert-Druck GmbH, Darmstadt Buchbinder: Geiger, Ammerbuch

Contents Acknowledgements Introduction

vii 1

Section I

21

1. Statistical Analysis, Basic Order

21

2. Functional and Formal Syntactic Analysis

39

3. Syntax and Reconstruction

107

Section II

148

Editorial Practice

148

V

151

V(S-pro|

170

V[S+pro]

190

fS-prolV

190

[S+pro]V

193

V[O-pro]

195

V[O+pro]

221

|O-prolV

222

[O+proJV

227

V[S-pro][O-pro]

232

V[S+pro][O-pro|

238

|S-pro]V[O-pro]

241

[S+pro]V[O-pro]

246

V[O-pro][S-pro]

251

SlO+projV

256

OV[S-prol

259

Enclitics in Compounds

262

Rel

264

Misc ?

276 283

Instt. 598-603

294

Notes

296

1. Verbal forms

296

2. Verbs

304

3. Absolute vs. conjunct

354

4. Developments of the PIE 5-aorist in Welsh

358

5. Future in -(h)awt I -(h)awr

368

VI

6. Oedd and the imperfect 7. Further forms of bot 8. The verbal particles 9. CA 1. 285 and archaic case forms 10. Gwant 11. CA 11. 1055-56/1078-79 12. CA awdl XCIII 13. Gwyraaeth 14. CA 1. 616 15. Kyui Bibliographical Abbreviations Bibliography Concordance of Lines Index of Words

371 380 383 393 399 413 419 423 428 436 444 445 467 471

Acknowledgements This work would not have been possible without help from many quarters. I am grateful first of all to the Alexander von Humboldt Foundation, who awarded me a fellowship for a generous two years, from July 1993 to June 1995, allowing me to conduct the research for the work uninterrupted at the Sprachwissenschaftliches Institut of the Friedrich-Wilhelms-Universität Bonn. I owe a particular debt to Emeritus Professor Karl Horst Schmidt of the Sprachwissenschaftliches Institut, for supporting my fellowship with the von Humboldt Foundation, and for sacrificing countless hours of his always priceless time to discuss aspects of the work and related issues with me. To the pioneering linguists of the 'Friday Circle1 at the Sprachwissenschaftliches Institut, Dr. Stefan Georg, Ms. Ina Lucht, Mrs. Melanie Suchy, Messrs. Peter Raulwing, Uwe Seefloth, Dieter Stern, and Professor Nico Weber, I am indebted for the warm welcome they gave me into their ranks, and for the fresh perspectives on old problems that discussions with them always bring. The debt of the work to their interest and ideas will be clear to them. I thank Dr. Ursula Marmo for kindly giving me a copy of her then unpublished dissertation, which provided useful further references. Various parts of the work were presented as occasional lectures in Bonn, Berlin and Marburg, and have benefitted from the discussions following these presentations. Over the years, Professor D. Ellis Evans has been a faithful correspondent, always ready with support and encouragement for various aspects of my work. I am grateful to Ms. Bettina Gade of Max Niemeyer Verlag for the sharp and critical eye with which she has guided the work through the press. Words do not suffice to express my gratitude to Maike, whose long-suffering patience and support allowed to me concentrate on the impracticalities of a project such as this. Needless to say, of all the infelicities which remain in the work, I alone am guilty. Finally, I cannot but mention my debt to Emeritus Professor Emrys Evans, who, in more ways than I can enumerate here, put me on the right path and made this work possible. I hope he will accept the dedication as a token of my gratitude.

Bonn, 27.10.95

G. R. Isaac

Introduction

o.o The original motivation for this work was the feeling that statements about the character of the syntax of early Welsh poetry in general, and that of the text of the Book of Aneirin in particular, were rather poorly supported by close and detailed textually based studies. In particular, the word order has been used in various ways in argumentation in the context of a wide range of often contradictory theories, without ever having been the subject of an extensive study in its own right It seemed to me that in order to make meaningful, and above all testable statements about the syntactic structure and function of the word-order patterns of the corpus, all instances, within relevant parameters must be collected and classified. In the case of constituent order at the level of the clause, this meant collecting all finite verbal sentences, and classifying them with respect to the positions of the arguments of the conjugated verb. This in turn meant that a collection of all finite verbal forms in the corpus would have to be produced. In putting this collection together, I have omitted the instances of the copula and the substantive verb, since it seemed to me that these involved a separate syntactic subsystem that should be studied in its own right Since the examination of the syntax associated with the other finite verbal forms was such an extensive undertaking in itself, it appeared that I would not be able to do justice to the syntax of the forms of hot within the confines of the same work. I therefore omitted these. A detailed preliminary study was presented by MacCana/Watkins (1958). Since I then had a comprehensive list of all finite verbal forms in the Book of Aneirin (with the omission just noted), I decided that the opportunity ought not to be missed to classify and analyse the forms morphologically and etymologically. Furthermore, since the classification of the verbal sentences involved considerable philological and analytical work to start with, it seemed not out of place to append several, more or less speculative notes on the analysis of certain passages which appeared particularly interesting to me. The result is the following book. As a whole, it is in essence a study of the language of the Book of Aneirin, from various theoretical and methodological perspectives. But as all the aspects of the language considered centre on the finite verb, it seemed wise to draw attention to this in the main title of the book, The Verb in the Book of Aneirin, the subtitle drawing attention to the three main aspects considered, Studies in Syntax, Morphology and Etymology. It is easy to imagine that some researchers may be particularly interested in certain aspects of the language to which I pay less attention in my discussion, or to which I give no consideration at all. But I hope that the study contains sufficient material to be of use to scholars working in a range of fields.

The division into two large sections follows from the nature of the analysis undertaken. Section I presents the close syntactic analysis of the sentences of the corpus, section II presenting the collection and classification of the instances which form the basis of this analysis, followed by the notes referred to above, on various morphological, etymological and interpretational matters arising from the text. Only in the syntactic discussion of section I have I attempted to produce a continuous, consistent line of argumentation. Section II, with its character, effectively, as a series of unconnected notes, may give a more fragmented impression. The overall result cannot be claimed to be definitive in any respect, regarding any of the matters considered. It naturally follows from the nature of the argumentation that at certain points in the discussion, certain results are assigned, for argument's sake, a quasi-definitive character. However, even when an author makes the claim that a certain result is definitive, in the end it is only posterity that really has the final say as to what actually is definitive. 0.1

Who do you think I have at my elbow, as happy as ever Alexander thought himself after a conquest? No less a man than leuan Fardd who hath discovered some old MSS. lately that nobody of this age or the last ever as much dreamed of. And this discovery is to him and me as great as that of America by Columbus. We have found an epic poem in the British called Gododin, equal at least to the Iliad, Aeneid or Paradise Lost. Tudfwlch and Marchlew are heroes fiercer than Achilles and Satan. (Letter from Lewis Morris to Edward Richard, 1758; Owen 1947-49: i, 349.) These oftquoted words may be thought to mark the beginning of modem interest in, and study of the text in question. Shortly after this announcement, Evan Evans (the leuan Fardd of the letter) published his Some Specimens of the Poetry of the Antient Welsh Bards (E. Evans 1764), which contained ten stanzas with Latin translations. The whole of the text was made publicly available in the first volume of the Myvyrian Archaiology of Wales (Jones et al. 1801). These works were not however based on the Book of Aneirin as it is known today, but on various late and rather imperfect copies of the mediaeval manuscript, at several removes. Evans found the text in a copy made by Thomas Wiliems of Trefriw towards the end of the sixteenth century. The copy is now part of the manuscript BL Add 31055. The relationship of the many post-mediaeval copies of the text to each other and to the Book of Aneirin is discussed by Huws (1988). The first edition of the text directly from the Book of Aneirin is that of Skene (1868), together with a translation in volume 2, by Silvan Evans. The edition was a great advance on that of The Myvyrian Archaiology but is still imperfect, and the translation, in common with all those made before the advances of twentieth-century scholarship is unusable. Other such translations are those of Probert (1820), Ab Ithel (1852), Stephens

(1888), Anwyl (1911): despite the imperfections, there is, as would be expected, a gradual increase in accuracy of these translations with time. Work on the text of the Book of Aneirin was placed on a firm foundation by the publication of J. Gwenogvryn Evans' facsimile and diplomatic edition (J. G. Evans 1908), though Evans' interpretation of the text, encapsulated in the introduction to the edition and in the later publications, J. G. Evans (1922, 1924), is disastrous, like his interpretation of the poems in the Book of Taliesin (introduction to J. G. Evans 1910, cf. the review monograph by Morris-Jones 1918). These interpretations did not have the chance to gain credence, since by that time, the philologists John Morris-Jones and his student Ifor Williams were already working carefully and systematically on the analysis of the texts. Morris-Jones' views are presented in his monograph, Taliesin (Morris-Jones op. cit.), and work on the text in this period culminated in the standard edition of Ifor Williams, Cam Aneirin (= CA; I. Williams 1938). 0.2

At this point it is opportune to state briefly the nature of the manuscript in question and that of the text it contains. The manuscript and its history have recently received authoritative description by Daniel Huws, in the introduction to the new, colour facsimile (Huws 1989; the work also reproduces Evans' diplomatic edition, with corrigenda). The manuscript is officially designated Cardiff, South Glamorgan Library ms. 2.81. It consists of nineteen leaves of parchment (with two seventeenth-century flyleaves at the beginning). It is incomplete, three leaves being lost from the last quire, leaving visible stubs. It cannot be stated whether these leaves were filled with writing, and whether there was any further quire, but the abrupt breaking off of the text at the foot of p. 38 indicates that the text did continue for some unspecifiable space on the lost leaves. The manuscript is datable on palaeographic grounds io the middle of the thirteenth century (J. G. Evans 1902:91-92; Denholm-Young 1954:44). The manuscript is written, in what can be described in the most neutral terms as a form of Middle Welsh, by two unknown hands, called A and B. A wrote pp. 1-23.5, and pp. 25-30.11. B wrote pp. 23.6-24, and 30.12-38, and presumably the continuation of the text on the lost leaves. On p. 1 of the manuscript is the rubric Hwn yw e gododin. aneirin ae cant This is the Gododdin. Aneirin sang it'. The significance of this rubric is discussed by O Hehir (1988: 57-58), but some of his conclusions must be handled with caution. Important for our present purposes is that the attribution of the text to Aneirin is explicitly made in the manuscript itself, and that at least some part of the text is called Gododdin. The text of pp. 25-30.11 consists of the so-called Gwarchaneu, Gwarchan Tudßvlch and Gwarchan Cynfelyn, poems celebrating the heroes named in the titles, between them, Gwarchan Adebon, apparently a short collection of versified maxims or proverbs, and following them, Gwarchan Maelderw, the obscurity of which has so far

defied satisfactory interpretation: here too, it will not be considered, though the time may be ripe to attempt an edition of this strange piece. It has not been possible for me to consult Klar/Sweetser (forthcoming). Surrounding these four independent poems, the rest of the text consists of a series of 130 shorter poems, or awdlau, 88 in hand A, 42 in hand B (the last being incomplete), celebrating many heroes apparently associated with the tribe of the Gododdin < Brit. *WotadJm, who appear in Ptolemy's Geography as Otadinoi, located in the south-east of what is now Scotland. Many of the individual poems have variants in other parts of the text (cf. the schematic presentation of the distribution of the variants by Dumville 1988: 9, from the point of view of the Α-text). The presence of these variants led I. Williams to adopt the dubious practice in CA of altering the order of the text by bringing all the variants of an individual awdl together so that they always followed the first variant to appear (cf. Isaac 1993a: 89-90 for consideration of problems arising from this practice). So much can be stated as fact Apart from refining details of the palaeographical description, whatever statements that are made that go further than these bald comments, regarding the nature and history of the text are necessarily the result of discussion, argumentation and speculation, in various proportions, so that it is unrealistic to suppose that there is consensus as to these matters, though one view, considered below, very nearly had this status until a few years ago. Various perspectives on the problems of the text will be considered in the following pages. To the issues covered there may be noted the discussions of the metrical properties of the poetry, CA Ixxff., Donovan (1975), D. E. Evans (1978), Klar et al. (1983/84), Sweetser (1988), Isaac (1992; 1993c). Here also, there is, as yet, no consensus. A convenient bibliography of works on, or relating to the text in all its facets is given by Jarman (1988: Ixxxix-xcviii). 0.3 Quantitatively, most research into the text in this century has been concentrated on two issues. The first is relatively straightforward, and concerns simply the understanding of the text as it stands. Such work has inevitably mainly taken the form of lexicography, a large amount of literature having grown up with the aim of explaining the many individual obscure words in the text. Apart from the two monumental works of LloydJones, Geirfa Barddoniaeth Gynnar Gymraeg, and I. Williams, CA (the notes to which are overwhelmingly of a lexicographic nature), many lexicographic notes have appeared in journals throughout this century. These notes are referred to in the commentaries in section Π of the present work. The other issue which has most taxed scholars, and not just in this century, is far

more problematic, and can be characterised as the question of the 'authenticity' of the

text. Is it largely the work of a sixth-century North-British poet named Aneirin, or is it the work of later ages, and if so how late? It is immediately obvious that these questions are reducible to the sober matter of dating the text: simply, how old is it? This could be thought of as a matter of philology which could be approached with a cold, critical eye, to come to the optimal, objective conclusion, within the parameters set by the nature and constitution of the text But in reality, the issue has been approached in a rather more subjective manner. It has been a matter of principle amongst some scholars to try to make the text as old as possible. And the matter has been further complicated precisely by the factor of 'authenticity'. This concept is ultimately rather incongruous in the debate surrounding the contents of the Book of Aneirin. It can be traced to the romantic eighteenth century and the controversy surrounding Mac Pherson's Ossian: were these pieces really translations of ancient Gaelic ballads, or were they the product of the imagination of that patriotic Scotsman, anxious to show to the world the glories of Scottish literature? In this case, there is a very clear issue of authenticity: Ossian was either what it was claimed to be or it was a modern production. Now it can and should be wondered what role such considerations can play in the analysis of a mediaeval text None, I would think. The Gododdin is a mediaeval text, and there are indications that it belongs to the Dark Ages. It is one of the tasks of philology to try to pin down as far as possible when it, or various parts of it came into being. But this is not a matter of 'authenticity'. The text is, and always will be, an authentic mediaeval text The matter of the attribution to Aneirin also has nothing to do with 'authenticity'. This attribution has rather the nature of a scholarly commentary. What has to be assessed by modem scholarship is whether the attribution to Aneirin by the mediaeval scholar is correct or not, or to what extent it is correct. But this is not comparable to the issue of whether Mac Pherson was a scholar of integrity presenting a startling discovery, or a well-meaning but, in the end, dishonest patriot Nevertheless, the issue has been discussed in terms of authenticity, and the arguments surrounding the matter fall into the two categories of historical vs. linguistic. I am not qualified to speak authoritatively on historical matters, and can only reproduce, as critically as possible, the arguments of others in this respect The historical argument for the 'authenticity' of the Gododdin, and of Dark-Age Welsh poetry in general, was first stated by Turner (1803). His title is symptomatic: Λ Vindication of the Genuineness of the Antient British Poems of Aneurin [sic], Taliesin, Llywarch Hen, and Merdhin, with Specimens of the Poems. While twentieth-century scholarship has naturally refined the arguments with more data, and more accurate data, and greater methodological precision, the historical arguments for the 'genuineness' of the poetry, presented by Morris-Jones (1918), Williams, in the introduction to CA and in the essays in Williams (1972), Jackson, in the introduction to GOSP, and Charles-

Edwards (1978), can all be regarded as being in direct line of descent from Turner's arguments. Simplistically, paired down to the barest minimum, the argument might be stated as follows: i) ii)

The Gododdin is attributed both within the manuscript and by other mediaeval sources to Aneirin. Aneirin (together with Taliesin) is recorded in one (twelfth-century) copy, Harleian 3859, fol. 188b, of the ninth-century Historia Brittonum, attributed to Nennius, as being a poet of the sixth century (cited from Morris 1980: 78): Tune Talhaem Tataguen in poemate claruit; et Neirin, et Taliessin, et Blucbard, et Cian, qui vocatur Gueinth Guaut, simul uno tempore in poemate Brittannico claruent

iii)

The geographical setting of the poetry is that of North Britain, southern

iv)

Scotland. The content of the poetry portrays a situation of independent North-British realms in this region fighting against advancing Germanic peoples, consistent with the supposed historical situation of the sixth century.

This is the essence. There are many other points which are alluded to in the argumentation of the various scholars mentioned. In particular, Charles-Edwards (1978) gives detailed consideration to the organisation of the society that is depicted in the poetry, concluding that the heroic principles that determine and regulate the behaviour and relationships of the figures concerned, and in which the poet himself has an active part to play, point to the sixth-century 'authenticity' of the text. More discussion of the role of heroic values in early Welsh poetry is to be found in Jarman (1967, 1982, 1988: xl-xlvii). Dumville (1988) has taken a very critical look at the historical argumentation that has been brought to bear on the question of the 'authenticity* of the text of the Book of Aneirin. He points up the real problem of the historian in connection with the text, referring to 'the scepticism which the historian must feel when faced with using as primary sources essentially undated mediaeval texts of a determinedly literary quality' (op. cit: 1). From the philological point of view, the dating of a text is, ideally, simply a goal, but for the historian, it is the means by which the contribution sources may make to the understanding of a period can be tested. On the other hand, he reduces the question of the 'authenticity' of the Gododdin to the single, but complex issue of transmission. The fundamental problem is to understand how the Gododdin came to be copied into a manuscript of unknown provenance in the middle of the thirteenth century. The model of transmission presented by Jackson in GOSP 62ff. has the corpus composed orally by

Aneirin in the sixth century, handed down from generation to generation of poets for two or three centuries, to be committed to writing in the eighth or ninth centuries. Dumville (op. cit: 5-7) has shown that this model is far from being the only possibility, even if it were in itself satisfactory. Also, Koch (1985/86) has argued for very early manuscript tradition of the poetry. It is quite symptomatic that hypotheses of oral composition and transmission of early European heroic literature are not viewed with the same enthusiasm today as they were in the 1950s and 1960s. 0.4

The other proving ground for the dating of the Gododdin has been its language. The general impression has always been that the language of the text appears rather archaic compared with the sort of Middle Welsh that is authoritatively described by Simon Evans in GMW. But the question has always been, how archaic? There is a fundamental problem about early Welsh texts which is pointed up by comparison with the situation in early Irish studies. Sagas and poems preserved in Middle Irish manuscripts can be dated to the Old Irish period, and even to different centuries within the Old Irish period, because we have an extensive corpus of Old Irish materials, mainly glosses, preserved in contemporary manuscripts which can be dated with great reliability by external criteria to the eighth, ninth and tenth centuries. That is, we can see the development of Old Irish happening before our very eyes in contemporary sources and are thus in a position to note which features appeared when, and disappeared when, and on that basis can use the absence or presence of such features in texts preserved in later manuscripts to determine the original date of such texts. That is in itself no easy matter, but the methods and resources necessary are available. When we turn to the early Welsh sources, we are faced with a great problem. The extensive corpus of datable Old Welsh materials is just not there. The Old Welsh sources that are available are too meagre for us to observe the appearance and disappearance of the features that would be needed for dating of texts preserved in later manuscripts. We do not see the continuum of development, but only a few points on the continuum, so that we cannot know whether such and such a feature is a genuine development or an idiosyncracy of a particular scribe or region, even when we can understand a particular source. The Old Breton sources, which could provide useful corroboration if used judiciously, are likewise too scanty to expand our view considerably. An example. The sort of morphological variation of verbal forms attested in Middle Welsh texts, and well illustrated by the collection from the Book of Aneirin given in note 1 of section Π, is no doubt analogous to the sort of variety seen in Middle Irish texts in this respect But in the latter case, it is generally possible sort out various strands and historical layers of development which have become confused in the texts, and to see how and when these strands and layers came into being. For Middle Welsh this is not

8 possible, at least by the same methods as are employed for Irish. All we can see directly is the confusion. Faced with this problem of lack of good attestation of earlier stages of Welsh, which would make the developments of the period accessible, the attempt must be made to reconstruct as much of these developments as possible, by familiar comparative-historical methods. The masterpiece of this approach is Jackson's Language and History in Early Britain (= LHEB). This work is in effect the expansion and refinement of the historical linguistic arguments presented by Morris-Jones (1918: 27ff., in particular 33-34), and Williams in the paper "When did British become Welsh?' (first published 1939, and available in Williams 1972: 1-15). The conclusion of all this work has been that Welsh is indeed old enough as a language for poems composed in northern Britain towards the end of the sixth century to survive intact to appear in Middle Welsh guise in the thirteenth century. However, masterly and comprehensive as LHEB is, the fact remains that it was written precisely with the hidden agenda of providing just such criteria as would make this possible. It too is a disguised 'Vindication of the genuineness ... etc.' A glance at the phonological developments placed by Jackson in the sixth century shows a number of decisive changes, which might be expected to have an effect on the poetry composed before they took place: middle of the sixth century, completion of the loss of final syllables, syncope of the composition vowel; mid to later sixth century, syncope of other unstressed internal syllables. The dating of these changes are all conveniently earlier than the supposed date of the Gododdin, and the poems of Taliesin, late sixth century. But how precise are these dates? How precise can they be? If one or two of these changes were in reality still taking place half a century later, then the dating of the poetry towards the end of the sixth century would be rather problematic. Then again, Dumville (1988: 1,3-4) has suggested that the middle of the sixth century must be included within the possible period of composition of the poetry, if it is genuine, on the basis of the historical background of the events apparently portrayed. Can we be sure that these changes would not have left a radical mark on any compositions from this period? Or can we push the phonological developments back a little in time to compensate? These are just phonological changes. Can we exclude the possibility that historical changes in the morphology would have made it impossible for a poem composed at this date to have been preserved later somewhat modernised in Middle Welsh times? Koch (1983a) has given some consideration to the historical development of nominal morphology, maintaining the line that all the changes required were early enough not to pose a problem for the hypothesis of sixth-century origin of the poetry. What of the verbal system? Cf. the reassessment of some issues of LHEB, though general accord, in Bammesberger/Wollmann (1990). Even if the results of LHEB are allowed to stand (which a large proportion undoubtedly must), it is one thing to point to the development of the Welsh language,

and say that it does not disallow the possibility that the poetry stems from the sixth century. It is quite another matter to find features in the poetry which positively point to that early date. Matters such as these were raised by Greene (1971) and MacCana (1971) to cast doubt on the early date from this perspective. And they insist that there is nothing in the Gododdin that forces us back to the sixth century, and that the ninth century is the earliest one can trace it, mainly on orthographic grounds (though Dumville 1977, 1985, suggests that even this date is, on purely orthographic grounds, rather optimistic). This view is confirmed by D. S. Evans (1978: 87-88): Yr ym wedi edrych ar y canu ο ddau gyfeiriad. Ο un o'r cyfeiriadau hyn mae gennym gyflawnder ο dystiolaeth i'n cyfarwyddo, a gallwn weld yn l o'r drydedd ganrif ar ddeg mor bell a'r nawfed, ond fawr ddim pellach. O'r cyfeiriad arall, ο gyfriod a chyflwr dadfeiliad y Frythoneg, go brin fod yn y canu ddim y buasai'n deg i ni ddisgwyl ei weld mewn gwaith a luniwyd yn niwedd y chweched ganrif.

'We have looked at the poetry from two directions. From one of these directions we have plenty of evidence to guide us, and we can see back from the thirteenth century as far as the ninth, but not much further. From the other direction, from the period and condition of the break-up of Brittonic, there is nothing really in the poetry that we could fairly expect to see in a work created at the end of the sixth century.' Despite these words of caution, or perhaps in answer to them, John T. Koch has, over the past decade and a half, taken up the position that the Gododdin, and some other works of early Welsh poetry, do indeed remount, to the sixth century, or at least parts of them. In his dissertation (Koch 1985a), and the publications derived from it (Koch 1983a, 1985b, 1985/86, 1987a, 1987b, 1988), he has extensively documented his view that there are indeed features of the language of the poetry that can be regarded as linguistic evidence of this early date. But Koch's arguments are intrinsically bound up with his idiosyncratic views on the development of Celtic and Brittonic grammar. Furthermore, while there are no doubt linguistic features in the poetry which can be considered consistent with the hypothesis of origin in the sixth century, there has been, as yet, no demonstration that any of these features uniquely force us to the positive conclusion that the works, or parts of them, must have been composed in the sixth century. Koch's position with regard to the syntax of the poems is considered in chapter 3, and sporadically in the preceding chapters. For other considerations regarding his arguments, cf. note 9 in section Π (on archaic case endings), and Isaac (1993c; on 'archaic' patterns of alliteration in the poetry). At present, it is best to regard the view of D. S. Evans, quoted above, as the authoritative statement on the matter of where the language of the poetry leads us with regards to its dating.

10 0.5

Compared with the large amount of work which has been done on the phonology and lexicography of the early poetry, and the reasonable amount on its morphology, the researcher soon finds that work on its syntax was, until recently, almost non-existent. It can be said that Henry Lewis initiated work in this field in his note on the position of the verb in the Hengerdd (Lewis 1928). He observes briefly that the verb is not restricted to the beginning of its clause in the early poetry, and is frequently attested medially, following its subject or object After drawing attention to word-order patterns of the Gaulish inscriptions available at the time, he concludes that the verb-medial orders of the Hengerdd are archaisms, reflexes of a period of Welsh or Brittonic when the verb was not fixed in position at the beginning of the clause. Lewis (op. cil) and later Lewis (1950) are no more than presentations of some very interesting observations. There is no attempt to develop a consistent theory which captures and explains the phenomena noted. The latter task is undertaken in two works, Lewis (1931 [1989]: chapter viii; 1942b). Here, his approach is diachronic, and he attempts to build up a picture of the early development of Welsh syntax. The theories presented in Lewis (1931) and Lewis (1942b) differ slightly in detail, but have the same foundation. The more elaborate theory of Lewis (1942b) will be summarised here, the difference to that of Lewis (1931) being noted afterwards. Lewis argues, following from the observations of the note from 1928, that Brittonic and early Welsh had a relatively free word order, that the constituents were not bound to any particular order in the sentence, so that either subject (S) or verb (V) could begin a clause, without any distinction in the meaning being discernible, e.g.: i)

echyngawd deigyr bard ywch y eneu Ά tear has afflicted the poet on his cheek1

ii)

(RBPcol. 1205. 2). VS Dewi differwys y eglwysseu 'Dewi defended his churches' (L1H 205.1). SV

In principle, all that happened between this early stage and the stage of Modern Welsh, where the order is strictly verb-initial in non-emphatic clauses, is that pattern i) has been generalised at the expense of pattern ii). However, Lewis' theory attempts to account for some other developments that took place in between, specifically the development of the Middle Welsh 'abnormal' sentence. Lewis notes that when an object pronoun is required, this is infixed in a compound verb in the oldest pattern, as in Old Irish, e.g. (subject-initial): iii) deus dy-m-gwares 'God save me' (BT 41.2).

π If the verb is simple, the pronoun cannot be infixed but must be placed before the verb, 'supported' by the 'meaningless' particle a, e.g.: iv) ttawurydet a-m dwc 'Sadness seizes me' (L1H 111.7). When the grammar so developed that compound verbs were no longer felt as separable, so that infixing was no longer a possibility in the language, the pattern with the particle a was generalised to such instances also, e.g.: v) Duw a-m difero 'God defend me' (L1H 33.30). And from the basis of patterns iv) and v) the particle was then further generalised to instances without object pronoun, giving the familiar Middle Welsh 'abnormal' sentence,

e.g.: vi)

Pwyll, pendefig Dyuet, a oed yn arglwyd ar seith cantrefDyuet 'Pwyll, chief of Dyfed, was lord over the seven hundreds of Dyfed' (PKM 1.1-2).

The difference between this theory and that of Lewis (1931) is that in the latter, the insertion of the particle α into the construction with initial subject, to give the 'abnormal1 sentence, takes place simply by analogy with the 'mixed1 sentence, with contrastive focus on the subject, e.g. in Lewis' example (1931: 117), Gwyra gyrchodd 'It was men who attacked'. Lewis notes (1942b: 275-76) that subject initial patterns are still sporadically used in modern colloquial Welsh, e.g. in the oaths Duw cato pawb or Duw cato ni 'God keep everyone/us'. Also, in the context described by Lewis as follows (loc. cit.): Again, if we chanced in walking along a country road to meet straying cattle, and inquired the cause, we should as likely as not be told y ffermwr adawodd y glwyd ar agor 'the fanner left the gate open1. We should be hearing an expression in a form that undoubtedly goes back to very early times. For the sentence does not mean 'it was the farmer that left the gate open1. There is no emphasis at all on the subject of the sentence. A similar little anecdote illustrates the same Modem Welsh construction in Lewis (1931: 118-19). The role of Gaulish, and of the reconstruction of earlier patterns of syntax, in Lewis' argumentation is discussed in chapter 3 below. Whatever the merits or dismerits of the theory, it clearly constitutes no detailed theory of the syntax of the text of the Book of Aneirin, or of early Welsh poetry in general. But it is central to the theory that the patterns seen in the early verse, where the subject may

12

precede the verb, i) are archaic, and ii) do not differ in meaning or function from verbinitial instances. As far as archaism is concerned, there is the obvious point that the priority Lewis assigns to the patterns of the Book of Aneirin is presumably influenced by the assumed very early (largely sixth-century) date of the corpus: if the corpus were demonstrably of twelfth-century date, say, then the historical-linguistic conclusions may have been somewhat different. The point is just that, since the date of the Gododdin and other examples of Hengerdd, is in doubt, the presence of a particular unusual linguistic feature in that corpus is in itself no conclusive indication of the antiquity of that feature. 0.6 Lewis' theory held the field of early Welsh syntax until the appearance of a paper by MacCana (1973). Regarding the syntactic patterns of early Welsh poetry, MacCana considers these unacceptable as indications of the syntax of the spoken language of any period: 'Strict verse is a notoriously poor guide to normal speech patterns, and its evidence can only be accepted when it can be tested against some sort of prose text' (op. ciL: 94). He also points out (op. cit.: 104) that Lewis' conception of emphasis is too limited to do justice to the possibilities of expression that may be present in different syntactic patterns (op. cit.: 104):

The basic error here, I would suggest, is that emphasis is considered solely in terms of contrasting emphasis relating to a specific element in the sentence, whereas ... a change of word-order may convey various nuances of emphasis affecting the substance of the whole statement. For Lewis' example of the farmer and the gate (and his earlier one from Lewis 1931), MacCana points out (op. cit.: 104-5) that it is a common strategy in both Welsh and Irish for elements of the sentence to be fronted in responses to questions. Cf. Greene (1971: 9), who also casts doubt on Lewis' theory from the point of view of comparativehistorical studies, discussed in chapter 3 below. Much of MacCana's paper deals with the construction known as nominativus pendens, so-called from the fact that some element of a sentence is placed, 'hanging' as it were, at the beginning of the sentence, in a language with case-marking morphology, in the nominative. The role of this "hanging' element in the sentence following is indicated by the presence in the appropriate position of a pronominal element, the so-called resumptive pronoun. E.g. (MacCana op. cit. 95, 97): OW grefiat guetig nis minn tutbulc hoi cenetl in ois oisou Tudfwlch and his people will require no title ever afterwards' (Surexit Memorandum). Grefiat 'title* is the nominativus pendens, the object pronoun -s is the resumptive pronoun. This construction is discussed in detail in chapter 2 of the present work, where, like MacCana, I suggest that it has played a role in the development

13

of the Middle Welsh 'abnormal' sentence. The construction was discussed at length in the context of Old Irish by Mac Coisdealbha (1976). T. A. Watkins (1977/78) likewise criticises Lewis' theory, for the same reasons, i.e. i) it relies too much on the assumption that patterns of syntax in poetry reflect well spoken usage, and ii) the theory takes no account of functional distinctions that exist between various constructions. Watkins' own solution to the matter of initial subjects (or objects) in the 'abnormal' sentence is to claim that the 'abnormal' sentence in fact does not exist as a separate construction, but just reflects a particular usage of the construction seen in contrastive statements, such as Gwyr a gyrchodd, Lewis' example quoted above. In Watkins' analysis, the construction with apparent relative is called the 'relative order1 ('y drefn berthynol'), and its uses are summarised as follows (op. cit.: 395; translated): i) ii) iii) iv) v)

Adjectival clauses (i.e. the traditional 'relative clause'). Contrastive statements (the 'mixed sentence' in Welsh grammatical terms). Pronominal questions, and responsive. Affirmative hortative. (In some circumstances) neutral statement

In later publications, Watkins (1983/84, 1987, 1988, 1991) develops and applies this analysis further, adopting the term 'cleft order' in place of 'relative order', for the construction in question in principle clauses. In chapter 2, below, it will be shown that the strategy of collapsing contrastive sentences and the 'abnormal' sentence into a single construction is both formally and functionally unsatisfactory. In this connection, it may be noted that Wagner (1967, 1977) too, regards the unusual word-order patterns of early Welsh (and early Irish) poetry as artificial constructions, with little relation to the grammatical structures of colloquial speech. Wagner, however, is arguing in quite another context, one which is given greater attention in chapter 3 of the present work. 0.7 Despite the fair warnings of MacCana and Watkins about the need to pay attention to subtle nuances of difference in the functions of various constructions, and about the dangers of using poetic examples as evidence in theories of linguistic development, the recent work of John T. Koch, represents, in syntactic matters, an explicit throwback to the methods and ideas of Lewis. Since Koch, however, draws much more extensively than Lewis on the reconstructive methods of comparative-historical linguistics, the nature of his argumentation, and particularly the role of the evidence of Gaulish is given consideration in chapter 3, below, on reconstructing syntax. One more general point may be made here, however. Utilising the unusual syntactic patterns of the Hengerdd, and the Book of Aneirin in particular, as evidence of the early date of the corpus forces us to tread a very thin line bordering on circularity of argument

14 The syntax can only be evidence of early date if it is demonstrable that it is genuinely older than more familiar patterns. But it will be remembered that one of the reasons Lewis, followed by Koch, was able to consider the subject-initial orders as archaic was because they were found in the supposedly ancient Hengerdd. The danger of circularity in the argumentation is evident Koch attempts to avoid this danger by means of the Gaulish data and comparative-historical reconstruction, but if these criteria themselves are in doubt (see chapter 3), then nothing remains to assign early status to the aberrant syntax, and that criterion falls away as evidence of the early date of the poetry. The use of linguistic criteria for dating early Welsh verse is fraught with such considerations. 0.8

The works on the syntax of the Book of Aneirin considered so far have been concerned with constituent order in the sentence, and its position in historical linguistic argumentation. One other syntactic matter has been the subject of debate in discussions of the nature and date of the Book of Aneirin. That of morphological agreement Adjectives modifying plural nouns appear in the text both in singular and in plural forms. In this work on the syntax surrounding the finite verb, this matter will not be considered, but a certain amount of attention will be given to the other controversial agreement pattern, that of the finite verb in the singular with a plural subject The text of the Book of Aneirin is, as far as has been ascertained till now, by no means consistent in this, and thus appears to oscillate between the familiar pattern of later Welsh, with a third person singular form of the verb with a nominal subject in some instances, even when the latter is plural, and a pattern of congruence of a plural verb with a plural subject in other instances. This circumstance has been variously interpreted. So DiUon/Chadwick (1967: 216ff.), MacCana (1971: 321-23), Greene (1971: 8) all voice their suspicion that the presence of instances of singular verbs with plural subjects is a feature indicating later rather than earlier date, and that a work supposed to date from the sixth century should much more consistently show concord of a plural verb with a plural subject, presumed to be the older pattern, on the basis of Old Irish and Indo-European grammarin general (incidentally, D. S. Evans 1971:44-45 shows that in all instances in Old Welsh with a plural subject, the verb has plural morphology). Jackson, GOSP 90 and (1973/74:2-3), on the other hand, argues back that this view rests on argumenta ex. silentio, that we do not and cannot know what the pattern of agreement between a verb and a plural subject was in North-British Welsh of the sixth century. Then again, D. S. Evans (op. cit, and briefly in 1978: 77) argued in the other direction entirely, that lack of agreement was always the rule in Welsh, and that instances of agreement that can be seen in early sources are due to the influence of Latin learning on the native grammatical tradition: in this he follows Morris-Jones (in Morris-Jones/Rhys 1894: xxvi; MorrisJones 1931:191). A less extreme view, that both presence and absence of plural subject-

15

verb agreement are idiomatic and old in Welsh, is expressed by Lewis (1942b: 16-17). These matters are taken up in chapter 1 of the present work. 0.9 The problem of the syntax of the language of the Book of Aneirin has, then, always been bound up with, and discussed along with, many other issues: dating poetry, historical syntax, Continental Celtic, Old Irish, etc. In particular, there has been a tendency either i) to hail it rather gratuitously as 'archaic', and therefore as evidence of early structure, or ii) to dismiss it as 'poetic', and therefore irrelevant to discussions of the grammar of early Welsh in general. My feeling is that neither approach does justice either to the text itself (which is, or should be, more than just a proving ground for historical argumentation), or to the contribution the text might make to the understanding of issues of wider context. My aim in the present work is to attempt partially to clear away the trees in order to see a bit more of the wood than has been visible to date, by making a thorough, internal study of the syntax of the text in question, free, to the greatest extent possible, of any ulterior theoretical motives other than that of understanding the syntax of the text in its own terms. By this means the paradox that the syntax of the Book of Aneirin is utilised in so much other argumentation without being understood in itself might be resolved.

0.9.1 The following questions may be regarded as encapsulating various considerations relevant to the discussion of syntax in the Book of Aneirin. I make no claim to have explicitly and exhaustively addressed in my study all of the issues outlined below. Rather, I consider these issues as constituting a loose programme for which I hope the present work might be viewed favourably as a preliminary essay, or, taking fairer account of prior work in the field, as an interim report on work in progress. 0.9.2 What are the syntactic structures employed by the poets of the Book of Aneirin? As stated above, I concentrate my attention on the syntax associated with the finite verb (in terms of constituent order in the verbal sentence, interaction of syntax and morphology, patterns of agreement, semantics of predication). However, further work might extend the analysis to consideration of other aspects of the syntax, regarding, perhaps, nominal sentences, syntax of the copula and substantive verb, syntax of adjectives, syntax of possession and other genitival relationships. Also, it would be quite natural in the future to extend consideration in the same detail to other works or corpora of the Hengerdd.

16

0.9.3 With the limitations on the scope of the present work presented in the previous paragraph, in mind, what are the relationships between the various syntactic structures of the Book of Aneirin itself? I shall consider ways in which different patterns may be in complementary distribution to one another, or whether some patterns really vary facultatively, to determine the manner in which the poets used different syntactic patterns, possibly motivated by semantic or poetic contentual considerations, possibly as purely formal variation, causa metri for instance. Thus, certain patterns may be more basic than others, certain patterns may be marked or contextually determined in ways which have escaped notice so far. This approach may partially be regarded as a preliminary attempt to formulate a 'poetic grammar1 of certain aspects of the language of the Book of Aneirin, independent of consideration as to what patterns were used in the language or register actually spoken at that period, as far as that might be determinable (see section 0.9.5).

0.9.4 What are the relationships between the syntactic structures of the Book of Aneirin and that of other, contemporaneous sources? It may be that it is possible specify in detail substantive differences between the syntactic practices of the poets of the Book of Aneirin and those of, say, the Gogynßirdd (12th-14th centuries, roughly contemporaries with the manuscript of the corpus under scrutiny in the present study), or those identifiable in Middle Welsh prose. Differences of genre might be identifiable in addition. That such differences between the various corpus types are indeed present has already been documented, and given rise to the historical speculations noted above. The methodological error up to now has been to interpret the data from the Book of Aneirin only in terms of the more familiar patterns of other sources (whether or not the latter have been correctly analysed), and unsystematically, so that examples have been taken arbitrarily as illustrations of a ready-made thesis, rather than being analysed systematically as individual realisations of the overall syntactic-structural conditions of the language, in possibly determined, and therefore, identifiable contexts.

0.9.5 What are the relationships between the syntactic structures of the Book of Aneirin and those of contemporaneous spoken Welsh? This is a particularly delicate matter, and is accompanied by the risk of giving way to groundless speculation, because the spoken Welsh of the period in question is, of course, directly quite inaccessible to us. Nevertheless, the question might be given

17 consideration, if only from a methodological perspective. It may be that various possibilities can be discussed, drawing together data from comparative-historical and synchronic analyses, and that the 'poetic grammar' referred to above might be seen to diverge in a systematic manner from the conventional grammar of the every-day Welsh of the period, and in ways which are accessible and analysable, such that the deviations might be interpreted as being conditioned by the properties of the conventional grammar itself. Poetic language cannot be an arbitrary distortion of the spoken language, for the simple reason that it must still, even if only to a minority of specialists within the population, be meaningful: it must make sense, and must therefore be grammatical, in precisely the same sense as an utterance in the colloquial register must be grammatical. The 'poetic grammar1 referred to may be considered as a systematic set of 'poetic licences', in the conventional sense, which may consist of the relaxation or restructuring of certain rules or tendencies that must be part of the grammar of the spoken language in its entirety. So it may well be that the structures of certain common syntactic patterns in the poetic context are virtually present in the spoken language, but for some reason or other, are obscured, or much rarer than in the poetry, because of the interference by other factors which are filtered out in the poetic context. But this makes the study of the patterns of the poetry no less relevant to the study of the language as a whole and its history, especially when the ordinary language of the period is inaccessible to us, as is the case here. In this way, the syntactic analysis of the language of the Book of Aneirin (and early Welsh poetry in general) might after all cast light on the structure of the everyday Welsh of the period in question, or even of the modern period.

0.9.6 What are the relationships between the syntactic structures of the Book of Aneirin and the history of Welsh syntax? This question is naturally related to the preceding one and also to the next one, but deserves separate consideration. The structures of the Book of Aneirin, or some of them, might be archaic, or they might not. If they are archaic, then in what way. Conversely, they, or some of them, might be a purely 'literary' creation, or they might not. The historical interpretation would naturally differ according to the solutions to these problems, or according to the hypotheses set up in relation to these issues, and in ways which might have implications for the analysis of the history of Welsh in particular, or historical change of language in general. This question is implicit in many of the studies referred to in previous sections of this introduction, but so far without being formulated in terms of a sound, text-intemally based foundation of understanding of the underlying structure of the language of the Book of Aneirin itself.

18 0.9.7 What are the relationships between the syntactic structures of the Book of Aneirin and those of other Celtic languages? If a sufficient level of understanding of the syntactic structures of the Book of Aneirin and their possible language-historical context can be achieved, this may be of value in discussing the syntactic patterns, and their histories, of the other Celtic languages, as well as being relevant to periods of development which may be accessible only by reconstruction for the prehistoric proto-languages of the attested languages.

0.10 In the light of the considerations of the preceding paragraphs, certain additional methodological difficulties readily present themselves.

0.10.1 The text of the Book of Aneirin is in itself one of the most obscure and enigmatic of our early Welsh texts, even within the context of the Hengerdd as a whole, which always presents difficulties in interpretation at the best of times. Despite the efforts of MorrisJones, Williams, Jackson, Jarman, as well as countless other scholars of this century, who have made contributions on a smaller scale to the study of the text, the mere understanding of the basic sense of many passages is still in question. Naturally, no single research project can have as its aim the clarification of all such obscurities. But a syntactic analysis of a sentence or passage assumes that the meaning of the sentence or passage is to a greater or lesser degree accessible. For this reason, rather than presenting only a bare list of instances of finite verbal sentences in the text, I have added commentaries of differing lengths to most of the instances classified and presented in section II. These commentaries have the purpose of at least drawing attention to, if not in most cases discussing whatever difficulties may appear in the interpretation of the individual instances, and of justifying any decision I might make concerning the interpretation. In addition, there is in the classification a not inconsiderable category of instances ('?'), in which the difficulties associated with the interpretation seemed to me to be too great to justify a syntactic analysis of any sort. 0.10.2 The poetic nature of the text means that we must always be watchful for features and processes that are conditioned by more than the usual formal and functional demands of linguistic communication. Indeed, any theory of the language of the text must explicitly incorporate such considerations if it is to have general validity. In the notes to the instances in section II, I have noted cases where such considerations do appear to interact with the linguistic form of the text, though I have probably not done this consistently. In

19 addition, the discussion and analysis of the following chapters of section I repeatedly make reference to these matters. 0.10.3 The Book of Aneirin gives clear indications of being rather heterogeneous in its constitution. Consisting as it does of a large number (130) of smaller poems, together with the four Gwarchaneu, this structure has naturally lent itself to processes of accretion, so that various layers of differing historical and geographical provenance must be reckoned with. Many individual poems in the text have variants in other parts of the text, and the nature of the variation is not always obvious. At the most external level, the manuscript is written by the two clearly discernible hands A and B, the relationship between the passages in these two hands being unclear. There are also orthographic differences between different parts of the text. It must be considered therefore that there may be differing treatments of syntax to be identified in different parts of the text (or there may not). My approach to this particular difficulty is the same as the one I adopted with regard to the metrics of the Hengerdd in Isaac (1992). Rather than being immediately blinded by the heterogeneity, with the result that effective progress is made impossible, I have treated and analysed the text in terms of a homogeneous overview. The theory developed on the basis of this overview may be in need of refinement when consideration is given to the ways different parts of the text are related to each other, or differ from each other. In this way a more detailed perspective might be developed by the application of the broad principles which appear by this analysis to specific problems of particular passages. But in the present work it was necessary to restrict attention first and foremost to the development of those broad principles. 0.10.4 It will be obvious from many of the points made so far that the historical linguistic context of the corpus embodied by the Book of Aneirin is far from clear. The only date that is reliable to within an accuracy of less than a century is that of the manuscript itself, the middle of the thirteenth century. This remains true despite the view (which is no longer so much of a consensus today as it was perhaps twenty-five years ago, if that is what it was then) upheld by Williams and Jackson amongst many others, and today championed energetically by Koch, that a substantial core of the poetry indeed remounts to the sixth century AD. There is a considerable portion of the text which can be reliably attributed to an Old Welsh exemplar or exemplars. But the poor attestation of Old Welsh (eighth-eleventh centuries) makes control in that direction difficult (cf. Dumville 1977, 1985). On the other hand there are indubitable differences between the language of the Book of Aneirin and that of the sort of Middle Welsh prose that was being written in the thirteenth and fourteenth centuries. In addition, the historical relationship between the

20 poetry of the Book of Aneirin and that of the datable poets of the Gogynfeirdd

is

ambiguous. These are all substantive issues which have been discussed on many occasions from differing points of view. There is no hope of solving such problems in the space of some brief introductory comments, as they have been, and are, rather, subjects for separate extensive studies. Since these rather philological problems cannot be solved here, so that the results may with confidence be built into the syntactic analysis, I adopt the simple, if extreme, measure of largely disregarding them in the present study. The issue of dating the text will not be addressed. In the analysis of chapter 2, it becomes necessary to draw on a comparison of the language of the Book of Aneirin with that of Middle Welsh prose. But if it is concluded in that discussion that the syntax of the Book of Aneirin presents a somewhat earlier stage of development to that of the Middle Welsh prose, this is at least, precisely, a conclusion, not an assumption, and only regarding relative chronology. Nowhere in the present study will an attempt be made to assign any part of the text of the Book of Aneirin to a particular century.

0.11 Some results that may be hoped to be gained by a study of the syntax of the Book of Aneirin may be summarised explicitly as follows: i)

Improved understanding of the contents of the Book of Aneirin as a text in its own right, a classic of Dark-Age European literature still largely overlooked outside of the field of Celtic Studies, and in any case, the central oevre in the corpus of early Welsh poetry.

ii)

A comprehensive data base and analysis for further work on early Welsh

iii)

syntax. A contribution to the study of the history of Welsh poetry.

iv)

A contribution to the study of the history of the Welsh language.

v)

A contribution to the study of language in a poetically conditioned context.

vi)

Clarification (to a certain extent, though probably by no means conclusively) of the position of the Book of Aneirin in the history of the Welsh language.

vii)

Clarification (qualified as in vi) of the position of the Book of Aneirin, and early Welsh poetry in general, in the context of the syntax of other Celtic languages, e.g. light might be thrown on 'archaic/poetic' syntax in Old Irish.

viii) In respect of issues of wider implication and application, identification of tendencies/influences/relationships of European scope, oral vs. literate features, linguistic/literary typology, etc.

Section I 1. Statistical Analysis, Basic Order 1.1

A glance over the word order patterns attested in the corpus, as they are presented in the overview at the beginning of section II, reveals at first sight a bewildering array of variations. Stripped of the classification into non-pronominal and pronominal arguments, we have the orders, in varying proportions, VS, SV, VO, OV, VSO, SVO, VOS, SOV, and OVS, not including the clear instances of relative clauses and 'mixed' sentences, with contrastive focus on the subject or direct object (the main component of the category "Misc.* in section II). However, the presence of such variation is quite normal in languages. There are many sorts of factors which can trigger word-order variation. One only needs to consider the types of word order that are common in Modem Welsh to get an idea of how that can work:

la)

S V O Mae Siön wedi darllen y llyfr. is John after readvn. the book

Ib)

S 9 V Mae Sion wedi'i ddarllen e. is John after his read.vn. he

lc)

S V Siön sy wedi darllen y llyfr. John is.rel. after read.vn. the book

Id)

s

9 v

Siön sy wedi'i ddarllen e. John is.rel. after his read.vn. he

le)

V O S Wedi darllen y llyfr maeSion. after read.vn. the book is John

If)

O V S Wedi'i ddarllen e maeSion. after his read.vn. he is John

2a)

S V Fe wnaeth Siön ddarllen y llyfr. ptl. did John read.vn. the book

2b)

S V Fe wnaeth Siön ei ddarllen e. ptl. did John his read.vn. he

2c)

V O S Darllen y llyfr wnaeth Siön. readvn the book did John

22

2d) 3a) 3b) 3c)

O V S Ei ddarllen e wnaeth Siön. his read he did John V S O Fe ddarllenodd Siön y llyfr. ptl. read.prt. John the book S V Siön ddarllenodd y llyfr. John read.prt. the book O V S llyfr ddarllenodd Siön. the book read.prt. John

This list does not exhaust the permutations. The assignment of the values S(ubject), V(erb) and O(bject) follows the semantic criterion of the predication represented by all these sentences, which are all variants of the core predication, JOHN (READ)past THE BOOK. Here is precisely the same range of word order variation, in terms of VSO etc., as observed in the Book of Aneirin. In the case of the Modem Welsh examples, we have no difficulty in determining the basic order. The series 1) and 2) both show periphrasis of different sorts, 1) with the verb *be' as auxiliary and a prepositional construction, giving resultative aspect, 'John has read the book', 2) with the verb 'make, do' as auxiliary. Examples Ib, d, f) and 2b, d) have pronominal objects, which are subject to certain formal constraints (the possessive pronouns are proclitic). The series 3) shows neither periphrasis nor pronominalisation, but only the inflected verb coding the predicate and two nominal arguments (in fact, pronominalisation of either of the arguments would not affect the word-order patterns in 3). It is clear that adding extra elements to create periphrastic constructions (as in 1 and 2) results in greater complexity than in constructions with only the minimum of elements necessary to code the predication (as in 3). The last factor affecting the word order in the Modern Welsh examples is that of contrastive focus. Examples Ic, d, e, f), 2c, d) and 3b, c) all add to the content of the sentence by placing a particular emphasis on a certain element of the sentence. Thus Ic) may be translated with a cleft sentence, 'It is John who has read the book' or with contrastive stress, 'JOHN has read the book'. Similarly, 2c), 'John READ the book', or 3c), 'John read the BOOK'. In terms of meaning, what this contrastive focus adds to the sentences, e.g. for 3b) Sion ddarllenodd y llyfr 'JOHN read the book', is the information that 'John read the book, and no one else read the book*. Such a statement must be used in context. It is not synonymous with the statement that John is the only person in the world who read the book. Rather, it means that in a given discourse context, with a limited number of candidates who possibly read the book, John is the candidate who really read the book, and none of the others. Whether this is really true or not is beside

23 the point; what matters is that the speaker believes it to be true, and considers it an important piece of information to convey to the addressee(s). In any case, it is obvious that adding such additional information to the predication forming the core of the content of a sentence results in a realisation of the sentence which is more complex than one which conveys the core predication alone. In this light then, we can see that the only sentence from the examples given above which conveys only the core predication,

JOHN (READ)past THE BOOK, and has no other structural

complications such as periphrasis or pronominalisation is example 3a) Fe ddarllenodd Siön y llyfr 'John read the book', with the constituent order VSO. This is the basic order then. In the case of the corpus of instances from the Book of Aneirin, we must be careful in applying from the start such simple analytical methods as illustrated in the preceding paragraphs. It will be remembered that the whole problem here is that it is unclear what the functions of all the variant word order patterns are. Clear cases of contrastive focus can be discerned, and are noted as such in the commentaries following the instances in section II. But the fundamental issue here is first of all to establish just what the basic order, in the terms just prescribed, of the corpus was. This is not a trivial exercise, precisely because researchers pursuing certain historically oriented approaches have been at pains to argue for either a basic SVO order in the early poetry, or clear reflexes of a slightly earlier SVO basic order. In order to establish the status of such constructions in the poetry, it is of primary importance to establish as a starting point, or base line, what the basic order of the corpus really was, regardless of knowledge gleaned from other sources or corpora (e.g. Old Welsh from contemporary manuscripts, historical or area! considerations). The basic order of the Book of Aneirin must be established on the basis of the internal criteria of the comprehensive list of verbal sentences presented in section II.

In the face of such data, we have no recourse but to fall back on statistical analysis. It was established above that basic order can be negatively defined as the order of constituents in a sentence in which structural and semantic complications of various sorts are absent: basic order is thus in an important sense the 'simplest' order. On this basis we may make the assumption that simpler constructions are more likely to occur often in a corpus than more complex ones, and that therefore, indeed, if the corpus is large enough, the simpler constructions should predominate. This assumption is rooted in the PragueSchool theory of markedness: more complex, less frequent structures are marked, less complex, more frequent structures are unmarked (cf. Greenberg 1966a for a discussion of markedness over a range of grammatical phenomena). I will have to make the additional assumption that 603 instances constitutes a large enough corpus for such an analysis.

24

1.2 The least problematic conclusions to be drawn from the data of section II are those concerning the relative positions of the verb and the direct object (the classification of section II takes no account of indirect objects, and apart from occasional comments in this and the following chapter, indirect objects will not come under discussion in this work). Of 286 instances with direct objects, 211, 74%, show the order V(...)O, with or without intervening material (not including antecedents of relative clauses with the object function of the clauses, or instances with clearly fbcussed, i.e. fronted, objects). Of the remaining 75 instances of O(...)V, the object immediately precedes the verb in all but four instances. 46 of the 75 instances have pronominal objects, and 29 have nominal objects. The instances with pronominal objects present a clearly defined grammatical rule. There is only one instance of a pronominal object following the verb, i.e. V[O+pro]: inst. 307 ket ryladefl] hwy 'Although they were killed'. In conjunction with inst. 198, this forms CA1. 1128, ket rylade[t] hwy wy ladassant 'Although they have been killed, they killed*. The appearance of the object hwy after the verb in the first part of this line is clearly motivated by the resulting chiasmus, KILL + hwy. wy + KILL. Apart from this instance, all other instances of pronominal objects appear preverbally, as proclitics. These proclitic object pronouns correspond historically and functionally to the infixed pronouns of Old Irish, and are indeed commonly called infixed pronouns in Welsh grammar (e.g. GMW 55). However, while there are clear cases of true infixing within the corpus (instt. 476-81), the most general statement as to the positions of these object pronouns has them immediately preceding the verbal stem: I shall continue to refer to them therefore as proclitic or prefixed (object) pronouns. The full distribution of objects relative to the verb within the corpus is therefore as follows (regardless of what other arguments of the verb or other material are present in the individual instances, and where): V[O-pro] 210

[O-pro]V 29

V[O+pro] 1

[O+pro]V 46

instt.

In percentages, the proportions of VO to OV for non-pronominal objects and for pronominal objects are as in the table which follows:

-pro +pro

VO

OV

88

12

2

98

25 The rule is clear that non-pronominal objects most often follow the verb, while pronominal objects precede it, in fact, are procliticised to it (the same applies to pronominal indirect objects). Since, as discussed above, pronominalisation adds complexity to a sentence, we must consider the basic order, with regards only to the relative positions of the verb and its direct object, as VO. The motivation for the 12% of non-pronominal objects that precede their verbs, and the structural constraints that apply to these constructions, are to be discussed in chapter 2.

1.3 If the position of the direct object relative to the verb in basic order is unproblematic,

the position of the subject relative to the verb and the object is rather more difficult to establish. Bearing in mind that the concrete, central issue is whether basic order was VSO or SVO, we might begin by simply comparing the statistical frequencies of these two orders in the corpus. Ignoring pronominalisation for the moment, we observe that VSO is attested a total of 35 times (instL 355-89), while SVO is attested a total of 45 times (instL 390-434). On this basis, one might be inclined to assign basic status to SVO and attempt to derive VSO from it in some as yet undefined context or contexts. However, as was seen above in the case of the position of the object relative to the verb, it is misleading to ignore pronominalisation in assessing the basic order of the arguments, and the most reliable guide is the position of non-pronominal arguments. Of the 35 instances of VSO, 20 have non-pronominal subjects (instt 355-74), 15 have pronominal ones (instL 375-89). Of the 45 SVO instances, 14 have non-pronominal subjects (instL 390-403), 31 have pronominal ones (instL 404-34). Considering only the instances with non-pronominal subjects, VSO is statistically the better attested order. Since there are, however, so few instances with both arguments of transitive verbs explicit and non-pronominal, it is necessary to consider other data in reaching a final decision as to basic order. In the first place, the status of the constructions with pronominal subject must be considered It was clear in the discussion of pronominal objects above that there can be particular grammatical rules affecting the position of pronominal arguments within the clause. Since there can be no doubt that pronominalisation creates a more complex construction than one without pronominalisation, it is likely that any change in the position of a pronominal argument relative to the position of the equivalent nonpronominal argument is to be regarded as a deviation from basic order. Indeed, with 15 instances of VSO with pronominal subject against 31 of SVO with pronominal subject, we might conclude that, a similar rule of proposing subject pronouns operates as in the case of object pronouns. Such a conclusion would, however, be rather premature. After

26 all, the proportion of V[S+pro]O is not insignificant, and we are therefore not in a position to think of a firm grammatical rule, as was possible for pronominal objects. It seems that there is a tendency at work to place pronominal subjects before the verb, but the motivation for this does not emerge from the present general statistical survey on the basis of the individual instances alone. This matter will be dealt in detail in section 2.9. In addition, consideration should be given to the position of the subject relative to the verb in all instances in the corpus with an explicit subject, not only those with both subject and object This means including consideration of the patterns V[S-pro] (instt 101-76 = 76), [S-pro]V (instt. 177-89 = 13), [S+pro]V (instt. 190-204 = 15; note the absence of the pattern V[S+pro] in the light of the preceding paragraph), VOS (instt 435-53 = 19), SOV (instt. 454-61 = 8) and OVS (instt 462-75 = 14). Note that, since all instances of SOV have pronominal objects, this pattern is simply the equivalent of SV[O-pro], with the addition of the rule for proclitic object pronouns. Similarly, with the exception of inst 469, OVS is the equivalent of VS[O-pro] with the proclitic object pronoun. Ignoring the presence of the object, in order to generalise these patterns to VS vs. SV, we attain the following proportion: VS = 144 = 64% vs. SV = 81 = 36%. With postverbal subjects outnumbering preverbal subjects by more than 2: 1, we may be further justified in considering the postverbal position as basic for the subject, that is, in conjunction i) with the frequency of attestation of VSO itself relative to that of SVO, and ii) with the considerations regarding pronominalisation. The question naturally arises, if VSO is indeed basic, then why does the pattern SVO occur with not insignificant frequency at all? This is a matter of contextual and cotextual motivation, and is one of the objects of the discussion in chapter 2. It may be observed that it is not yet, in fact, VSO that has been established as basic. What has been established is that the postverbal position is basic for both subject and object, i.e., for the object we expect VO, and for the subject, VS. It has not, however, been clearly established in what order subject and object occur relative to each other in basic position. The attestation of V[S-pro]O 20 times in the corpus is challenged by the fact mat VOS (both arguments always non-pronominal) is attested 19 times. All that has been established so far is that we expect a verb-initial basic order nothing seen so far excludes VOS as a possible basic order, unusual as this is, especially for a European and Indo-European language, though in that respect not significantly more unusual than VSO. The easy way out could be taken by simply following common sense and accepting that, if we can establish verb-initial basic order corpus-internally, then our general knowledge of the Insular Celtic languages will surely point to VSO as the full basic order. It may, nevertheless, be considered whether there is not some internal criterion by which the matter can be settled.

27 In fact there is just such a criterion. In the following, I present all instances of V[S-pro]O and VOS in turn, noting the syllable count of the two nominal arguments involved, in the form, e.g., 2-4, i.e. for VSO, 'subject has two syllables, object has four syllables', for VOS, Object has two syllables, subject has four syllables'. V[S-pro]O: instt. 355 ny memws gwrawl gwadawl chwegrwn The brave man did not demand a father-in-law's dowry' (2-4); 356 goreu (yw) hwnn kyn kystlwn kerennyd enneint crei ac angeu oe hermyd 'Ramer than declaring peace, this man made a bath of blood and death of his enemy' (2-6); 357 treiglessyd llawr lloegnvys giwet The lone hero overthrew the English mob' (1-4); 358 gorgolches e greu y seirch "His blood washed over his armour1 (2-2); 359 ny deliis meirch neb marchlew 'No one's horses caught up with Marchlew' (2-2); 360 yt rannei rygu e rywin The most dear one would share out his plentiful wine1 (2-3); 361 e gwnaei varchlew waet()lin 'Marchlew would make bloodshed* (2-2); 362 ny weles vrf Jun e dot 'not one saw his father (again)' (2-2); 363 na wnelei kenon kelein That Cynon would not make a corpse' (2-2); 364 amuc moryen gwenwawt mirdyn 'Morien defended the fair song of Myrddin' (2-4); 365 nyt emduc mam mor eiryan gadarn "No mother gave birth to one so far (and) powerful' (1-5); 366 Er()kryn e alon ar(-)afery brwydrin "His enemies fear the weapon of the daring one of battle'(3-5); 367 dimcones [e] lovlen benn eryron llwyt "His glove/hand satiated the beaks of grey eagles' (3-5); 368 Dyfforthes meiwyr tnolut nyuet 'Cowards joined in the praising in the hall' (2-4); 369 mal ban llad llew llywywc 'As a lion kuls a ?...' (1-2); 370 na deliis meirch neb marchlew "That no one's horses caught up with Marchlew' (2-2); 371 ardwyei waetnerth e gerth or gat 'Gwaednerth would defend his rights through battle1 (2-2); 372 ladaut map nuithon ο eur()dorcnogyon cant ο deymet The son of Nwython killed a hundred gold-torqued princes' (3-10); 373 Niforthint ueiri molut niuet 'the stewards could not bear the praise in the hall' (2-4); 374 dirtis(h)ei etarif ]uet()iuet 'Edar derserved his mead-drinking' (2-4).

28 VOS: instt. 435 mob botgat gwnaeth gwynnyeith gwreith e law The son of Botgad, the deed of his hand wrought vengence' (2-3); 436 Gwyr a aeth gatraeth gan waver dygymyrrws eu hoet eu hanyanawr "The men who went to catraeth with the dawn, their mettle shortened their lives' (2^4); 437 neus goreu deufrj bwyllyat neirthyat gwychyd "Neirthiad the splendid performed his brave purpose' (3-4); 438 ny chryssyws gatraeth mawr mor ehelaeth e aruaeth "No great one of such extensive purpose rushed forth to Catraeth' (2-8); 439 dyfforthynt lynwyssawr gelorawr hir "Long biers would carry the corpses' (3-4); 440 diebyrth e gerth e gynghyr "His counsels deprived him of his rights' (2-3); 441 dryllydei vedgym eillt mynydawc The man of Mynyddawg deserved meadhorns' (2^); 442 nyt emda daear... mor eiryan gadarn 'One so fair and powerful does not travel the earth'(2-5); 443 Nyt efborthi gwarth gorsed senyllt The court of Senyllt bore no shame* (1-4); 444 nyphyrth mevyl moryal efnj dilyn 'Morial does not bear shame in pursuit' (1-5); 445 oed mor diachfajr yt ladei esgar gur()haual en [bedin] 'It was so cruelly that Gwrhafal in the army would kill the enemy' (2-6);

446 ny deliit kywyt kywrennin benn The skilful chieftain does not deserve sin(-ful treatment)1 (2-4); 447 dyrllydei med melys maglawr gwrys aergyngtys gan wawr The attack of the champion's court at dawn deserved the sweet, ensnaring mead' (5-6); 448 an dyrllys molet med melys maglawr 'the sweet, ensnaring mead deserves praise from us' (2-5); 449 dyrllydei vedgyrn vn mabferuarch The only son of Fferfarch deserved meadhorns' (2-4); 450 ardwyei conn ()wr arwr mynawc The noble hero used to defend a hundred men' (2-4); 451 ny phorthassan worth wyr ny thechyn The men who would not flee bore no shame* (1-4); 452 Llithyessit adara()da(a)m ... edeiuiniat The hand of the chieftain fed the birds' (2-5); 453 ny cheri gyfofhi gyvyeith The compatriate did not love being afraid' (3-2). In all instances of V[S-pro]O, the object either has a higher syllable count than the subject, or, in instt. 358,359 (the same sentence in the Α-text as inst. 370 in the B-text), 361, 362, 363, 371, both arguments have the same syllable count, in all these cases, as it happens, two syllables. In all instances of VOS except inst 453, the reverse is the case,

29 the subject has a higher syllable count than the object. It appears that there is a strong tendency to position the argument with the higher syllable count after the one with the lower syllable count This phenomenon would remain an interesting curiosity of this poetry if it were not for the fact that it is a well known principle of word order, observable in a wide range of languages, if, indeed, it is not a universal tendency (it cannot be regarded as a rule). It was first discussed by Behaghel (1909), and then given the status of 'Behaghel's fourth law' of word order, the so-called 'Gesetz der wachsenden Glieder', after that scholar's observations on word order in the fourth volume of his Deutsche Syntax (Behaghel 1932: 6): '[Das Gesetz] besagt, daß von zwei Gliedern, soweit möglich, das kürzere vorausgeht, das längere nachsteht'. This 'law' is presumed to lie behind, or at least contribute to, the order of words in, for instance, such English expressions as 'rough and ready', 'hot and sweaty', 'dead and buried', Tang and country', 'bow and arrow1. The 'law' appears in various guises in various contexts, e.g. in Hawkins' (1983: 89ff.) 'Heaviness Serialisation Principle', which is invoked to explain the relative order of the modifiers, adjective, genitive and relative clause, of nouns in a study of crosslanguage typological aspects of word order. In Isaac (1993b: 22ff.), I invoked a principle 'light constituents first', in an attempt to account for the order of constituents in the Old Irish verbal complex. This principle is, in effect, a minor-image formulation of Behaghel's fourth 'law', a fact which escaped me at the time. Particularly interesting here, in view of the poetic context, is the role of the principle in the metre of the Finnish epic Kalevala (Sadeniemi 1951, cf. Kiparsky 1970: 168), where the tendency is strong to locate longer words nearer the end of the line, shorter words nearer the beginning. It would be unwise to conclude that Behaghel's fourth 'law' was a part of the grammar of early Welsh in general. There is no data to support this conclusion, and it would in any case not be relevant to the present study. The fact remains, however, that the poets quite obviously observed the 'law' in forming the verb-initial sentences of their poetry, ensuring that the syllabically longer argument followed the syllabically shorter one, if the object was longer giving VSO, if the subject was longer giving VOS. The practice has the appearance of an aesthetic preference rather than a grammatical rule, but in an artistic medium such as poetry, we must seriously consider the role such aesthetic factors may have played in the grammatical, here, syntactic structuring of the language. The exception of inst. 453 is unexplained in this analysis, but does not constitute a serious challenge to it, in the face of the consistency with which the principle is followed in all forty-eight other instances examined. It occurs in any case in the marginal and not entirely clear Gwarchan Adebon. If VSO (which must now be taken to include those instances with the always monosyllabic pronominal subjects), and VOS are to be regarded as equivalent verb-initial constructions, subject to the aesthetically motivated constraints of Behaghel's fourth

30 'law', it remains to consider whether either one or the other can be identified as ultimately basic, or whether we must be satisfied with specifying the basic order as Verb-initial' only. The answer is presumably that, in those instances where subject and object have the same syllable count (instt 358, 359 = 370, 361, 362, 363, 371), VSO is chosen. In the absence of any other interfering factor, VSO is the default order for the verb and its arguments. VSO is, therefore, under this analysis, truly basic. The 'interfering factors' giving rise to the subject-initial and object-initial constructions attested in the corpus are considered in the next chapter.

1.4 Greenberg's (21966b) implicational universal 6 says that, 'All languages with dominant VSO order have SVO as an alternative or as the only alternative basic order'. The data from the Book of Aneirin is thus fully consistent in this respect with results from general linguistic, syntactic typology. Before examining the distribution of the SVO constructions from a textual and pragmatic point of view in the next chapter, there is one more set of facts to be considered regarding its grammatical distribution. The following table shows the distribution of various constructions from the point of view of the particles (e.g. negative ny(t), affirmative neu, yf, etc.), and from the point of view of principle vs. subordinate clauses. Verb-initial Principle

-ptl.

+ptl.

VS 30

PVS 27

VSO 18

PVSO 12

vos

PVOS 8

ovs

POVS 12

10

1

Subject-initial

Subordinate

Principle

Subordinate

-ptl.

+ptl.

VS 19

SV 20

PSV 1

VSO 5

SVO 45

SPV 5

χ .·.·.·.

VOS 1

sov

SPOV 1

'.·

7

SV 2 ·.·

•'•'.•'.•l·'.·'.·':

:·:·:· ·:·:·: *:¥

ΐ

Si ±

'••Ά : :χ:

The instances in principle clauses, with particles preceding the verb are denoted PVS, PVSO, etc. Those instances in subordinate clauses are always preceded by conjunctions, e.g. pan, ene etc. Only those orders with explicit subjects are shown (i.e. the orders V, VO, OV are not shown, since the position of the direct object relative to the verb in basic order is not in doubt, see section 1.2). The relative clauses with explicit subject within the clause always have VS, and are therefore also not included Inst 469 gufejm rybud ο gat dydygei gant is the only

31 instance of OVS with non-pronominal object, and thus constitutes an exception to the general pattern of equivalence, [O+pro]VS = VS[O-pro], noted above. It is therefore not included in the table, and is discussed in section 2.7. The results seen so far, regarding pronominalisation and Behaghel's fourth 'law', allow the generalisation of all verb-initial patterns thus, (O)V(O)S(O) = VS, and subject initial patterns thus, S(O)V(O) = SV. Conflating the data of the above table, in terms of the distinction VS vs. SV, the full distribution of particles and conjunctions in these orders is as follows, remembering that in subordinate clauses, there is always a particle or conjunction:

-ptl. +ptl.

VS

SV

59 84

72 9

And the distribution of the orders between principle and subordinate clauses is as follows:

Principle Subordinate

VS

SV

118 25

79 2

There are three obvious tendencies to be noted: i) While the verb-initial orders are well-attested both with and without particles and conjunctions (59% with, 41% without), the subject initial orders seem to have been positively avoided when the clause required a particle or conjunction (11% with, 89% without). ii) Even more emphatically avoided is the use of subject-initial order in subordinate clauses: 98% of instances of subject-initial order are found in principle clauses, against 83% of verb-initial clauses. iii) Correlating with tendency ii), 60% of principle clauses, and 96% of subordinate clauses with explicit subjects have a verb-initial order. Since, whichever parameter is chosen, the verb-initial orders always have a much wider range of distribution amongst the differing syntactic contexts, there is thus further support for the conclusion that these orders reflect, in various guises, the basic order of the language of the poetry. The subject-initial orders are unmistakably an essential part of the syntactic apparatus of the poetry, but they are equally unmistakably subject to

32 stringent constraints as to their use and distribution. These constraints are discussed in the next chapter. In the final two sections of the present chapter, I shall consider the interaction of the different word orders with two aspects of verbal morphology.

1.5 In an enlightening paper, Thomson (1978) discussed the aspectual contrast between the imperfect ('amherffaith') and the preterite ('gorffennol') in early Welsh poetry. Both morphological categories are strictly 'past' with respect to tense, but they differ in aspect, the imperfect coding progressive ('parhaol1) aspect, as well as habitual ('arferol') aspect, while the preterite codes perfective ('perffeithiol1) aspect, that is, indicating the completion of the action/event. The progressive and habitual aspects coded by the imperfect tense (also the present) can be generalised as 'imperfective' aspect, since neither represents the action, event or state from the point of view of its completion, in contradistinction to the perfective, which explicitly represents the action or event in question as completed. Thomson pays particular attention to factors of verbal semantics in the distribution of the tenses, noting (op. cit: 201) that there are 'a hundred and fifteen verbs [in Caw Aneirin] which occur in one of the tenses but not the other1 (my translation from the Welsh). In a similar vein, it may be asked whether there is any correlation between the uses of the tenses in the poetry and the uses of the differing word orders. The following table indicates the distribution of the tenses between the orders in the Book of Aneirin, excluding V, and relative clauses and the category 'Misc.'. No account is taken of mood, so 'prs.' and 'ipf.' are to be understood as including present and imperfect indicative and subjunctive

vs sv vo [O-pro]V

prs.

ipf.

prt.

21 2 11 3

26 12 46 11

21 14 42

fut.

jjrs.hab.

plpf.

1 2 2 3 .·.·.·.·.·.·.·.·.·.·.·.·.· χ*:*:·:*:·:·:·:·:· -.ν.·.·.·. ν.·.·.·.ν. 1 :·:·:·:·:·:·:·:·:·:·:·:·:· ·:·:·:·:·:·:·:·:·:·:·:·:·:

3 3

tA

lU+proJ V

VSO SVO VOS SOV OVS

11

prf.

1U

5 4 4 χΧχΧχΧχΧ:

13

9 25 8 5 1

21 16 7 3

•χ·:·:·:·:·:·:·:·:·:·:·]:·:·:·:·:·:·:·:·:·:·:·:·:· wi MANNER The hierarchy is read thus: an argument with a role further to the left in the hierarchy is more likely to appear as subject of a clause than an argument further to the right (in the cases of DAT/BEN and INST/ASSOC, no decision can be made as to relative topicality). The hierarchy means that arguments with roles to the left are generally, prototypically, more topical in an utterance than those to the right, that is, they are more likely to be the main centre of attention in the utterance, the main topic, appearing normally as syntactic subject Arguments with other roles are also topics, that is, the utterance is also about them, but they are of secondary (or tertiary) importance. At least part of the motivation for the order of this hierarchy lies in the concept of cognitive proximity. An agent is necessarily animate, and usually human (metaphorical and mythical usage apart). Since the users of natural language are (so far) exclusively human beings, it follows that the semantic role of agent is the one which is felt to be closest in character to the pragmatic role of the speaker, who will himself be represented as an agent, if his very act of speaking is in turn represented in an utterance. It is this proximity which is the reason for the position of the role of agent at the top of the hierarchy: it is that role which has the features most similar to those of the speaker. The positions of the other roles follow quite naturally. The dative role, being one of a conscious entity, is prototypically also taken by human arguments, but the non-agentive features of the role mean that it has similarities to the patient role. The latter is the first role in this direction of the hierarchy that can be a feature of an inanimate entity, since it

42 assumes neither volition nor even consciousness on the part of an argument The similarity of the dative role to the patient role puts the former further from the active speaker than the agent is, and the possibility of inanimacy of the patient places this role even further away. A location is perceivable as a presence, but the abstract, relational concepts involved push it cognitively further away from the speaker than the patient The roles lower than this on the hierarchy have a rather abstract nature, and are thus quite distant from the speaker. It is intuitively obvious that the further away, or more different, from the speaker a concept is felt to be, the less likely it is to be a frequent centre of interest in discourse, it is therefore prototypically of less interest than a concept which is perceived as being nearer the speaker. Hence the topic hierarchy. The functioning of the topic hierarchy can be illustrated by some simple examples: ΤΙ Τ2 3. John ate the apple. ACT PAT Tl T2 T3 4. John gave Mary the book. ACT BEN PAT

In both cases, John, as agent appears as subject, i.e. as the most prominent topic, the main topic, Tl, arguments with other roles occupying the less prominent positions of direct and indirect object. Such instances represent the prototypical case. But the topic hierarchy is not a rigid rule for all instances, but rather, precisely, the rule for the prototypical instance. It does not take much thought to realise that there are a wide range of syntactic constructions which are blatant violations of the ordering represented by the topic hierarchy, e.g. passive: Tl T2 T3 5. Mary was given the book by John. BEN PAT AGT Tl T2 6. The book was given to Mary. PAT BEN

The language thus provides means for the same event to be represented from various perspectives regarding the relative importance or interest attached to different arguments. In 5, the agent is demoted to the final position in the sentence, and subordinated to a preposition. Interest is concentrated on the roles of Mary and the book in the event, most particularly on Mary, who thus appears as subject In 6, the relative importance of Mary and the book is reversed, but here, the agent is so unimportant, or perhaps unknown, that it is entirely absent from the linguistic representation of the event Note that

43 semantically, the agent must be present: someone has to give the book to Mary. But the agent argument here is so unimportant that it is not attached to any topic.

This

distinction, argument vs. topic, is of great importance below. The argument belongs to the semantic component of the utterance as a representation of something in the world. The topic belongs to the pragmatic component of the utterance as communication of information, some of which is more, some less important (this concept is elaborated in the context of the syntax of the early Welsh poetry below in sections 2.5ff.). The fact that the passive appears to violate the topic hierarchy does not, however, invalidate it, rather, it confirms its importance. Since the topic hierarchy presents the prototypical instance, and representations in passive constructions present deviations from the prototypical instance, the passive is syntactically a marked (rarer, more complex) construction. This markedness of the passive is a reflex of the prototypical status of the relations characterised by the topic hierarchy. Such syntactic devices as the passive are utilised (amongst many other functions) to maintain the pragmatic high importance of topics in an utterance, where strict adherence to the topic hierarchy would change the main topic, e.g.: Tl T2 T3 ΤΙ Τ2 7. John didn't give Mary any present, but he was still invited to the party. ACT BEN PAT DAT LOG

Throughout the whole sentence, John is the main centre of interest, the other topics serving only as the setting for the narrative concerning him. In the second clause, John has the role of dative, which, prototypically, according to the topic hierarchy, is not so important as the role of agent, which would presumably be Mary here. But Mary is of so little interest (at least for the speaker), that she does not appear in the second clause at all, and John, referred to by the anaphoric he, retains subjects status, by means of the passive construction. Still, there are many cases where the communication involves a change of the primary centre of attention, e.g.: Tl T2 ΤΙ Τ2 8. John kissed Mary and she slapped him. AGT PAT AGT PAT

The representation of Mary actively reacting to John's action brings with it here an automatic switch of primary attention from John as main topic in the first clause to Mary as main topic in the second Indeed, human discourse is characterised as a whole as a complex chaining of units of information, with numerous devices (syntactic, morphological, lexical) for maintaining

44 the prominence of topics, but also with very frequent switches of attention from one topic to another, back again, introduction of new topics, and so on. The need on the pan of the speaker to ensure that the hearer can follow such switches of attention results in the use of various strategies for coding them, with great variety within a single language, and even greater variety between languages. A survey and analysis of even a sample of such devices is far beyond the scope of the present study. Much of Givon (1984, 1990, particularly 739ff., chapter 3) is concerned with elucidating these devices. Here, attention will be concentrated on one such device, topicalisation. Once again, it turns out that this single term conceals a wide variety of phenomena. For the purposes of this study, it will be considered as a means of bringing a particular topic to the special attention, that is, beyond the usual parameters of the topic hierarchy, of the hearer (/reader), by placing it in a special position in a clause, that is, first in the clause. One common form of this strategy is that of left-dislocation, for nominal arguments resulting in the nominativus pendens construction (see Introduction). Examples of the form and use of this construction are to be found scattered throughout Givon (1983b), and it is discussed in detail in Givon (1990: 757ff.). The construction is characterised formally thus (Givon op. cit.: 759): L-dislocation typically involves the following three syntactic characteristics: (a) A separate intonation contour for the dislocated NP (= Noun Phrase); (b) Neutralization of the case-marking of the topic NP; (c) An anaphoric pronoun resumption of the topic NP within the clause. Example la above illustrates the construction for English. A short passage can be used as a further illustration, including the use of the construction in discourse. The passage is taken from the short story An Episode of Cathedral History by M. R. James (James 1992:434): And the Dean he gave a kind of sniff, and walked straight up to the tomb, and took his stand behind it with his back to the screen, and the others they come edging up rather gingerly. Henslow, he stopped on the south side and scratched his chin, he did. The topic the Dean is introduced left-dislocated in the first sentence as main topic, then taken up by the anaphoric pronoun he, which then functions as syntactic subject of the three following verbs ...gave...walked...took.... After the clause containing the last of these verbs, attention is switched to the others, which argument is duly left-dislocated as a new main topic, with resumptive they. And finally, a further individual is singled out for status as new main topic, Henslow, again, left-dislocated. The passage illustrates briefly how the device is used to introduce new main topics. By 'new* here is not, however, necessarily meant concepts which have not been mentioned at all up to that point in the discourse. Usually, there is a store of various

45

topics which play a role in the discourse, and which are alternately treated as main topic, dropped for a while then brought back, etc.: it will be clear that the passage quoted above is taken from a wider context in which the players mentioned appear repeatedly in different roles with differing importance. As Givon (1990: 757) puts it, 'Left dislocation is typically a device to mark topics - most commonly definite - that have been out of the focus of attention for a while, and are being brought back'. The device has therefore to do with topic accessibility, in that it brings (often back) to the fore topics which are not immediately accessible to the knowledge or memory of the hearer, in a given discourse context The device also has to do with topic continuity, in that the topic it establishes as main topic is by definition discontinuous from the point of view of the immediately preceding clauses, since it has not played a role there, but highly continuous from the point of view of succeeding clauses, since, having once been established as a main topic, it can remain so for an indefinite number of clauses, in principle without end, in practice, until a new main topic is established. Having introduced the basic theoretical ideas adopted in this analysis, and having presented, as an example of topicalisation the construction of left-dislocation, a term which includes the nominativus pendens construction mentioned in the introduction to this volume, I now turn, in the following sections, to consideration of the application of these ideas to early Welsh, first, by way of illustration, in Middle Welsh prose, then later, in the early poetry which is the subject of this study. The digression into the problems of the Welsh 'abnormal' sentence, may seem at first non sequitur from the point of view of the preceding discussion, but I hope its relevance will became clear in later discussion. I present the 'abnormal' sentence as an example of a topicalisation strategy, at this stage to be regarded as formally distinct from the strategy of leftdislocation.

2.2 In what follows, I shall not attempt an exhaustive discussion of the 'abnormal' sentence in all its formal and functional variety. For instance, the role of the construction VN + a + GWNEUTHUR will not be considered. The example of the 'abnormal' sentence in Middle Welsh prose is only used i) as an illustration of the sort of effects pragmatic factors of information structure have on word order in early Welsh, and ii) as a basis for consideration of the formal and functional status of various types of word-order variation in the early Welsh poetry which is the actual subject of the present study. In the following selections, a simplified glossing is adopted, in comparison with that used in the list of instances from the Book of Aneirin which follows chapter 3. Further, only certain key sentences are given interlinear glosses: for sentences with less importance for the argumentation, only the translation is given (from Jones/Jones 1974).

46

Main topics are indicated by MT; 'ABNORMAL SENTENCES' are printed in capitals; subscript letters, j, j, etc. indicate coreference; ptl. = particle. I) Breudwyt Ronabwy (Richards 1948: 1-2): MT1 1. MADAWC UAB MAREDUDj A OED idawi Powys yn y theruyneu. M. son M. ptl. was to-him P. in its boundaries 2. Sefyw hynny, o Porford hyt yg Gwauan yg gwarthaf Arwystli. it is that from P. till in G. in top A. new topic 3. Ac yn yr amser hwnnw BRAWTj A OED idawi, nyt oed kyuurd gwr ac eft. and in the time that brother ptl. was to-him not was equal man with him 4. Sefoed hwnnwj, lorwoerth uab Maredudj. it was that I. son M. 2 MT 5. A HWNNWj A GYMERTHgoueileintmawryndawj a thristwcho welet and that ptl. took care great in-him and sadness from seeing yr enryded a' r medyant a oed y\ vrawt ac ynteuj heb dim. the honour and the possession ptl. was to-his brother and he without thing 6. Ac ymgeissaw a oruc a'e\ gedymdeithon a'ej vrodoryon maeth, ac ymgyghor ac wynt beth a wnelei am hynny. 7. Sefa gawssant yn eu kyghor, ellwng rei onadunt y erchi gossymdeith idawj. 8. Sef y kynnigywys Madawc\ idawj, y pennteuluaeth a chystal ac idaw\ ehun, a meirch ac arueu ac enryded. 9. A gwrthot hynny a oruc lorwoerthj, a mynet ar herw hyt yn Lloeger, 10. a Had kalaned a llosgi tei a dala karcharoryon a oruc lorwoerth j. 11. A chyghor a gymerth Madawci a gwyr Powys ygyt ac eft. 12. Sefy kawssant yn eu kyghor, gossot kanwr ym pop tri chymwt ym Powys o'ej geissaw. 13. A chystal y gwneynt rychtir Powys, o Aber Ceirawc yn Hallictwn vet yn Ryt Wilure ar Efyrnwy a'r tri chymwt goreu oed ym Powys. 14. Ac ny vydei da idaw ar teulu ym Powys ar ny bei da idaw yn y rychtir hwnnw. 15. A hyt yn Nillystwn Trefan yn y rychtir hwnnw yd ymrannassant y gwyr hynny. MT3 16. A GWRfcA OEDar y keis hwnnw, and man ptl. was on the quest that 17. sefoed y^ enw, Ronabwy^. 18. Ac y doeth Ronabwyi a Chynnwric Vrychgoch, gwr ο Vawdwy, a Chadwgawn Vras, gwr ο Voelure yg Kynlleith, y ty Heilyn Goch uab Kadwgawn uab Idon, yn ran. Jones/Jones (op. cit.: 137): 1. Madawg son of Maredudd held Powys from end to end, 2. that is, from Porffbrdd unto Gwafan in the uplands of Arwystli.

47 3. And at that time he had a brother[, who] was not a man of equal rank with himself: 4. he was lorwoerth son of Maredudd. 5. And he felt great heaviness and sorrow at seeing the honour and power that were his brother's, whereas he had naught. 6. And he sought out his comrades and foster-brothers and took counsel of them what he should do about it 7. They decided by their counsel to send some from amongst them to demand provision for him. 8. The offer Madawg made him was the captaincy of his war-band, and equal standing with himself, and steeds and arms and honour. 9. And lorwoerth rejected that and went harrying into Lloegyr. 10. And lorwoerth made slaughter and burned houses and carried off prisoners. 11. And Madawg took counsel, and the men of Powys with him. 12. They decided by their counsel to place a hundred men in every three commots in Powys to seek him out. 13. And they reckoned Rhychdir Powys, from Aber Ceriawg in Hallictwn as far as the ford of Wilfre on Efyrnwy, as equal to the best three commots that were in Powys. 14. And the man would not prosper with a war-band in Powys who would not prosper in that cultivated land.

15. And as far as Didlystwn [Trefan] in that cultivated land, those men took their quarters. 16. And there was a man on that quest, 17. his name was Rhonabwy. 18. And Rhonabwy and Cynwrig Frychgoch, a man from Mawddwy, and Cadwgawn Fras, a man from Moelfre in Cynlleith, came to the house of Heilyn Goch son of Cadwgawn son of Iddon for lodgings.

Notes to I) 1. Madawc nab Maredud: first main topic (MT) of discourse, fronted in 'abnormal' sentence. 3. Brawt 'a brother' is a new topic, introduced by the 'abnormal' order, but is too nonspecific to displace Madawc asMT. 4. The referent of brawt is specified, lorwoerth uab Maredud. 5. The masculine singular demonstrative hwnnw refers to lorwoerth, is definite and in 'abnormal' order, establishing the referent lorwoerth as MT, displacing Madawc. 6-15. Various new topics are introduced, "his comrades and his foster-brothers' (6), 'the men of Powys' (11), and Madawc, the first MT reappears (8, 11), but no topic appears in the 'abnormal1 order, so no topic displaces lorwoerth as MT of the paragraph. The main topic status of lorwoerth is seen most clearly in that anaphoric pronouns may appear which refer to him without his having been named in the same

48 clause, 8, 12. Such pronouns can only refer to Madawc, however, when he has been named immediately before, 8, 11. (contrast 3, with Madawc as unnamed MT, and as referent to anaphoric pronouns). 8. Sefy kynnigywys Madawc idaw. Madawc is recalled, but in basic VS order, so does not oust lorwoerth as MT; idaw refers to lorwoerth who has not be mentioned by name since four sentences previously. 16. A new topic, and new MT is introduced in 'abnormal' order, gwr, named immediately after (17) as Ronabwy. 18. Many new topics are introduced, some even as syntactic subjects, but Ronabwy remains as MT, so no 'abnormal' order, but unmarked VS instead.

II) Historia Peredur vab Efrawc (Goetinck 1976: 42-44): 1 . ARTHUR A OED yg Kaer Llion ar Wysc A. ptl. was in K. LI. on W. new topic 2. a mynet a wnaeth y hela a Peredur^ gyt and go pU. he-did to hunt and P. together

ac ef with him

MT2 3 . Λ PHEREDURj A ELLYGAWD y gi ar hyd, and P. ptl. let-loose his dog on stag

(SMT) 4. A "R KI A LADAWD yr hyd mywn diffeithwch. and the dog ptl. killed the stag in wilderness still MT2 5. Ac ym pen ruthur y wrthaw eft a welei arwyd kyfanhed and in end rush from by-him he ptl. saw sign dwelling

6. a thu a" r kyfanhed y death. Ί. Ac efi a welei neuad, 8. ac ar drws y neuad eft a welei tri gweis moelgethinyon yn gware gwydbwyll. 9. A phan death y mywn efi a welei teir morwyn yn eisted ar leithic, ac eurwiscoed ymdanunt mal y dylyei am dylyedogyon. 10. Ac efi a aeth y eisted attunt y'r lleithic, MT3 11. ac VN O'R MORYNYONj a edrychawd ar Peredur yn graff, and one of the maidens ptl. looked on P. ptl. keen

12. ac wylaw a wnaeth. and weep ptl. she-did MT2 13. APHEREDURjA OFYNNAWD idij beth a wylei. and P. ptl. asked to-her what ptl. she-wept

[Dialogue] M: 'Roc dryccet genhyfgwelet lleassu gwas kyn deccet a thi. '

49 P: 'Pwy a'm lleassei i?' M: 'Pei na bei pyt it arhos yn y lie hwn, mi a'e dywedwn it.' P: 'Yrmeint uoygwrthretamafyn arhos, mi a'egwarandawaf.' M: 'Ygwr yssyd tat inni bieu y llys hon, a hwnnw a lad pawb o'r a del y'r llys hon heb y ganhat.' P: 'Py gyfryw wr yw awch tat chwi, pan allo lleassu pawb uelly?' M: 'Gwr a wna treis ac anuod ar y gymodogyon, ac ny wna iawn y neb ymdanaw.' still MT2! 14. Ac yna y gweleiefiy gweissonyn kyfodi ac yn arllwyssawy clawr and then ptl. saw he the youths asp. rise and asp. clear the board o 'r wenn. of the pieces 15. Ac efi a glywei twrwfmawr, and he ptl. heard noise great new topic le.acynol y nvrwfefiO welei[wr du mawrvnllygeityawc\kyn dyfot and in track the noise he ptl. saw man black big one-eyed asp. come y mywn. in MT4 17. A 'R MORYNYON A GYFODASSANT yn y*. erbyn, and the maidens ptl. rose at-him 18. α diot y^ wise y amdanaw^ a wnaethant and remove his garb from around-him ptl. they-did MT5 19. ac YNTEUkA AETHv eisted. and he ptl. went to sit 20. A gwedy dyfot y^ bwyll idaw^ ac araftiau, edrych a oruc ar Peredun, a gofyn pwy y marchawc. [Dialogue] M: 'Arglwyd,' heb yr hitheu, 'y gwas teccafa bonhedigeidafo'r a weleist eiroet, ac yr Duw, ac yr dy syberwyt, pwylla wrthaw.' BM: 'Yrot ti mi a bwyllaf, ac a rodafy eneit idaw heno.' MT2 21. Ac yna PEREDURi A OOETHattunt wrthy tan, and then P. ptl. came to-them by the fire 22. ac a gymerth bwyt a llyn, and ptl. took food and drink 23. ac ymdidan a 'r rianeda oruc. and converse with the ladies ptl. he-did 24. Ac yna y dywawt Peredu^ gwedy yi urwyscaw... and then ptl. said P. after his get-drunk

50 Jones/Jones (op. cit.: 207-8): 1. Arthur was at Caer Llion on Usk, 2. and he went to hunt, and Peredur along with him. 3. And Peredur loosed his dog upon a stag, 4. and the dog killed the stag in a wilderness. 5. And some way off from him he could see signs of habitation, 6. and he came towards the habitation. 7. And he could see a hall, 8. and at the hall door he could see three bald-headed swarthy youths playing gwyddbwyll. 9. And when he came inside he could see three maidens seated on a couch, and royal apparel about them as was meet for folk of noble birth. 10. And he went to sit with them on the couch, 11. and one of the maidens looked closely at Peredur, 12. and she wept. 13. And Peredur asked her why she was weeping. M: 'Because it is so grievous to me to see slain a youth as fair as thou.' P: Who would slay me?1 M: "Were it not perilous for thee to stay in this place, I would tell thee.' P: "However great my danger be, staying, I will hear it.' M: 'He who is our father owns this court, and he slays every one that comes to this court without his leave.1 P: 'What kind of man is your father that he can so slay every one?' Μ: Ά man who does treachery and malice to his neighbours, and he makes no redress to any for it.1

14. And then he could see the young men arising and clearing the board of the pieces. 15. And he could hear a great clatter, 16. and after the clatter he could see a big black one-eyed man coming in. 17. And the maidens arose to meet him, 18 and they drew off his garb from about him. 19. And he went to sit down. 20. And after he had recovered himself and was at his ease, he looked upon Peredur and asked [who the knight was]. M: 'Lord,' she replied, 'the fairest and noblest young man thou hast ever seen, and for

God's sake, and the sake of thine own pride, deal gently with him.' BM: 'For thy sake I will deal gently with him, and will grant him his life for to-night.1 21. And then Peredur came to them near the fire, 22. and took meat and drink,

51 23. and conversed with the maidens. 24. And then Peredur, having grown tipsy, said to the black man ... Notes to II) 1. The first MT of the discourse, Arthur, is introduced by means of the 'abnormal' order. Arthur in fact, after 1-2, plays no further role in the narrative, and these sentences have a scene-setting function. 2. A new topic, Peredur, is introduced. 3. Peredur is established as MT by the 'abnormal1 order. 4. The narrative continues for one sentence centered on the two minor topics of 2, the dog and the stag, with the dog as MT of the sentence. However, 5. when the narrative returns to Peredur, he is still overall MT and does not need to be explicitly re-established as such (i.e. named in 'abnormal' order). Cp. note 13. 5-10. This is a chain of narrative containing sentences with efa welei (x4)... 'he could see ..." and efa aeth... 'he went..." as their cores. While the order is 'abnormal' , its use with the anaphoric pronoun, itself not a topic, but rather referring to an already established topic, does not introduce a MT, but rather maintains one as such in the face of a veritable barrage of new topics in a descriptive passage, 'habitation' (5), 'hall' (7), 'three bald-headed swarthy youths' (8) etc., etc. Cp. note 19. 11. A new MT is introduced and remains for two clauses. 13. Peredur is re-established as MT in the 'abnormal' order. The question arises, why must Peredur be explicitly re-established as MT in the abnormal order in this sentence, whereas in 5 above, that did not have to be done, even though Peredur had been absent for one sentence, 'and the dog killed the stag in a wilderness' (4)? The answer is derivable from the topic hierarchy. In 4, the fact that Peredur is linguistically absent from the event narrated means that his status as MT of the whole paragraph is not interfered with: the topicality of 'the dog1 extends over only that sentence, and once it is dropped from the narrative, the overall MT of the paragraph remains unaffected. The dog' is a subordinated MT (SMT), salient only in parenthesis, in a single sentence, and not over an extended passage. Furthermore, the 'dog', as a nonhuman participant in the action, is prototypically less salient than a human participant: its presence as most salient argument within a short clause is not sufficient to replace the human Peredur as MT over an extended passage. However, sentence 13 is preceded by 11-12 "and one of the maidens looked closely at Peredur, and she wept' in which Peredur is indeed a participant, but in a role of low relative topicality, as patient, governed by the preposition a; with the human 'maidens', in agentive role as most salient argument: in 11, Peredur has been demoted on the scale of topicality. In order to re-establish Peredur as MT, to promote him back to that status, over an extended passage (not just a single sentence), it is not enough to have him back in the

52 subject position, but the subject role must itself be promoted to sentence-initial position, producing 'abnormal' order. Dialogue: between Peredur (P) and a 'maiden' (M). 14. The narrative is resumed, with no explicit MT marking, but rather a basic VSO sentence y gwelei efy gweisson 'he saw the youths' preceded by the temporal adverb yna 'then'. The subject pronoun ef refers to Peredur, last named in 'abnormal' order (13) before the dialogue, in which he participated, and by virtue of that, still MT. High topicality is maintained despite the interruption of the narrative by the dialogue, and need not be explicitly restated in 'abnormal' order. Contrast note 21. 16. A new topic is introduced wr du mawr vnllygeityawc 'a big black one-eyed man', as patient 17-18. The maidens' temporarily take over MT status as Peredur is removed from the narrative. 19. An anaphoric pronoun, in the so-called 'conjunctive' form ynteu, coreferent with the 'black man' appears in 'abnormal1 order, and establishes him as MT. The 'conjunctive' form of the pronoun establishes the reference of the pronoun in hightopicality position to a previous constituent in low-topicality position, here, reference of a subject pronoun in 'abnormal' order to a previous patient (16), possessor (y wise 18), locative (amdanaw 18). 20. Peredur is reintroduced into the narrative, as patient. Dialogue: between the maiden (M) and the black man (BM). 21. In contrast to 14, where Peredur was maintained as MT from before the preceding dialogue, in 21, he is re-established as MT: before this dialogue he was a lowtopicality patient, and the black man was MT. Note that the presence or absence of 'abnormal' order neither depends on nor conditions the presence of sentence initial adverbs: basic VS(O) Ac yna y gwelei efy gweisson (14), Ac yna y dywawt Peredur ... (24) vs. 'abnormal' S a V Ac yna Peredur a doeth attunt (21). Thus these distributions are not conditioned by formal properties of sentence structure, but purely by the pragmatic demands of information structure. 24. Peredur is still MT, so the sentence Ac yna y dywawt Peredur... has basic VS order after the adverb.

2.3 Ideally, it will be seen that the analysis of word-order variation in Middle Welsh as presented so far is not only plausible, but also supportable by textual analysis. However, an important confirmation of the plausibility of the argumentation would be the demonstration that the pragmatic factors involved have relevance and influence in another subsystem of the grammar of (Middle) Welsh, independent of the subsystem of

53 constituent order. This is indeed the case, in the system of so-called 'conjunctive pronouns' (Middle Welsh forms, GMW 49). The designation 'conjunctive' derives from the observation that they tend to occur after conjunctions, e.g. Τι a minnau 'You and me', etc. This impression however conceals what may be regarded as their primary function, in Middle Welsh at least, that of coding a switch in the salience or topicality of an argument. In the example Τι a minnau 'You and me', this switch occurs since the argument 'Isg.' is completely absent (i.e. with zero salience) in the expression ft' '2sg.': there is thus a direct switch from 0% topicality of'Isg.' to 100% topicality, and the switch is coded by the 'conjunctive1 form (cf. in general MacCana 1990a). I refrain from speculating on the formal etymology of this series of pronouns here: I note Russell's (1982) paper to this effect. In more complex texts, this phenomenon has, as to be expected, rather more complex features. Still a relatively simple Middle Welsh example is the following passage from Culhwch ac Ohven, 11. 1061-64 (Bromwich/Evans 1988 = 1992: 37): simplified glossing is used as above; SW = switch: 1. A gwedy disgynnu A rthur y 'r tir, and after descend A. to the land 'And after Arthur had landed, 2. dyuot seint Iwerddon attaw y erchi nawd idaw. come saints Ireland to-himtoask protection to-him 'the saints of Ireland came to him to ask his protection. 3. Ac y rodes ynteu nawd udunt hwy, and ptl. gave he.SW protection to-them they 'And he gave them his protection, 4. ac y rodassant wynteu eu bendith idaw ef. and ptl. gave they.SW their blessing to-him he 'and they gave him their blessing.' In sentences 1 and 2, the argument Arthur functions as main topic, introduced as such this time not by means of an 'abnormal' construction, but simply as subject, under genitival government by the verbal noun disgynnu. Its status as main topic is confirmed by the fact that in sentence 2 (actually the main clause to which the temporal clause 1 is subordinate), Arthur is referred to by means of anaphora (cf. the note to sentences 6-15 in the passage from Breudwyt Ronabwy quoted above). In sentence 3, anaphora is still used, but here, the pronoun, subject of the verb rodes, appears in the 'conjunctive' form ynteu. The reason is that the last pronominal form referring to Arthur was in the prepositional form idaw 'to him'. An argument appearing as complement of a preposition has a lower prominence in the discourse than the full subject of a conjugated verb. The same phenomenon was observed in the note to sentence 13 in the passage from Peredur. Arthur, in the form of the pronouns referring anaphorically to him, has thus been

54 promoted from a role of relatively low prominence (complement of preposition) to one of very high prominence (subject of finite verb). It is this promotion in prominence that is coded by the 'conjunctive' form of the pronoun. In sentence 4, precisely the same phenomenon is observable, but this time with reference to the seint hverddon 'the saints of Ireland' anaphorically referred to by a 3pl. pronoun. At the end of sentence 3, they appear in low-prominence role in the prepositional form udunf. in sentence 4 they are promoted to subject of the finite verb, and thus appear in their turn in the 'conjunctive' form. In this instance, the two arguments involved in the passage of discourse differ in number, one singular, the other plural. There could not therefore be any ambiguity in the identification of which pronoun referred to which topic, even if the 'conjunctive' pronouns were not used. This shows that the use is such pronouns is not simply a matter of establishing patterns of coreferentiality, but genuinely has the function of coding switches in relative prominence of topics in discourse. Cases can certainly arise where the use of these pronouns has a disambiguating effect by establishing the correct relations of referentiality. The following example is taken from Peredur again (Goetinck op. cit.: 28.28-29): 1. Ymwan eissoes a orugant, fight nevertheless ptl. did 'Nevertheless they fought, 2. a Pheredur a uyrywys y marchawc. and P. ptl. threw the knight 'and Peredur threw the knight. 3. Nawd a erchis ynteu. protection ptl. asked he.SW 'He asked for quarter.' The 'conjunctive' pronoun (not following any conjunction!) ynteu in sentence 3 unambiguously refers to 'the knight1, y marchawc, of the preceding sentence, since there the knight has the syntactic role of object, but in sentence 3 is promoted to subject The promotion is coded by the 'conjunctive' form. If the sentence had the form Nawd a erchis ef, the pronoun ef would have to be interpreted grammatically as referring to Peredur. note the ambiguity of the, after all rather unidiomatic, English translation. On the basis of the two passages so far examined in this section, it might be concluded that the function of the 'conjunctive' pronouns could be restricted to that of coding promotion of a preceding non-subjective argument to the syntactic role of subject The following passage from Culhwch ac Olwen, 11. 1193-96 (Bromwich/Evans 1988 = 1992: 41) demonstrates the fallacy of this conclusion:

55 1. Kacamwri, ual y tynnit efy uynyd, K. as ptl. pulled he to mountain 'As for Cacamwri, as he was being pulled up, 2. y tynnei deu uaen ureuan ynteu y V affwys. ptl. pulled two stone quern he.SW to the abyss 'two quem stones pulled him down into the abyss. 3. Oslo Gyllelluawr, yn redec yn ol y twrch, O. big-knife asp. run yn track the boar 'As for Osla Big-Knife, running after the boar, 4. _y dygwydwys y gyllell o 'e wein, ptl. fell his knife from his sheath 'his knife fell from his sheath, 5. ac y kolles; and ptl.-him lost 'and he lost it; 6. a 'e wein ynteu gwedy hynny yn llawn ο V dwfyr, and his sheath he.SW after that ptl. full from the water 'and since his sheath was after that full of water, 7. ual y tynnit ef y uynyd as ptl. pulled he to mountain 'as he was being pulled up 8. y tynnei hitheu ef y'r qffwys. ptl. pulled she. S W he to the abyss 'it pulled him down into the abyss.1 This complex passage consists of two periods, each beginning with their respective topicalised arguments, Kacamwri and Osla Gyllelluawr, as main topics, in a nominativus pendens construction. 'Conjunctive' pronominal forms occur in sentence 2, referring to Kacamwri, 6, referring to Osla, and 8, referring to y wein 'his sheath'. The comparison between the two parallel clauses as follows is instructive in the use of the 'conjunctive' pronouns: MT Kacamwri Osla...

Temporal clause V S O LOC ual y tynnit efy uynyd, y tynnei deu uaen ureuan ynteu y'r affwys. ual y tynnit efy uynyd, y tynnei hitheu ef y'r qffwys.

The difference in the two clauses is that in Kacamwri, ual y tynnit efy uynyd, y tynnei deu uaen ureuan ynteu y'r qffwys, the argument deu uaen ureuan, subject of tynnei, is a brand-new topic in the prominent subject role, so that the object pronoun referring to Kacamwri takes the 'conjunctive' form to maintain its high topicality in the face of new competition. In Osla... ual y tynnit efy uynyd, y tynnei hitheu efy'r affwys

however,

the subject of the tynnei clause is not a new topic, but an old one (y wein) which has been promoted from a low-topicality role (as patient in sentence 6 a'e wein ynteu gwedy hynny yn llawn o'r dwfyr) to a high-topicality role (subject); it assumes the 'conjunctive' form to establish referentiality across the promotion (promotion to subject role as in the

56 examples from Culhwch ac Olwen and Peredur above). But since it is only a promoted, previously low-status topic, and does not appear in the abnormal order, it does not compete in prominence with the object pronoun e/referring to Osla, still the main topic of the passage, which thus remains in the independent form. A 'conjunctive' pronoun has, however, already appeared referring to Osla, that is ynteu in sentence 6. Here, as in sentence 2 (object ynteu referring to Kacamwri) there is apparently a demotion. Immediately preceding sentence 6, Osla was (anaphorically) subject of the verb kolles. In a'e wein ynteu in sentence 6, Osla is referred to anaphorically by a pronoun in a prototypically low-salience genitive role: Osla, however, remains the main topic of the period, the pronoun codes the reference to the main topic in the 'conjunctive' form. The instances of the 'conjunctive' pronouns in this passage illustrate the two types of topicality promotion which can be coded by that means: i) Syntagmatic: a topic is promoted from a previous low-topicality role to a hightopicality role (e.g. syntactic subject). Reference of the pronoun to a previous lowsalience topic is established (sentence 8, and previous instances above). ii) Paradigmatic: a topic is promoted to high topicality in a prototypically low-topicality role (e.g. syntactic object, genitive or, not illustrated here, complement of preposition). Reference of the pronoun to a previous high-salience topic is established (sentences 2, 6). It is apparent then that rather than coding a simple promotion as such, what the 'conjunctive' pronouns code is a switch in topicality. Such switching in the relative topicality of arguments as is coded lexically by the 'conjunctive' pronouns is precisely what, it has been suggested above, is coded by another subsystem of the grammar, that of constituent order and its pragmatically conditioned variants. This constitutes a demonstration that the grammar of Middle Welsh as a whole was sensitive to such factors, and that the analysis of the word-order variation presented for certain passages above is indeed plausible. For completeness, it may be pointed out that the scopes of the two subsystems of pragmatic coding discussed here, 'conjunctive' pronouns and constituent order, differ by degree. The type of phenomenon coded by the 'conjunctive' pronouns operates with reference only to the last clause: whether the 'conjunctive' forms appear or not in a clause depends solely on the nature of the role of the topic in the immediately preceding clause (including whether it had a role at all there or not). The subsystem of word-order variation operates over a much wider range, at a macro-textual level, coding and regulating relative topicality over extended stretches of the discourse, forming topical paragraphs. This difference of degree between the two subsystems allows them to complement each other in the information structure of the texts.

57 2.4 Whatever its etymology, the particle o, OW ha(i), is, without a doubt, in origin, a relative marker. This is its clearly attested function in the Old Welsh sources, e.g. Chad 2 tir telih. haioid Hau elcufilius gelhig 'The land of Telych which was in the hand(s) of Elgw son of Gellhy', Computus Fragment ir serenn, hoi bu in arcimeir .o. The star [= constellation] which was opposite o\ and in a cleft sentence, is did citnan. had 'It is a full day that you will get' (interpretation of I. Williams 1927: 260). The relative marker a precedes an affirmative relative clause modifying a nominal

expression which has a structural-case role in that clause, subject or direct object. It therefore serves both to link the relative clause with its antecedent and to indicate the nature of that relationship in terms of syntactic roles. Historically then, the relative function of a is primary. But by the Middle Welsh period, it is open to debate whether it is meaningful or useful any more to regard α as a relative particle, which secondarily assumes different functions in various syntactic contexts. A generalisation about the function of the particle is being missed like this (analyses which treat α as a relative pronoun are even wider of the mark). Even less meaningful would be to attempt to attribute varying function to the relative clause itself. It seems unlikely that helpful results can be expected if a canonical 'abnormal' sentence like Gwawl a gerdawd ryngthaw ay gyuoeth 'Gwawl started for his realm' (PKM 15) were to be treated as if it contained a relative clause functioning as a main clause: there is something wrong with the terminology here (cf. 'y drefn berthynol', Watkins 1977/78, 'cleft order', Watkins 1991). Whatever the origin of the particle a, we clearly need a definition of its function in Middle Welsh which captures is use in relative clauses, in focus constructions and in the topicalising 'abnormal' sentence'. I formulate such a definition operationally as follows: A declarative clause is preceded by the particle a when the case role of the subject or direct object of the clause is a feature of an argument preceding the clause, and there is no other particle (negative ny(t), aspectual ry) in this position. This definition generalises the function of α seen in relative clauses, e.g. y gwr mwyaf a gery The man whom you love most' (PKM 12); in focus constructions, e.g. Mi a'e heirch 'I'm the one who asks for her' (Culhwch ac Obven 1. 562, Bromwich/Evans 1988 = 1992: 21); and in the topicalising 'abnormal' sentence, e.g. Ac Arthur a wisgawd ymdanaw 'And Arthur dressed himself (Owein 1. 513, Thomson 1975: 19). These examples do not illustrate three differing functions of a, or three functions of relative clauses, or the 'relative order1. There is just one function, to indicate that the case role of subject or direct object in the clause following α is a syntactic feature of the preceding constituent (cf. the account, in somewhat different terms, in Fife/King 1991: 134, giving the phenomenon a universalist background). The appearance of the particle in the 'abnormal' sentence is not, as Henry Lewis would have it (see the discussion in the

58 Introduction), a matter of the insertion of a meaningless particle into the construction by analogy with other constructions, but constitutes a generalisation of the (very concrete) function of the particle itself. If the particle a has been inserted into the topicalising 'abnormal' sentence by generalisation of its function from other constructions, it is meaningful to seek the structural predecessors of the 'abnormal' sentence in some sort of construction in which the particle did not originally appear. Certain considerations regarding the distribution of the 'abnormal' sentence in Middle Welsh prose give clear indications as to where these structural predecessors are to be found. Proinsias MacCana (1973) was the first to discuss the possible role of the nominativus pendens construction in mediaeval Welsh (and Irish) texts (for Irish, this discussion was picked up and developed in detail by MacCoisdealbha 1976: for discussion of such historiographical matters, see the Introduction above). It is informative to present MacCana's (1973:96) examples of the construction from Middle Welsh prose: 1) Ynteu Wydyon goreu kyuarwyd yn y byt oed. 'Gwydion was the best teller of tales in the world.' PKM 69. 2) Ynteu Wydyon kyuarwyd da oed. "Now Gwydion was a good teller of tales.' PKM 82. 3) a hynny yn Llys Eueyd Hen y gwnaeth o aghynghor. 'and he wrought that at the court of Hefeydd the Old, illadvisedly.' PKM 64. 4) ac ual y gyt ac y doethant hyt y Uelen Ryt y pedyt ny ellit eu reoli o ymsaethu. 'but as soon as they reached Y Felenrhyd, the men on foot could not be restrained from shooting at each other.' PKM 72. 5) Ansawd Pwyll hyspys oed gantaw ef. 'Pwyll's appearance was well known to him.1 PKM 24. 6) Ac wrth hynny er rey ereyll en kyflavn ofyn adav e dynas a orugant. 'And therefore the others, full of fear, left the city.' Brut y Brenhinedd (B. F. Roberts ed.), l. 712. It is to be observed that all of these examples contain not only the nominativus pendens construction, but also various other complex grammatical devices. Examples 1, 2 and 5 contain nominal and adjectival predicates, fronted under focus before a form of the copula; 3 contains an adverbial preposition phrase fronted under focus; in 4, the clause following the nominativus pendens, y pedyt, consists of a form of the modal auxiliary gallu with the verbal noun reoü as complement, the case role of y pedyt in the clause, coded by the possessive pronoun eu, being dependent on the verbal noun; and in 6, the clause in which er rey ereyll would have the subject role has the periphrastic construction adav e dynas a orugant, VN + a + GWNEUTHUR.

59 It would be reasonable to ask whether examples of the nominatlvus pendens construction might not be found in less complex clauses. Taking the instances with focussed predicate of the copula, e.g. Yrueu Wydyon kyuarwyd da oed, the conditions for focus arise out of the context, where it is desirable to place contrastive emphasis on the predicate kyuarwyd da 'a good storyteller', as justification for part of Gwydion's complex plan to bring Gilfaethwy to bed with Goewin. The resumptive pronoun, characteristic of the construction, to indicate the nature of the case role of the fronted constituent in the clause, does not have to appear explicitly when that constituent has the subject role, but is identified by the verbal flexion (MacCana 1991: 50; contrast the passage quoted from M. R. James in section 2.1, where the resumptive subject pronouns were always explicit). It can be wondered then what the construction what look like if the conditions for focus were not present. My contention is that this would be precisely an 'abnormal' sentence, e.g. Pwyll, Pendeuic Dyuet, a oed yn arglwyd or seith cantref Dyuet 'Pwyll, Chief of Dyfed, was lord over the seven hundreds of Dyfed1 PKM 1, or, with omitted or elided particle, Math uab Mathonwy oed arglwyd ar Wyned 'Math son of Mathonwy was lord over Gwynedd' PKM 67. A few lines lower on the page than this instance occurs the sentence, Ac yn yr oes honno Math uab Mathonwy ny bydei uyw, namyn tra uei y deutroet ymlyc croth monvyn ... 'And in that time, Math son of Mathonwy would not be alive, except while his two feet were in the lap of a maiden ...'. Is this an 'abnormal' sentence, or the nominativus pendens construction? Formally, there is no distinction here. The comparison of the following sentences from Owein is interesting: A'r iarll pieu y goer honno goreu gwr am vwyt yw 'And the earl who owns that fortress is the best man for food' (Thomson 1975:11. 707-8; = nominativus pendens); Ygwrda bieu y goer yssyd uarwyrawrhon "The nobleman who owns the fortress has died this hour' (op. ciL: 11. 337-38; = 'abnormal' sentence). Here again, the only structural, syntactic difference between the constructions is that the former contains a fronted, focussed predicate, the latter does not In this case, with present tense forms of the copula, the complex [a + y w] is realised by the 'relative' form yssyd: historically, this form was undoubtedly originally purely relative (< *esti-yo), but by the Middle Welsh period, precisely the same considerations apply to its function as discussed in detail above for the particle a. Another instructive comparison is to be seen when the constituent fronted has the role of direct object in the subsequent clause, e.g. A'r ulwydyn honno a dreulwys pawb ο honunt hyt oet y wled yn llys Eueyd Hen 'And every one of them spent that year until the time of the feast in the court of Hefeydd the Old' PKM 15-16; A gwr du mawr a wely ym penn yr orssed... 'And you will see a great, black man on top of the mound ..." Owein, Thomson (1975:1. 108). It appears that such (relatively rare) instances give the lie to the assertion that the 'abnormal' sentence is identical to the nominativus pendens construction, since they contain no resumptive (object) pronoun. But when instances are

60 found with fronted object and negation (an even rarer occurrence), an interesting thing happens, e.g. a chynhebfygfrwyd

y'r twryfhwnnw nys ryglywssei eiryoet 'and the like

of that noise he had never before heard' Breudwyt Ronabwy, Richards (1948: 3.30-4.1); Amsathyr dynyon nac alafoed nys gwelei yn y coet... 'He could see the tracks of neither man nor herds in the forest ..." Peredur, Goetinck (1976: 22.24-25). Here, there is a resumptive object pronoun, -s, producing canonical examples of the nominativus pendens construction. I conclude that it is a feature of the construction that for object topicalisations in affirmative clauses, zero anaphora is licensed as a resumptive element, just as in the cases of subject topicalisation, except that in the latter, the verbal flexion identifies the subject: the morphological characteristic of Welsh verbal flexion that the subject can be identified by agreement but not the object appears irrelevant for the syntactic structuring of the clause in the affirmative. That the explicit resumptive pronoun appears in negative clauses is attributable to the fact that an object in a negative clause is necessarily less affected by the event, action encoded than in an affirmative clause, and is therefore less central to the situation represented. The resumptive pronoun in the negated construction fixes the role of the constituent in question, whereas this is unnecessary in the affirmative clause, where the more strongly affected object has a more central, and thus clearer role. The relationship of zero anaphora and the explicit resumptive is further discussed at the end of section 2.8 below. In keeping with the conventional view, Poppe (1991a: 196-97) makes a distinction between the 'abnormal' sentence and the nominativus pendens construction, subsumed under the wider term of 'left-dislocation'. He assigns functional motivation to this distinction in the following way (loc. cit): There is clearly a functional difference between left-dislocated elements, which establish a separate information unit, and topics [in the 'abnormal' sentence], which establish the starting point for the statement.' Poppe gives one example of left-dislocation (for nominal constituents = nominativus pendens), the only one to be found in Cyfranc Lludd a Llefelys, the text which is the subject of the paper in question (= Roberts 1975:11. 164-63): Dy noil golledeu eiryoet... o"r a wneuthum i ytti, mi a'e hennittafitt yn gystal ac y dugym 'All your losses which I have ever caused for you, I shall win them (back) for you as well as I took them'. Poppe regards the left-dislocated phrase, Dy holl golledeu ... etc., as 'establishing] an important concept as a separate information unit1. This, according to Poppe, is necessary because in the preceding sentence, the losses had been 'introduced as a secondary idea' as part of the implicit conditions leading to the request for the granting of nawd 'protection'. Textually however, there can be no doubt that the phrasal argument Dy holl golledeu ... etc., introduces a brand-new, and in the context fundamentally important topic into the discourse, in precisely the same manner as discussed extensively above in section 2.2, and is thus subject to fronting for those reasons. The placing of the

61 pronominal subject mi in topicalising position follows from the context of the dialogue, where the speaker identifies himself for the first time as a another prominent topic in the statement in question (he has already referred to himself in a subordinate position in the relative clause o'r a wneuthum i ytti, but this subordinate position clearly does not identify him as a major participant in the events represented in the main clause of the utterance). In Poppe's schemata (op. cit.: 188), the sentence has the structural conditions of schema 2, topic ι + (topic2 + comment). Thus, Dy holl golledeu eiryoet.. .topic ι + (mit0pic2 + a'e hennillaf.. .comment) Since topic ι is outside the complex (topici + comment), it appears once more explicitly within that complex in the form of the resumptive object pronoun in mi a'e hennillaf. But the appearance of this resumptive pronoun is therefore conditioned by the presence of the second fronted topic mi, which, as stated, arises from the conditions of the dialogue. I can only assume that, had these conditions not applied, i.e. had the Isg. pronoun not required topicalisation, the construction would have rather taken the form, Dy holl golledeu eiryoer... o'r a wneuthum i yrri a enniltafi in... etc., that is, as an 'abnormal' sentence. The distinction between left-dislocation (more narrowly, nominatives pendens) and the 'abnormal' sentence, arises here again out of quite predictable formal characteristics of the constructions in question, which therefore turn out to be formal variants, in complementary distribution, of the same functional strategy. To summarise what has been argued in the preceding paragraphs. The 'abnormal' sentence is not simply related to the nominativus pendens construction, as argued by MacCana (1973, 1991). In a certain sense, it is the nominativus pendens construction, appearing in instances where the constituent topicalised, i) has the role of subject or direct object in the following clause; ii) where, if it is the object, the clause is affirmative; iii) where there are no conditions for contrastive focus of any constituent in the clause; iv) where there is no more than one topicalisation. (Left-dislocations involving adverbial expressions form a separate subsystem of fronting, and will not be considered here.) If all these conditions apply, and there is in addition no other particle, ry or ny required in the clause, then the particle a is inserted before the finite verb, in its generalised function of indicating that a constituent appearing before the clause has the role of subject or direct object in the clause (with the additional lexical idiosyncracy of yssyd appearing for fa + yw\). The result is the 'abnormal* sentence. If any of these four conditions are broken, then the construction reverts to the classical form of the nominativus pendens as it has been recognised and extensively documented in the

62 literature, and as illustrated by various examples in the preceding paragraphs (cf. MacCana 1973: 113-14 for a similar observation). The agreement which is generally observed between the 'subject' (actually the topic with the role of subject in the following clause) and the verb in the 'abnormal1 sentence (GMW 180), is simply the agreement of the verb with the non-explicit resumptive subject pronoun, just as observed for the nominativus pendens above. The reason there is generally no such concord in relative clauses and clauses with focussed constituents is that, in such instances, there is no resumptive element to indicate the case role of the constituent in question in the clause: the case role has simply been removed to the preverbal position along with the constituent occupying that position. There is nothing within the clause with which the verb can agree, so it assumes the default personal form, identical with the high-frequency 3sg. form. The insertion of the particle a into the nominativus pendens construction to produce the familiar Middle Welsh 'abnormal' sentence is clearly a secondary development (shared by Cornish and Breton), as stated above, an extension of the original functional range of the particle. The simple question which this secondary development suggests is, can the topicalising construction lying historically behind the 'abnormal' sentence be found without the particle dl And the simple answer can hardly be other than that this is the construction which appears in the early poetry under scrutiny in the present study, taking the conventional form SV, e.g. moon dychiorant (inst. 178), OV, e.g. Teithi etmygant (inst. 312: on such object topicalisations, see section 2.8), or SVO, e.g. beird byt bamant wyr ο gallon (inst. 394), and some more complex variants discussed below. The 'subject'-verb concord observed here is once again actually concord between the verb and the non-explicit resumptive subject pronoun referring back to the topicalised constituents in initial position. In such instances as beird byt bamant wyr ο gallon then, we see neither i) archaic verbal concord with a plural non-pronominal subject in archaic SVO order (as Koch et al.), nor ii) innovated verbal agreement under the influence of Latin (as Evans 1971: 46ff. with references). Such constructions may, on the other hand be regarded i) as archaic indeed, in the sense that they show the original topicalising construction without the particle a, and ii) at the same time poetic, in that their usage may be subject to the specific constraints of poetic diction. The distribution of these constructions in the corpus in question is discussed at length below. Out of interest, it is to be noted that what is one of the oldest attested sentences in Welsh, on the Tywyn Stone, shows the old topicalising construction: GINGEN CELEN TRICET NITANAM 'Cinien's body lies beneath', in the reading and interpretation of Morris-Jones (1918: 261-63), contrast I. Williams (1973: 33-37), cf. Koch (1985/86: 66). The topicalisation of the name of the subject of the epitaph is consistent with the context: naturally the name of the person in question is the most important topic in the sentence. The appearance of the verb in apparently absolute form here provides further

63

indication of the validity of the analysis suggested here of such constructions: although there is material in the sentence before the verb, the absolute form suggests that it is, after all, in an important sense, clause initial, that is, that the preceding material is extraclausal, appropriate if this material is a topicalisation, but not if it is simply subject in an old basic order SVO. Cp. golut byt eyt dydaw The wealth of the world comes and goes' RBP col. 1055 for a similar instance. The generalisation of the particle ο in the nominarivus pendens construction, resulting in the 'abnormal' sentence, for subject and object topicalisations, may be envisaged as follows. In its original function as affirmative relative marker before clauses with subject or object antecedents, a would have had a natural place in relative clauses such as: ... yr arwr a laddodd y gelyn... "... the hero who killed the enemy ...' The example is morphologically and orthographically anachronistic, but structurally illustrative. The focussing 'mixed' sentence probably has its origins in cleft sentences of the type: Oedd yr arwr a laddodd y gelyn. 'It was the hero who killed the enemy.' The relative clause here is well motivated by the universal structural properties of cleft sentences. But it is clear that from a very early period (it is a moot point how early), the copula tended to disappear from such constructions, resulting in the familiar 'mixed' or 'inverted' sentence: Yr arwr a laddodd y gelyn. 'The hero killed the enemy.1 With the omission of the copula, the original motivation for the relative clause in such a construction will have been obscured, so that on the basis of the constructions: i) Rel.

... yr arwr a laddodd y gelyn...

ii) Focus

'... the hero who killed the enemy ..." Yr arwr a laddodd y gelyn. ' The hero killed the enemy.'

the particle a was no longer felt as a relative marker, but rather, as defined above, as a marker that a constituent preceding the affirmative clause has a structural-case role (subject or object) within the clause. Side by side with i) and ii), the language will have had the left-dislocated nominativus pendens construction, with topicalisation: iii) Yr arwr lladdodd y gelyn, The hero killed the enemy.'

64 From the base of i) and ii), the particle a was generalised to iii) in its function as a structural-case marker of a constituent preceding the clause. The result was the 'abnormal' sentence: iv) Yr arwr a laddodd y gefyn. The influence of the patterns of object pronoun infixing, with the key role in Lewis' model, sketched in the Introduction above, is not to be excluded. But it is to be stressed that the generalisation of a was not as a 'meaningless* element, but as a case marker, the function of which was extracted from the relative and focal constructions, with or without pronominal objects. Evans (1971: 45-46) notes instances where plural subject-verb concord does not occur in the Middle Welsh 'abnormal' construction, but admits, "However, in such sentences the normal practice in Middle Welsh is for the verb to agree with its subject in number and person1. Instances with lack of concord in the Middle Welsh prose are indicative of the development of the construction further away from its roots in the nominativus pendens construction in the direction of further analogy with the relative and focal constructions, where there is typically no concord, since there is no anaphoric resumptive element there, the case role of the subject being syntactically deleted from the clause when it is realised on a preceding argument. This development ultimately results in such instances as Mi syddfachgen ifanc,ff&l Ί am a foolish young boy' (Bugeilio'r Gwenith Gwyn, Gruffydd 1946: 45), where the present tense of the verb bod assumes its full relative form for all persons. The earliest pattern in the Isg. is seen in mi ytwyfeur detyf ...mi ytwyfllew L1H 16 (Gwalchmai's Gorhoffedd, J. E. C. Williams et cd. 1994: 198), Mydwyf Taliessin BT 69.9, with the original non-relative morphology. The apparently univerbated form of the latter instance, mydwyf < mi ydwyf, is possibly a further indication of an obsolete pattern undergoing reanalysis. It is to be noted that by the time of the Middle Breton and Middle Cornish sources, the 'abnormal' sentence has already been fully assimilated to the relative and focal constructions: Middle Breton examples of the 'abnormal1 sentence from Bvhes Cathell, An sanies a respontas... me a lauaro dit... an roe a lauaras dezi The saint (fern.) replied ... I shall say to you ... the king said to her' (Lewis 1966: 58, 59). The Welsh, Cornish and Breton evidence for the construction taken together should not be taken as an indication that these languages share their origins in a proto-language with the syntactic feature of basic order SVO, as Koch (1988: 31) suggests. The phenomenon is rather evidence that the Brittonic proto-language had well developed topicalisation strategies which were inherited by the daughter languages, and which underwent certain parallel structural developments in the course of their history in those languages. Instances with initial pronouns in the Book of Aneirin are discussed at length below. As far as the later development of this pattern is concerned, we can suggest briefly here

65 that it was originally a topicalising pattern for maintaining high topicality in main clauses (as observed in section 2.2, in the note to sentences 5-10 of the passage from Peredur discussed there). In the post-Middle Welsh period, while the 'abnormal' sentence became increasingly marginalised with nominal main topics, the construction with pronouns was formally maintained (presumably due to high frequency in colloquial discourse) in declarative main clauses, Mi a welais ..., Ef/Fo/Fe

a welodd..., etc. As the functional

motivation for such constructions became more and more indistinct with the marginalisation of the 'abnormal1 pattern as a whole, the initial pronouns were divested first of their topicality, then of their referentiality, so that they ultimately ceased to be pronouns at all, but assumed their characteristic role in most Modem Welsh dialects as presentential particles in affirmative declarative main clauses (with prosodic deletion of the particle a), Mi weleis i..., Mi welodd ο..., Fe weleis i..., Fe welodd e ..., etc. Fife/King (1991: 144-46) discuss their view that the 'abnormal' sentence was a Middle Welsh 'fad', suggesting that it may have its pragmatic origins in the need, or tendency, to express information-structural subtleties in a literary medium, in the absence of the usual suprasegmental clues, contrastive stress, intonation, that would have been the carriers of this information in the spoken text It is however, in my view, rather difficult to accept that the system of pragmatically conditioned word-order variation as illustrated above, by the selections from Breudwyt Ronabwy and Peredur, is the result of tendencies, conscious or unconscious, which were without structural precedent in the colloquial language. Furthermore, I find this explanation inadequate in accounting for the contrast in the use of word order as a carrier of pragmatic information between the Middle Welsh period on the one hand, and the preceding and subsequent periods of Old Welsh (as poorly attested as this is) and Modern Welsh (cf. Fife/King op. cit.: 148, n. 7, Fife 1993: 30). I would have expected the 'abnormal sentence', as a particularly literary feature of the standard language, to have been maintained as a characteristic of literary Welsh in the post- Middle Welsh period: but this period saw the decline of the construction. I suggest that the following sketchy scenario may be nearer the truth. In the Old Welsh period, where our sources are restricted to glosses, legal documents, charters, those responsible for these 'texts' may have contented themselves with the baldest, least expressive declarative sentences (the Old Welsh poetry that we have in contemporary manuscripts is too sparse to allow a meaningful judgement). In the post-Middle Welsh period, there is clear evidence of the emergence of a strict, prescriptively regulated standard literary language that was to deviate through the centuries ever more extremely from spoken usage. Considering that modem literary Welsh continues in principle the standard set by William Morgan's translation of the Bible from 1588, itself leaning heavily on the poetic usage of that and the preceding century, it is no wonder that modern colloquial Welsh is sometimes met with bewilderment by non-Welsh speakers familiar

66 only with the literary language, and vice versa, that native speakers of Welsh often react with exasperation to the archaism of literary Welsh. Be that as it may, it is, I think, conceivable that precisely the Middle Welsh period presents us with an interface between spoken and literary usage that has not been reproduced since. The interaction of oral and literary media implied by the writing down of sagas and romances may have been accompanied by an infusion of colloquial usage into the literary norms, which may have included the formal coding of pragmatically conditioned, information-structural phenomena, previously ignored and subsequently abandoned. It is not inconceivable that the large number of translations out of generally subjectinitial Old and Norman French, with additional Latin and English influences, may have abetted the spread of the 'abnormal' sentence to functional contexts, in which it would not have originally been appropriate. And the type of multiple fronting discussed by Fife/King (1991: 144-45, cf. Poppe 1990, I991a) is also in all likelihood a purely literary extension of the original system. Nevertheless, the consistency and above all clearly discernible pragmatic motivation for the word-order variations in the passages discussed above, for which no precedent could be found in any mediaeval French, Latin or English sources, suffice to indicate to me that we are dealing with a subsystem of the native Welsh grammar. The reasons for the subsequent abandonment of such structures are undoubtedly complex, but I would assign a decisive role to the following factor, which appears not to have been considered thus far. The formation and cultivation of a literary norm assume more or less conscious grammatical analysis on the part of those responsible for it. They assume that scribes, grammarians, literati can come to a consensus (explicit or tacit) upon what is a (prescriptively defined) correct expression of a certain content and what is inadmissible. The nature of the inventory of prescribed, admissible structures is thus fundamentally dependent on the nature of the grammatical models available to the authorities responsible. Now, pragmatics is a very new branch of linguistic and grammatical theory (at least in western tradition). Even today it is viewed in certain circles with scepticism regarding its scientific value. In the tradition of classical, Latin-based grammar that was available to the mostly nameless mediaeval and Renaissance scholars who would have been responsible for the gradual crystallisation of standard literary Welsh, such concepts as semantic roles, the topic hierarchy, topic maintenance and salience in the discourse, to name but a few of the phenomena relevant to the present discussion, were unknown. Thus, any grammatical constructions which we may now be in a position to classify and analyse as being dependent on or motivated by such phenomena, could only have been opaque to those trying to standardise the means of expression. It is therefore not surprising that such grammatical devices were gradually rejected as 'undesirable', stylistically 'deviant', indeed 'abnormal', as part of the formation of the literary norm. On the other hand, as Henry Lewis in a roundabout way

67

observed (see the Introduction above), the 'abnormal' sentence is indeed a normal part of modem colloquial Welsh, if, perhaps, stylistically more restricted than it appears in Middle Welsh texts. Furthermore, the consistent use of such a pragmatically motivated subsystem of word order in Modern Breton may be attributed partially to the very late and superficial mediaeval literisation and standardisation of Breton (similar considerations may be applicable to Cornish). 2.5 The sort of analysis of the distribution of various word-order patterns over an extended passage that was undertaken for Middle Welsh prose in section 2.2 above is not possible in the poetry which is the actual subject of the present study. The poetry does not constitute a single continuous text, rather, each short awdl constitutes a complete text in itself, and any pragmatic conditions applying specifically to a single awdl do not extend beyond the boundaries of the awdl to others. It is therefore not possible to trace the continuity of individual topics over extended passages consisting of a number of thematic paragraphs. Nevertheless, it is my contention that, within the boundaries of the individual awdlau, precisely the same pragmatic considerations as in the Middle Welsh prose do indeed condition and largely determine the observed distributions of the word orders employed by the poets. In order to justify this claim, it is necessary to investigate the occurrences of the word-order patterns in their full discourse cotexts, that is, to determine what role the various patterns play in the awdlau in which they occur. The essential issue here is the distribution of basic VSO vs. topicalised SVO. For that reason, only these constituent orders will be examined in detail here. Furthermore, since, as will be seen in section 2.9, the presence of a pronominal subject imposes certain additional restrictions on the type of word order employed, only those instances of VSO and SVO in which the subject is substantival ([-pro]) will be considered here. The patterns with subject but without explicit object, VS and SV, may be expected to follow the same pattern of distribution as VSO vs. SVO, but since V[S-pro] outnumbers [S-pro]V by nearly 6: 1, there can hardly be any doubt as to the basic nature of the former compared with the latter (further considerations concerning the structure of these patterns in section 2.10 below; the related [S+pro]V is also considered in sections 2.9 and 2.10, and the pattern with topicalised object [O-pro]V in section 2.8). Finally, as was observed in the previous chapter, the pronominalisation of the object results in the structures [O+pro]VS and S[O+pro]V: since the former is practically restricted to the formulaic expression nys/neus adrawd..., instt. 361-74, a comparison of these patterns would not be meaningful for the present corpus. In the quotations of the awdlau following, punctuation has been introduced, and the usual editorial conventions adopted in this work have been followed. Where the correct

68 interpretation of a particular line or phrase is in doubt, no attempt has been made to reach a definitive conclusion here, as long as this has no bearing on the analysis of the wordorder patterns under consideration. Otherwise, the emendations or interpretations are those available in the standard works (though my precise translation may differ in detail from these), or are discussed under the relevant instances in the classification given in section II. Exceptions to this practice are the discussions of the first line of awdl B19, of a line in B34, and of a line in A72, where I differ significantly from previous commentators. In each awdl quoted, the position of the relevant instance of VSO or S VO is noted, as well as any other instances appearing in the classification in section II. The awdlau are lineated individually here for convenience of reference in the discussion below. Syntactic notes to the individual awdlau appear in the form of footnotes.

V[S-pro]O CA 11. 74-83 = AwdlA9.

5

10

5

1

Gwyr a aeth Gatraeth, ueduaeth uedwn,1 fyryf, frwythlawn, oed cam nas kymhwyllwn. E am lavnawr coch, gorvawr gwrmwn, dwys, dengyn ed emledyn aergwn. Ar deulu Brenneych beych bamasswn. [E]ilyw dyn en vyw nys adawsswn. Kyueillt a golleis, diffleis oedwn, rugyl en emwrthryn, rynn riadwn. Ny mennws gwrawl gwadawl chwegrwn, VSO maban y Gian o Vaen Gwynngwn.

(= inst. 489) (= inst. 336) (= inst. 106) (= inst. 309) (= inst. 310) (= inst. 542) (= inst. 355)

The men who went to Catraeth, a mead-nurtured ?host, firm, effective, it would be wrong were I not to mention them. Around red blades, great dark-blue sockets, in close ranks, grimly fought the battle-hounds. On the war-band of Brenneych I should have judged it a burden. I should not have left even the shape of a man alive. A comrade I lost, I was faithful, speedy in combat, the steadfast leader.

LI. 1-6 of the awdl present general images of battle, centered on the warband of the Gododdin, the 'men who went to Catraeth1, as a whole, and indeed, the awdl begins with the topicalisation of Gwyr a aeth Gatraeth, as the main theme of the first part, as discussed in note 13 of section II. LI. 5-6 include the further object topicalisations of beych 'a burden' and fejiliw dyn 'the shape of a man1, as quite unpredictable topics in the context. L. 7 then introduces the individual hero of the poem, in a sentence with focus, '(It was) a friend I lost'. The focus on 'friend', and the unambiguous reference of this word to one of the heroes of the poetry suffice to establish this, as yet unnamed hero as the main topic of this part of the awdl, so that when the hero appears as the subject of a verbal form in 1. 9, as the substantivised adjective gwrawl 'the brave man', no topicalisation is required. Information as to the identity of the hero is given only in the last line of the awdl. This is a common strategy in this poetry, and must be kept in mind in consideration of the pragmatic-syntactic structuring of many individual passages, some of which are the subject of discussion in these notes. In this case, the information about the hero's identity is incomplete: we are given his father's name, and the noun maban implies that the hero was young.

69

10

The brave man did not demand a father-in-law's dowry, the young son of Cian from Maen Gwynngwn.'

CA11.97-104 = Awi//A12. Gwyr a aeth gatraeth gan dyd;2 (= inst. 492) neus goreu ο gadeu gewilyd; (= inst. 210) wy gwnaethant en geugant gelorwyd (= inst. 405) a llavnawr llawn annawd em bedyd; goreu (yw) hwnn, kyn kystlwn kerennyd, VSO (= inst. 356) enneint creu ac angeu o'e hennyd. Rac bedin Ododin pan vudyd, (= inst. 11) neus goreu deu[r] bwyllyat Neirthyat gwychyd. (= inst. 437) The men who went to Catraeth with the day; he made of armies shame; they made biers inevitable with blades full of cruelty in the world; this (hero) made, rather than declaring peace, a bath of blood and death of his enemy. When he ?... before the army of the Gododdin, Neirthiad the splendid performed his brave intent.1

CA11. 258-67 = AwdlA23. Aryf angkynnull,3 agkyman dull, agkysgoget, 2 The complex interweaving of images associated with the warband as a whole and with the individual hero at the beginning of this awdl are discussed in note 13 of section II. LJ. 1, 3-4 are concerned with the warband, while 11. 2,5-8 deal with the individual hero. In 1. 2, the hero is already treated as a given topic and is thus untopicalised, the subject of goreu in this line being an implicit 3sg. pronoun, with cataphoric function, referring forward ultimately to the naming of the hero as the very last words of the awdl, Neirthyat gwychyd. So also in 1.5, the demonstrative pronominal subject appears in basic order, as no new concept is introduced. Here, the element of deixis involved in the demonstrative hwnn gives the poetically desirable effect of enhancing the impression of the immediate presence of the events narrated. In the VOS structure of the final line, the subject is delayed, extraposed, achieving the effect of naming the hero in the last words of the awdl. This, paradoxically, too, has a topicalising effect, in that the hero is thus the concept uppermost in the mind of the audience as soon as the time following the perception of the awdl begins. The hero is made the main topic of any discourse beyond the boundary of the awdL The essence of the heroic immortality of fame is indeed that the name of the hero persists in time, and naming him last is one strategy for ensuring that he metaphorically lives on beyond any direct physical or linguistic experience of him. 3 Much the same considerations concerning the syntactic status of the hero in this awdl apply as in the last, except that the first four phrases of the passage here present images figuratively representing the hero, and especially his effects on the enemy: he was a weapon which caused disorder in the ranks of the enemy, causing them to break their battle formations by his violence, he was unshakable in standing his ground, he was the cause of great destruction. These rather complex images are expressed minimalistically by the pregnant phrases Aryf angkynnull, agkyman dull, agkysgoget, trachywed vawr. The syntactic effect of these phrases is that by the time the hero is more or less explicitly introduced as the subject of a verb, as Uawr 'the lone hero', the initial phrases have already topicalised him, assuming understanding on the part of the audience with respect to the tradition-bound diction of the poetry, whereby such phrases are to be perceived as referring to the hero. As subject of the further two finite verbs of the awdl, implicit cataphora identified by the verbal flexion alone suffices to maintain the hero as main topic, leading up to the naming of the hero in the last line but one.

70

5

10

5

10

trachywed vawr, treiglessyd llawr Lloegrwys giwet. Heessit eis yg kynnor [tr]eis, yg cat uereu. Goruc wyr ll[y]dw a gwraged gwydw kyn no'e angheu. Greit uab Hoewgi,[ ]r()ac ysberi, y beri [c]reu.

VSO (= inst. 357) (= inst. 222) (= inst. 223) (= inst. 569)

'Disordering weapon, (causing) broken array, unshakable, great destruction, the lone hero overthrew the English mob. He sowed spears in the van of combat, amongst the battle-spits. He made men low and women widows before he died Graid son of Hoy wgi, before spears, caused blood.'

CA 11. 277-91 =AwdlA24. Arwr y dwy ysgwyd adan e dalvrith,4 ac eil tith orwydan, bu trydar en aerure, bu tan, bu ehut e waewawr, bu huan, bu bwyt brein, bu bud e vran, a chyn edewit en rydon,

(= inst. 22)

4 While the hero is named once again only in the last line, this awdl differs in a significant respect from those so far considered in the present notes, in that this is the first case in which the very first word of the awdl consists of an explicit (non-figurative) reference to the individual hero in question, here in the form of the word anw 'the hero'. It is unfortunate that the precise meaning of the first line in doubt, but there can be no doubt that the appearance of anvr as the first word constitutes a forceful expression of the high prominence of the hero in the discourse constituted by this passage of the poetry. The usual strategy of anaphora is employed for the subject of the series of bu in 11. 3-5, as well as for the impersonal edewit in 1. 6, and for ytoed in 1. 13. The hero further appears as a possessive argument in 11. 10-14 (further on the syntactic significance this phenomenon below). L. 9 consists of the famous, striking beird byt bamant wyr ο gallon 'Poets of the world pronounce judgement on men of heart', with initial, topicalised subject. Now, it may be wondered, why there is topicalisation here, if it is considered that, as was often the case in the use of the Middle Welsh 'abnormal' sentence discussed above, the topicalisation here has the effect of fixing attention on a new main topic. The 'poets of the world' are certainly not that in the thematic context of the whole awdl, for they only have relevance in this single, parenthetical clause, and thus clearly do not interfere with the status of the hero as the main topic of the passage. However, the relevant feature of these 'poets' in this context is not their importance over an extended passage, but rather their high (in principle 100%) unpredictability. The presence of beird byt as subject in the line in question is completely devoid of motivation from the cotext (though not too great a surprise from the point of view of the poetic context). This lack of cotextual motivation resulting in high unpredictability of the new topic is the trigger for the topicalisation, producing the SVO structure: putting it simplistically, the audience must be surprised into noticing that this line (and only this line) is about something else, and the surprise effect is achieved by the topicalisation. In the instance of VSO in 1. 14, no such surprise effect is needed: the "blood1 and 'armour' have indeed not been mentioned before, but their presence in a text concerned with the last feats of a warrior hero are cotextually quite unsurprising. And since it is still the hero who is the main topic, and who appears in the clause as well, in the form of the possessive e, y, topicalisation of the subject e greu would have been quite inappropriate. The same considerations apply to the VOS clause in 1.10, also containing possessive pronouns referring to the hero, but here with reversal of the verbal arguments, in accordance with the principles discussed in section 1.3.

71

10

15

5

10

15

gan wlith, eryr tith tiryon, ac o du gwasgar gwanec tu bronn, (beird byt barnant wyr o gallon), diebyrth e gerth e gynghyr. Diua oed e gynrein gan wyr. A chynn e olo a dan Eleirch Vre, ytoed wryt en[ ]e arch, gorgolches e greu y seirch, Budvan vab Bleidvan dihavarch.

SVO (= inst. 394) VOS (= inst. 440)

VSO (= inst. 358)

The hero ... his grey forehead, and like the movement of a steed, he was clamour on the hill of battle, he was fire, his spears were swift, he was a sun, he was the food of ravens, he was profit for a raven, and before he was left in ?... with the dew, the eagle of graceful movement, and beside the spray of the wave by the hillside, (poets of the world pronounce judgement on men of heart), his counsels deprived him of his rights. His warriors were destruction in the opinion of men. And before he was buried under Eleirch Fre, there was valour in his trunk, his blood washed over his armour, Buddfan the mighty son of Bleiddfan.1

CA 11. 300-31 \=AwdlA26.

5

10

Bu gwir mal y meud e Gatlew,5 ny deliis meirch neb Marchlew. Heessit waywawr y glyw y ar llemenic llwybyr()dew. Keny vacket am vyrn am borth, dywal y gledyual emborth. Heessyt onn o bedryollt y law y ar veinnyell vygedorth. Yt rannei rygu e rywin. Yt ladei a llauyn vreith o()eithin. Val pan vel medel ar vreithin e gwnaei Varchlew waet()lin.

(cp. inst. 559) VSO (= inst. 359) (= inst. 227) (= inst. 23) (= inst. 228) VSO (= inst. 360) (= inst. 24) (= inst. 119) VSO (= inst. 361)

5 The pragmatic structuring of this and the following three awdlau is not as clearly defined, in the terms used here, as in those so far discussed, for various reasons. Here it is to be observed that the hero of the awdl is introduced by name first in the second line as object in the bask-order clause ny deliis meirch neb Marchlew 'No one's horses caught up with Marchlew'. This expression of Marchlew's great speed, together with the fact that no other hero is named subsequently, automatically force Marchlew to the status of main topic for the whole awdl. And no other unexpected topics (as beird byl above) appear in main clauses: the subordination of the clausal simile val pan vel medel ar vreithin 'As when a reapingparty strikes in unsettled weather', 1. 11, effectively excludes the subject here from topicalisation strategies: the 'reaping-party1 is certainly new and unexpected (that is part of the poetic effect of the line), but the subordinate role of (be clause to the flow of the sense of the passage as a whole blocks any topicalisation (contrast the extreme asyndetic nature of the parenthesis beird byt barnant wyr ο gallon in the previous example). The result is that no topicalisations are employed, and all clauses appear in basic order, with or without explicit subject.

72

5

10

'It was true as Cadlew ?..., No one's horses caught up with Marchlew. He sowed spears in battle from a leaping, broad-tracked steed. Although he was nurtured ?... fierce was his sword-blow ?... He cast ash from the four gaps of his hand from a slender, bay, steaming steed. The most dear one would share out his plentiful wine. He used to kill with a blood-stained, fierce blade. As when a reaping party strikes in unsettled weather would Marchlew make bloodshed.'

CA 11. 343-52 =Aww//A34.

5

5

Ny wnaethpwyt neuad mor orchynnan,7 (= inst. 231) mor vawr, mor orvawr y gyvlavan. Dyrllydut medut, Moryen tan. (= inst. 232) Ny thraethei na wnelei Kenon kelein. VSO (= instt. 29, 363) Un seirchyawc, saphwyawc son edlydan. Seinnyessit e gledyf em penn garthan. (= inst. 126) Noc [y ]c()escyc carrec (vyr) vawr y chyhadvan, (= inst. 127) ny mwy gysgogit Wit uab Peithan. (= inst. 233) 'So clamorous a hall was never made, so large, so great for slaughter. You deserved the enjoyment, fiery Morien. He would not relate that Cynon would not make a corpse. An armoured one, wielding a spear of very broad fame. His sword resounded on the top of the rampart Than a rock of great circumference budges, Gwid son of Peithan would not be budged more.'

CA11. 459-69 = AwdlMO.

5

10

5

Am drynni, drylaw drylenn, am Iwys, am[ ]diffwys dywarchen am gwydaw gwallt e ar benn, y am wyr eryr Gwydyen, gwyduc neus amuc ae wfialen], ardullyat, diwyllyat e berchen. Amuc Moryen gwenwawt Mirdyn, a chyvrannv penn prif eg weryt ac an nerth ac am hen. Trywyr yr bod bun Bratwen, deudec Gwenabwy vab Gwenn.

(= inst. 339) VSO (= inst. 364)

'For the battle, sad destruction, for the fair, desert turf, for the falling of hair from the head, for the men, Gwyddien was an eagle, wildly he defended them with his spear, a musterer, a deforester its owner. Morien defended the fair song of Myrddin, and placed the head

7 This envt and the next, awdl A40, unfortunately present a rather confused picture. Symptomatic of this is the fact that cnvdl A34 mentions three heroes, A40 no less than four (not counting the clearly subordinate Mirdyn), without any, as yet, discernible motivation or progression in the switching from one to the other (contrast the case of the preceding owtfl A30). Note also the sudden switch from second person address to third person narrative in 11. 3-4 of ωνΛ Α34. Such switching is by no means unparalleled in early Welsh poetry, and indeed, in a case like poem III of Canu Taliesin (addressed to Urien), it produces particularly artful paragraphing with respect to content and form. But there at least the subject of the OHX remains constant, whether expressed in the second or third person. The effect in the present instance is rather abrupt, and the switch is not just one of person, but apparently of reference also, from Morien to Cynon. It is possible that there is a particular poetic principle involved here which we are missing, or perhaps the manuscript text is here after all a little corrupt, the latter impression being particularly strong in the case of awdt A40, especially side by side with its Β-variant, BS. I shall not attempt an analysis of these awdlau at this time.

74

10

of the chieftain in the earth with ?... Three men for the sake of a maiden was Bradwen, twelve, Gwenabwy son of Gwen.'

CA 11. 553-60 = G[w]roled gogled, gwr a'e goruc,8 llary vronn haeladon 'n[ ]y eissyllut. Nyt emda daear, nyt emduc mam mor eiryan gadam, haearn gaduc. Ο nerth e kledyf, claer e'[m] hamuc. Ο garchar anwar daear e'm due. O gyvle angheu, o anghar dut: Keneu vab Llywarch dihauarch, drut.

(= inst. 458) VSO (= instt. 442, 365) (= inst. 341) (= inst. 342)

The heroic deed of the North, the hero did it, a lord of generous heart in his nature. There does not travel the earth, no mother gave birth to one so fair and powerful, cloaked in iron. With the strength of his shining sword did he defend me. From the cruel prison of the earth did he take me. From the place of death, from unfriendly country: Mighty, brave Genau son of Llywarch.'

CA 11. 608-16 = Aw MANNER That which is to the left in the hierarchy is prototypically more salient than that which is to the right, and thus has more immediate access to the syntactic status of subject The predicate 'washing over* involves a non-conscious participant, i.e. a patient, a liquid, here "his blood', and a location, where the 'washing over' takes place, here, "his armour'. The role of patient is higher on the topic hierarchy than location, so the former is coded syntactically as subject, the latter as direct object Impersonal and antipassive constructions arise (for the simplest instances) when there is only one topic to be paired, but a predicate which selects two arguments (in traditional valency terminology, a two-place verb). As examples can be taken instt 299 sengit guid gunet The trees of battle would be trampled* (impersonal, i.e. with demoted or excluded

88 subject argument), and 196 eflladawd Ήε used to kill' (antipassive, i.e. with demoted or excluded object argument, cf. Givon 1990: 624ff.). Topic

Clause 0 1 segit

«=

Predication Agent I Predicate

«=

Patient

Clause S

I

I lladawd

I DAT/BEN > PAT > LOC > INST/ASSOC > MANNER A dative argument is prototypically more salient than a patient argument in terms of this hierarchy. As a result, a dative argument has greater accessibility to the subject position than a patient argument. In the sentence beird byt bamaru wyr ο gallon, the topic beird byt occupies the highly prominent preverbal position. Therefore, if the sentence were construed with topicalised object, TiVS(Oi) (= apparent OVS), this would not be in agreement with the topic hierarchy. There would be a pairing of the highly prominent topicalised constituent with the relatively low-salience patient argument, the object In reality, the highly prominent topicalised constituent is automatically, by default, paired with that argument in the clause which is relatively higher in the topic hierarchy, here the dative argument, which takes precedence over the patient argument in occupying the subject position. Beird byt bamant wyr ο gallon is thus construed with subject topicalisation, TiV(Sj)O (= apparent SVO). This is the default case. There are naturally instances of TjV(Si)O (= apparent SVO) where the selectional properties do indeed unambiguously identify the correct pairing of topics with arguments, e.g.: Inst. 397 twrch goruc amot The boar made a pledge': gwneuthur selects an agent as subject = the animate twrch "boar1 (metaphorical for the wild hero) and patient as object = the inanimate, abstract amot 'pledge'. Inst. 401 tavloyw ac ysgeth lavle[i] wydrin a med Tafloyw with a ?shrug would knock back a glass of mead': same argument selection as in previous instance; agent subject = animate tavloyw, patient object = inanimate wydrin a med. Note that the selection of agent 'as subject' and patient 'as object' does not really have to be specified in the description. It would be enough to say that tavlu or gwneuthur select agents and patients as arguments: the syntactic roles or these argunents in the clauses follows from the topic hierarchy. There is also an instance in which both arguments are patients, i.e. have equal status in the topic hierarchy: inst 400 peleidyr pwys pferjiglyn benn periglawr 'Weighty spears would endanger the chief endangerer'. The TjV(Si)O (= apparent SVO) analysis is adopted, with the highly prominent topicalised constituent paired with the prototypically most salient syntactic role of subject Though neither the selectional properties of the verb nor by default the topic hierarchy force an unambiguous pairing, nevertheless the semantic features of the predicate and arguments involved guide to the correct pairing.

92 The verb periglu, peryglu can be construed semantically as a causative, 'cause to be in danger'. Since 'danger1 is a concept prototypically indicating the potential of harmful change in the condition of an at least animate, more typically human argument, a predicate 'cause to be in danger' will almost inevitably be directed towards a human patient: The chief endangerer would endanger the heavy spears', while not exactly inconceivable, would be an extremely marginal interpretation, and is practically excluded by the feet that the structure allows the prototypical interpretation with 'heavy spears' as subject and 'the chief endangerer1 as object. The fact that in the default instance, the topic hierarchy is maintained, the highly prominent topicalised constituent being paired with the most salient argument as subject, does not however mean that the topic hierarchy is inviolable. Violations of the hierarchy are in any case part of the usual grammatical apparatus of the language in passive/impersonal constructions for instance. Thus, the verb gwanu 'strike (down)' naturally selects an agent and a patient as arguments, which are in turn prototypically subject and object in the default instance, in accordance with the topic hierarchy. But in an instance such as 292 pan wanet yg kyueilh 'When my comrade was struck down', the agent argument has been so stripped of salience that it does not appear in the utterance at all, and it is the patient argument which has all the prominence of the sentence. This constitutes a violation of the topic hierarchy, which naturally assigns greater topicality to the agent than to the patient It is this violation, or better, by-passing of the hierarchy which is morphologically coded by the impersonal flexion of the verb. It may therefore be expected that there are also topicalisations which by-pass the topic hierarchy. This is indeed the case. And it may further be expected that this can only take place when certain conditions apply to neutralise the default analysis. This too is the case. There is one instance in the corpus of apparent OVS, i.e TjVS(Oi), with topicalised, non-pronominal object: inst. 469 gu[e]ro rybud ο gat dydygei gant Ά hundred (men) would bring a bitter warning of battle'. Here, the default analysis of pairing the topicalised, but inanimate gufejro rybud ο gat with the subject argument is blocked by the selectional properties of the verb dydwyn 'bring' which requires an agent argument, necessarily animate, which is therefore syntactic subject by the topic hierarchy.

Topics gow(anim.)

Clause =* S

«-

i dydygei I +gu[e]ro rybud (inan.) =» Ο

Predication Agent

I «= «=

Predicate I Patient

93 The patient argument is given higher prominence than the agent, though the prototypical relations remain constant, so subject and object roles are assigned in accordance with the hierarchy. Pairing of topics with arguments takes place unambiguously because of the lexical properties of the constituents involved (argument selection, animacy). Note that the topicalisation of the patient object is not the only indicator of the decrease in the prominence of the agent subject. The internal structure of the latter itself also shows this demotion. The expression gant 'a hundred' is elliptical: 'a hundred men/warriors/heroes' is meant, and it is these 'men/warriors/heroes' that constitute the actual agent argument of the predicate. With the suppression of the prominence of the agent, these 'men/warriors/heroes' do not appear explicitly in the sentence, but are indicated elliptically by the numeral gant. It may be concluded that the combination of the topicalisation of the object and the ellipsis of the subject expression form a very special set of pragmatic conditions, which together work to license this particular by-passing of the topic hierarchy by neutralising the default analysis. The extreme particularity of these conditions may account for the uniqueness of the structure of this instance in the corpus, the only instance of Tj VS(Oi) (= apparent OVS).

2.9

There are however more instances in the corpus than this of topicalised objects. There is a not insignificant number of instances of TiV(S)(Oi), that is, apparent OV structures (non-pronominal object), with the topicalised constituent paired with the object argument (which does not appear explicitly postverbally, see further below for the distribution of resumptive pronouns vs. zero anaphora). There is, as to be expected by now, no explicit subject, which essentially means that the subject is pronominal and identified solely by the verbal agreement, or the construction is impersonal. It should be clear that such structures, by-passing the topic hierarchy in a highly marked manner, should only be attested under certain stringent conditions. Let us consider a hypothetical instance, which, on the face of it, would appear quite appropriate in this heroic poetry: esgar Uadei 'He used to kill the enemy'. The agent subject, 3sg.m. 'he' is identified by the verbal form, esgar 'enemy' is topicalised and paired with the patient argument of Uad. No instance comparable to this hypothetical one is attested in the corpus. The reason is derived from the need to pair topics unambiguously with arguments. There is no way to achieve this in an instance such as esgar Uadei, to arrive at the interpretation TjV(S)(Oi) (= apparent OV). Uad selects an agent and a patient as arguments. Esgar 'the enemy', as an animate noun, could fulfil either role. There is thus no unambiguous pairing possibility which would identify the topic esgar with the patient argument as object The default strategy would therefore be engaged, the topic hierarchy, and the high-

94

prominence topic would have to be paired with the agentive subject argument. A sentence such as esgar lladei is therefore excluded from the interpretation TjV(S)(Oi) (in fact, such a sentence is not attested in the corpus at all, i.e. it is also excluded from the interpretation TiV(Sj)(O) = apparent SV, see below). It can be seen then that the TjV(S)(Oi) structure should only arise when the topicalised constituent can be unambiguously paired with the object argument of the verb. In principle, there should be two ways of achieving this: i) argument selectional properties, and ii) unambiguous verbal flexion. i) Selectional properties: this strategy has been considered in detail above in another context. What it specifically implies here is that the topicalised constituent has certain lexical features such that there is only the possibility of pairing it with the object argument of the verb. Examples are: Inst. 313 blaen bragat briwei 'He would shatter the front(-line) of the army'. Briwaw selects an agent and a patient Blaen bragat 'the front of the army' is inanimate, and cannot therefore be paired with the necessarily animate agent as subject It must therefore be paired with the patient as object Bragat 'army' alone is animate, but as a relative spatial concept, blaen 'front' is inanimate. Topics 3sg.

Clause =» S

«=

I briwei

I *=

I +blaen bragat (inan.)



Ο

Predication Agent Predicate

I

S

VO. The syntactic change consists of changes in the syntax and pragmatics of the enclitic auxiliary pronouns and in the case-selecting properties of the verbal noun. These latter changes may well be motivated by the typological factors indicated above, but the syntactic means by which the typological change is realised cannot be accounted for in a theory of typology. Rather they present language-specific and construction-specific analyses and reanalyses. Another example. Both Irish and the Brittonic languages have postposed demonstratives. In Old Irish, these appear as indeclinable, enclitic particles, e.g. in leborso 'this book', ind libuir-se Of this book' or 'these books' (the variation so ~ se is mechanically conditioned by the quality of the preceding consonant). In Welsh on the other hand, the demonstratives appear as accented, inflected adjectives, e.g. y llyfr hwn 'this book' (masculine), y dref hon 'this town1 (feminine), y ttyfrau hyn, y trefi hyn 'these books, these towns'. The word order, noun + demonstrative, is typologically consistent with the VSO basic order of the Insular Celtic languages. However, there are good reasons to suppose that the articles of these languages have developed from preposed demonstratives in their respective immediate forerunners (Isaac 1993b: 10; 1994a: 339-40, cp. Gaulish σοσιν νεμητον, Lejeune 1985: 205-9; sosin celicnon, Lejeune 1988: 147-55). The diachronic change so postulated, demonstrative + noun > noun + demonstrative, is once again of a typological nature, consistent with the transformation of a VSO language from a (possibly much) earlier SOV language. But as well as a diachronic change of this sort can be captured in a theory of typological change, such theories say nothing about the precise syntactic character of the individual developments in the attested languages. In other words, the typological classification of different types of change does not capture, for instance, the construction-specific and language-specific syntactic details of the development of the Old Irish demonstrative particles and the Welsh inflected, demonstrative adjectives. But precisely these developments, and other, similar ones, would have to be the object of a theory of diachronic syntax change. Typological analyses and theories are undoubtedly interesting and indispensable contributions to our understanding of language in all its aspects, but

138 they do not take the place of a theory of diachronic syntax, which, in order to facilitate reliable reconstructions, would have to include a 'principle of the regularity of syntactic change'. An analogy from the field of phonology again highlights the problem. We can state that the Goidelic occlusives /t, k, kw, b, d, g, gw/ are reflected in Old Irish in intervocalic position by the spirants /θ, χ, β, δ, γ/ (with loss of the labial element in the original labio-velars). This statement can be generalised in a consistent phonological theory, in the form the following rule (using the articulatory features of Chomsky/Halle 1968): [-sonorant, -continuant] > |+continuant] / f+vocalic] _ [+vocalic] Cp. the form of the general rule of phonological change discussed in section 3.6 above. For the Brittonic languages, this rule is valid in the variant: [-sonorant, -continuant, +voice] > [+continuant] / [+vocalic] _ [+vocalic] This rule captures the Brittonic change of /b, d, g/ to /β, ο, γ/ in intervocalic position (the later treatment of/γ/ in the individual Brittonic languages can be left aside here). For the voiceless occlusives, /p, t, k/ (p < PIE *kw), another rule applies: [-sonorant, -continuant, -voice] > [+voice] / [+vocalic] _ [+vocalic] It is to be noted that the definition of the feature [±voice] is not entirely unproblematic, cf. Ball (1984: 13ff.), Ball/M ller (1992: 83-86), but can be taken as a primary term for convenience here. These rules are descriptive, not explanatory. They constitute no attempt to explain why these changes took place in these positions, they simply capture the fact that they did, and in such a manner as to allow them to be stated within a general theory of phonology. But these rules (and the many others required for a full description of the phonological development of the attested Insular Celtic languages), together with the principle of the regularity of phonological change suffice to make reconstruction possible. That is, even if we had no attestations of Celtic languages at all outside historical Insular Celtic, it would still be possible to compare W cad, Olr. cath («-stem), and reconstruct reliably a shape *katu-, or W budd 'profit' with Olr. buaid (/-stem), and reconstruct a shape *bodi-, if not *boudi- (in both cases taking account of the relevant additional rules). Such rules can, of course, be formulated in a number of different ways, depending on the type of phonological theory adopted. But this terminological variation does not alter the fact that by means of such rules, language change at the phonological level can be

139 consistently and adequately described. For a given language or language family, it is not a priori predictable that such a rule will necessarily operate: that must be established by empirical observation. But by means of a theory of phonological change, scholars are not restricted to mere observations, but are in a position to classify rigorously and comprehend the nature of various changes. And by the application of the principle of the regularity of phonological change, reconstructions of proto-forms can be achieved. When we turn back to the syntactic example of the Insular Celtic postposed demonstratives, however, it turns out that the difference in the syntax of the Old Irish particles, and the Welsh inflected adjectives cannot be described and comprehended in any way within a general theory of syntax that would lead us to a reliable reconstruction of the proposed demonstratives of the proto-languages. The typological observation of the change in word order from demonstrative + noun to noun + demonstrative does not capture the syntactic development to particles in Irish and adjectives in Welsh; while the reconstruction of the articles of the languages as preposed demonstratives in the protolanguages, e.g. Olr. in etc. (in various declensional forms) < *sindos-etc., is once again a purely phonological reconstruction. However we look at the situation, we have no way of comparing the syntax of Olr. -so and W hwn, for instance, and arriving at a syntactic reconstruction of preposed demonstratives, inflected or not That is, if we had neither the evidence of the Insular Celtic articles, nor the Gaulish demonstratives (which perhaps have only indirect bearing on the arguments after all), we would have no information about the syntax of the demonstratives in Celtic prior to the historical languages, since there is no theoretical apparatus which captures and consistently describes such developments of a syntactic category, and in such a way that anything can be reconstructed. Useful discussion of various theories of diachronic typological change is to be found in Bean (1981: 18ff.). Attempts to unify typological insights with results from general theories of grammar are made by Vennemann (1974a, 1975), Hawkins (1983, 1989), Aristar (1991), Dryer (1992).

3.8 It should be clear by now that syntax is something so different from phonology and morphology that quite different methods are needed in order to reconstruct it, whatever these methods may be, and not just the same methods, slightly differently applied. In the following, I shall give some consideration to the question as to why this should be. Every utterance consists essentially of, at least, two representations: its external, acoustic form, and its meaning (however defined). Only a little thought is required to come to the conclusion that these two types of representation are so different that it is a

140 wonder that they can be brought into a relationship at all, not to mention the intimate relationship which exists between them in the utterances of a natural language. This idea was made explicit by Chomsky (1965: 137-38):

The final effect of a grammar, then, is to relate a semantic interpretation to a phonetic representation - that is, to state how a sentence is interpreted. This relation is mediated by the syntactic component of the grammar, which constitutes its sole "creative" part.

This conception of a model for the structure of a grammar has remained in the centre of generativistic research, into the so-called 'Minimalist Programme1, as presented by Chomsky (1992: cf. in particular 1-4). This model can be informally represented as follows:

Lexicon

Interpretation Derivation External form

Meaning

It cannot be maintained that Chomsky's model, so interpreted, is accepted by all researchers. It is too well-known that that is far from being the case. But the essential matters here are not specific features of individual models, but rather the empirical and conceptual foundations of all approaches. Thus, for instance, for Talmy Givon (1984: 32), 'In terms of coding ... pragmatics is jointly coded with propositional semantics via syntax.' I.e. syntax is the iconic coding strategy of pragmatics and propositional semantics in a language. This assumes that an utterance has a perceptible, observable form, which is conditioned by the way the syntax codes the contentual elements of the utterance, the meaning in the broadest sense. The meaning of an utterance is mediated by the syntax. Givon's conception can be graphically illustrated as follows:

141

Prepositional semantics

Pragmatics

Discourse

Different as these two conceptions of a theory of grammar are, in neither case is it possible to avoid the idea that syntax is the means by which the content of an utterance is related to its acoustic form (realistically, we must assume at least one additional representation of phonological form, but this can be left aside here). For Chomsky, syntax takes the form of an autonomous system of grammatical principles, for Givon, an iconic coding strategy for pragmatics and prepositional semantics. The definitions of the elements involved in the conceptual system of the grammatical theory differ, but the relationship between these elements is the same, with syntax as the mediator of form and content. Without turning this study into a list of recent (and old) theories of grammar, it can be stated that the same is true of other approaches, even at the extremes of, e.g. Langacker's 'Cognitive Grammar1 (e.g. Langacker 1987, cf. Fife 1990), where syntactic structure in itself is abolished, subsumed into a continuum of 'grammatical structure', which 'consists in the conventional symbolization of semantic structure (Langacker op. ciL: 76). Semantics is assumed to be the basis of the utterance, and the form of the utterance results from the hierarchical semantic relations existing between various elements of the utterance. The terminology is different, but, whether in a continuum or not, the components of the grammar exist in similar relations to each other as was sketched above. Simplistically, whether one says one wants to reconstruct 'syntax' or 'the conventional symbolization of semantic structure1, the task and the problems are the same. Newmeyer (1992) presents an interesting discussion of the relationships between generativistic and functionalistic theories of grammar, with particular attention paid to the concept of iconicity. The idea that elements of language have both a form and a content is much older than the boom in research into syntax of the past forty years. It is derived from the insight that

142 language is a system of signs, and is to be traced through the ideas of Glossematics (cf. Hjelmslev 1943; Uldall 1957), to De Saussure's signifiant and signifie (De Saussure 1980: 99 et passim), as well as through diverse philosophical traditions. But the matter in hand here is very specifically the nature of the connection between the form and content of utterances in discourse, and the position of this connection in a theory of grammar (whichever one is taken). This is conceptually subordinate to the concept of language as a whole as a system of signs, but can be pursued independently of philosophical issues about the nature of language. It can perhaps be agreed then that the reconstruction of syntax involves nothing less than the reconstruction of the mediator between form and content/meaning of an utterance. In order to achieve such a reconstruction, we would have to have access to the form and content of a corpus of utterances in the proto-language. In short, we would have to have access to the utterances themselves, which is precisely what we do not have access to in the case of Proto-languages. It would be easy to say that this problem alone is sufficient to ensure that attempts to reconstruct syntax by means of the comparative method must fail. But it is still not inconceivable that, at some future date, refinements in the methodology could overcome or by-pass even these difficulties. For that reason it is perhaps not irrelevant to pursue this line of argument further, so that as broad a spectrum as possible of the problems associated with the undertaking can be aired. An utterance implies not only a superficial, acoustic form, and a content, semantically and pragmatically defined, but also intentionality. Discourse-pragmatics and semantics present the information to be communicated. But the utterance will also of necessity be influenced by the intention of the speaker to communicate precisely this information. This intention is not necessarily pragmatically derivable from the discourse context, because, on the one hand, I can regard this information as interesting and relevant, even though the context may not justify my view. And on the other hand, my utterance can be the expression of a purely egoistic wish on my part, independent of any attempt at conforming to given conditions of a context at a particular time in a particular situation. It is to be expected that the factor of intention be built into any comprehensive method of reconstructing syntax. But how can intentionality be reconstructed? Further, assuming that syntax really is the mediator between (at least) the form and meaning of an utterance, it is then not the mediator between two closed classes of entities. When the utterance is initiated by the intention of the speaker, syntax is rather the means by which two completely new members of the class of possible forms and the class of possible meanings are brought into a systematic relationship with each other. This property of syntax to mediate between completely new entities constitutes its creative nature. How is such creativity to be incorporated into a reconstruction?

143 It may be objected that a lexical reconstruction, i.e. the reconstruction of a lexical item of the proto-language, includes the relationship between a form and a meaning anyway. The reconstruction PIE *ph2tcr 'father1 consists both of a form, expressed in terms of a series of phonemes, and a meaning, expressed as a gloss. Why should not the same be possible at the level of the sentence? Apart from the restrictions on reconstruction already mentioned, it seems that a further problem lies in the concept of 'meaning' itself. I shall not attempt to answer the question, 'What does, "What does X mean?", mean?', rigorously here. The question is a central issue of semantics, lexicography and the philosophy of language, matters far beyond the scope of this brief study. One thing may be clear, however: the question, 'What does the word X mean?', asks something quite different from the question, 'What does the sentence X mean?'. That is, the meaning of a word is something quite different from the meaning of a sentence. Whether we understand the statement as an analogy, or as an empirical statement about the nature of human knowledge and memory from a cognitive perspective, the meaning of a word is an entry in an information storage system. How this entry is internally structured, and by what mechanism it can be retrieved and put into use are further problems of semantics. But a meaning (or function in a broader sense) must be bound to a word, or be a feature of a word, such that it is part of a unit of information that can be retrieved and used, in other words, an entry in an information storage system. The full class of such entries, including elements such as derivational affixes, flexional morphemes, etc., constitutes the lexicon of a natural language. The meaning of a sentence, on the other hand, cannot be an entry in an information storage system, since sentences themselves cannot be stored in this way, but must be spontaneously produced and understood, even when a given sentence has never before been part of the linguistic experience of the speaker-hearer of the language. Whether one pursues a representational semantic theory (the meaning of the sentence is processed as a whole unit in a unitary representation), or a compositional one (the meaning of the sentence is processed On-line* as a function of its individual components), the character of the meaning of a sentence as a spontaneously created/analysed entity, and not a stored/retrieved entity, is not to be overcome. The meaning of a sentence is therefore something completely different from the meaning of a word. It is not a matter of two extreme points on a scale, from the meaning of the smallest elements to the meaning of the largest, but one of two, qualitatively different entities. It is no wonder that, while we are able to reconstruct a relationship of form and meaning of a lexical item of a proto-language, i.e. to reconstruct a word, in trying to reconstruct the relationship of form and meaning at the level of the sentence, clause, or even phrase, i.e. to reconstruct syntax, we run into serious difficulties. After all, in the latter case, we are dealing with a pair of infinite classes of entities (possible forms and possible meanings). The creation of the relationship between two individual

144 entities from these classes is simultaneous with the creation of the entities themselves, and to cap it all, the creation of all three quantities, form, meaning and the syntactically mediated relation between them, is influenced and largely determined by such intangible and, even purely synchronically, controversial factors as creativity, intentionality, and the spontaneous pragmatic conditions of the individual discourse context Under these conditions, it is rather self-evident that the reconstruction of syntax, if possible at all in any realistic sense, has to be pursued with quite different methods from those which suffice for the reconstruction of phonemes, morphemes, lexemes. What these methods may look like is, at present, quite uncertain, though there is no dearth of recent work devoted to to identifying and testing possible methods. One example of a such work has been referred to above, that is Campbell (1990). After discussing many of the problems referred to in the present chapter, and illustrating them, with data from the Uralic, particularly Balto-Finnic, languages, Campbell goes on to claim that, by being careful, we can avoid such problems, and thus effectively defuse the objections to syntactic reconstruction which are raised on that basis. In conclusion, he lists various features of reconstructed Proto-Uralic syntax (op. cit.: 87-90), which he regards as grounds 'for rejoicing over how much [syntax] is accessible to reconstruction' (op. cit.: 91). These features are: i) ii) iii) iv) v)

SOV word order, no copula in the present tense; no partitive case; no agreement within the noun phrase; no subordinating conjunctions, but rather various participial constructions to this purpose;

A little consideration suggests that these 'successes' are less a matter for rejoicing than may seem apparent at first sight. i) I have discussed the status of typological matters within the reconstruction of syntax above. SOV refers to basic word order, which, as has been stressed on a number of occasions in the present work, tends to be statistically rather infrequent in discourse. SOV is a type. It is not syntax. The syntax is the interpretative/encoding system in which the basic order is embedded as an ideal output or realisation. As a comparison of, say, Hittite, Japanese and Georgian will readily show, there are many ways a basic order SOV can be embedded in widely differing syntactic systems. ii), iii), iv) and v) all have the obvious common characteristic that they are zeroreconstructions. On the basis of comparisons of attested forms and structures in attested languages, nothing, zero is reconstructed for the proto-language. Expressing this 'reconstruction' affirmatively is possibly rather disingenuous. Basically what is being said is that we cannot reconstruct these phenomena for the proto-language, a negative proposition. Campbell seems quite at ease with the argument, Teature X cannot be

145 reconstructed for proto-language Y, therefore proto-language Υ did not have feature X'. I find this argument rather unconvincing. For one thing, the term 'proto-language' has two quite distinct meanings in the two propositions of the argument. In the first proposition, it means 1) 'the reconstruction (a theoretical construct) of an inventory of forms and features on the basis of comparison of forms and features of attested languages which are sufficiently similar to each other to suggest their genetic descent from a common ancestor'. In the second proposition, 'proto-language' means 2) 'the common ancestor of a group of genetically related languages' referred to in meaning 1). I am not persuaded that the identification of meaning 1) with meaning 2) is unproblematic. Even if it turns out to work, it can hardly be taken as given at this stage. Glossing over the problems with the term 'proto-language', the question must be asked, 'Just because we cannot reconstruct something for a proto-language, can we conclude that it was not there?' 1 fear the answer is *No'. It is a matter of record that neither passive finite verbal morphology nor a passive syntactic construction of any sort can be reconstructed for Proto-Indo-European. Did Proto-Indo-European therefore have no passive? Experience teaches that a language without a strategy for decreasing the prominence of prototypical subject arguments, or fading them out altogether, is inconceivable, as this appears to be a basic syntactic requirement in the flow of natural discourse, enabling the speaker to maintain the addressee's attention on more salient topics when these do not correspond to the prototypical subject (see section 2.1, above). Our inability to reconstruct a finite passive morpheme for the verbal system does not allow us to conclude that there was no such morpheme, nor does it allow us to reconstruct passive syntax. A natural language without a copula in the present tense, or without NP agreement, or without partitive morphology is perfectly conceivable and well attested. But in view of the situation with regard to zero-reconstruction of passive in Proto-Indo-European, what is the status of these zero-reconstructions for Proto-Uralic? It does not appear to me that these matters can be taken lightly. A further point is to be mentioned, concerning the second part of reconstruction v), the use of participial constructions for subordination in Proto-Uralic. If I reconstruct participial morphology for a proto-language, then presumably I make the tacit assumption that the participles were actually used. I have not reconstructed any syntax, only morphology, a significant result, to be sure. But to turn this round and claim that the tacit assumption that this morphology was actually used in sentences is really a significant 'success' of syntactic reconstruction seems a rather unattractive methodology. In short, these examples at least of Proto-Uralic syntactic reconstruction do not appear to me to be grounds for rejoicing. Presumably one would want to stress that these 'successes' are just a sample of the sort of 'results' that can optimistically be expected.

146 But I feel that the nature of these examples does not inspire great confidence. I find it hard to accept a statement such as, The sentence may contain a participle', as a reconstruction of syntax. One might also want to stress that for syntactic reconstruction we must lower our expectations relative to phonological and morphological reconstruction. But we must be prepared for the possibility that 'lowering our expectations' in this case means accepting that we cannot do it.

3.9 In all the preceding discussion, I have maintained, quite consciously, a rather hard line with regards both to the reconstruction of Celtic syntax and to the possibility of syntactic reconstruction at all. My aim has not been to sweep away all previous work in these fields, to prepare the way for some radical new theory. Rather I have tried to draw attention to some of the problems that are involved in such work, and attempt some explanations as to why different researchers can come up with such radically different ideas as to the developments concerned The problems are not derived solely from the sparse nature of the evidence itself, but are to a great extent conditioned by the nature of the object of investigation, syntax, and by the methodologies used. The study of the prehistory of Celtic syntax would certainly be far poorer without the many excellent studies mentioned in the preceding pages. Still, the feeling persists that what has been revealed so far is only an imperfect fragment of the invisible reality lying behind the reconstructed prehistory of Celtic syntax. The mainly formally oriented approaches applied so far to the developments are not necessarily wrong, as statements about, exactly, formal conditions. Thus there can be no question that Wackernagel's Law did operate in early forms of Celtic, and as a descriptive observation, Vendryes' restriction that the enclitics can only follow a part of the predicate is impeccable. The question remains, however, to what extent these phenomena are, historically seen, causes of change or restructuring, or symptoms of them. We know so little about the structure of early Celtic that it is hardly realistic to suppose that we can, with all confidence, claim to understand the conditions under which such phenomena arose and were distributed. It is not inconceivable that there were quite different principles operating, that are wholly inaccessible to us, or which manifest themselves in such subtle ways that they are easily overlooked in all the 'background noise' of historical development which separates Proto-Celtic from the attested Insular Celtic languages. Compare the principle of'light constituents first'discussed in Isaac (1993b: 22ff.), seen to interact with Wackernagel's Law in producing the structure of the Old Irish sentence, under a certain sort of analysis, which is by no means the only possible analysis. And the data of Continental Celtic are so meagre that, as was pointed out, they are probably consistent with a wide variety of even mutually exclusive hypotheses.

147 It is quite possible that the sort of historical argumentation presented by Vendryes, or Watkins, or Koch, based exclusively on formal factors such as Wackernagel's Law or basic order in attempting to reconstruct and explain the prehistory of Celtic syntax, are completely on the wrong track. The role of functional and pragmatic factors, that have played such a prominent role in the present study, may have been large, even decisive. Givon (1984: 204-6) draws attention to the case of Ute (a Uto-Aztecan language), where the basic word order in the sentence can be observed actually in the process of being transformed from an earlier SOV pattern to something else, at the moment a highly flexible word order for the clause. The factors involved are purely pragmatic, conditioned by precisely the phenomena of topic prominence and accessibility which formed the basis of the analysis of early Welsh word order in chapter 2. In Ute, highly inaccessible and unpredictable topics, i.e. new topics, are introduced into the discourse in preverbal position, giving rise to orders such as SV, OV, SOV (reflecting earlier basic order), while highly accessible and predictable topics, i.e. old topics drift into postverbal position, giving rise to orders such as VS, VO, VSO (in both cases with various other realisations such as SVO, OVS, VOS, depending on the relative prominence of the arguments). Such a scenario is wholly consistent with the Celtic data, as far as we understand them. In certain sorts of Celtic (Gaulish, Brittonic, Goidelic, without prejudice to the internal historical relationship of these), a system of pragmatically conditioned wordorder variation may have arisen, such that distinctions between S(O)V(O) structures and VS(O) structures were utilised to distinguish new topics with low-accessibility from old topics with high accessibility, just as the system was observed in the early poetry, and a similar one in the Middle Welsh prose, in chapter 2. That the grammars of the Celtic languages were sensitive to pragmatic factors was shown in section 2.3 by the so-called 'conjunctive' pronouns, found both in Brittonic and in Irish. Such factors would have established the order VSO as the unmarked order, in that this was the order utilised when all topics were maximally accessible and there was no need to place any particular emphasis on any one of them. The formal factors of Wackernagel's Law, Vendryes' restriction, perhaps 'light constituents first', and who knows what else, will have interacted with the pragmatics, to give rise, by complex transformations, to the pattern of Celtic syntax familiar to us from the historical languages. Gaulish might represent just such an intermediate stage as Givon (loc. cit.) claims for Ute. And Celtiberian maintains the oldest, strictly SOV pattern. None of this can be proven, now or perhaps ever. I simply air the possibility, as an example of the sort of phenomena which might be reckoned with in discussing the history and prehistory of Celtic syntax.

Section II

Editorial Practice The instances are given in such a way as to present a meaningful text, while allowing the reader in every case to reconstruct the original reading of the manuscript, except that a) the only abbreviation used in the manuscript, the rare use of the stroke " to denote n has been silently expanded, and b) the punctuation of the manuscript (indicating, but not consistently, metrical units, usually 'lines') has not been reproduced. Nor has modern punctuation been introduced, except in a few discussions of more extended passages. Otherwise all departures from the manuscript reading are shown thus: i)

Square brackets are placed around insertions, e.g. fej lovlen. Also around insertions of a word boundary when two words are written as one in the manuscript, e.g. vr/ ]un.

ii) Square brackets are also placed around emendations of individual letters in the manuscript text, e.g. rifojdnefs}, and around more radical emendations of whole words, or strings of words, e.g. en [bedin]. In cases such as these, the reading of the manuscript is given immediately following the translation of the instance, e.g. (ms. riwdrec) resp. (ms. en y bet). In cases i) and ii), notes following the instance either discuss the reading, or, where it is deemed certain or unproblematic enough, references to the justification of the emendation in the literature. iii) Round brackets are placed around letters to be omitted from the reading, e.g. cy(h)uran, i.e. ms. cyhuran, lege cyuran. Also, word division to be deleted within compounds which are written as two words in the manuscript, e.g. waei()lin, i.e. ms. waer ün, lege waetlin. Abbreviations in the glossing: dem. = demonstrative f. = feminine hab. = present habitual int = interrogative ipf. = imperfect ips. = impersonal ipv. = imperative irr. = irrealis m. = masculine neg. = negative = object pass. = passive

pi. = plural plpf. = pluperfect prs. = present prt = preterite ptl. = particle rfl. = reflexive sbj. = subjunctive sg. = singular supl. = superlative unspec. = unspecific pronominal vn. = verbal noun

149 Indicative mood is not glossed in the instances, i.e. as the unmarked mood. Verb forms formally in the third person (3sg. or 3pl.) are only glossed thus when the subject is pronominal, either explicit (sg. m. ef, L hi not attested in the text, pi. wy), or left implicit in the verbal flexion itself. When the subject is substantival, the third-person forms are strictly not marked for third person as such, but are used as default personal forms (personal forms for non-speech-act-participants), so such instances are glossed only for number, sg. or pi. The same applies to the verbs of relative clauses where the antecedent is the subject of the relative clause. Object pronouns are glossed by person and number in the first and second persons, followed by O, e.g. 2sg.O, Ipl.O. Third person object pronouns do not distinguish number, and are glossed accordingly 3.O. Where the object pronoun has a function other than to denote the direct object of a transitive verb, this is discussed in the notes. The translations given aim to reflect as accurately as possible the syntactic conditions of the original, while not straying into wildly unidiomatic English. Where my interpretation differs from that of previous commentators, this is discussed in the notes following the instance. Where my translation differs only in minor details of wording from previous ones, no comment is added. Focussed phrases are shown underlined in the translation.

The texts: A = the text of Hand A in the manuscript B = the text of Hand B in the manuscript GT = Gwarchan Tudfwlch. GA = Gwarchan Adebon. GK = Gwarchan Kynfelyn. The obscure Gwarchan Maelderw has not been included in the study.

150

Overview of instances: V:instt. 1-100=100. A: instt. 1-62 = 62. B: instt. 63-97 = 35. G A: inst. 98 = 1. GK: instt. 99-100 = 2. V[S-pro]: instt. 101-76 = 76. A: instt. 101-55 = 55. B: instt. 156-71 = 16. GT: instt. 172-73 = 2. G A: instt. 174-75 = 2. GK:inst. 176=1. V[S+pro]:noinst. [S-pro]V: instt. 177-89 = 13. A: instt. 177-84 = 8. B: instt. 185-87 = 3. GA: inst. 188 = 1. GK: inst. 189 = 1. [S+pro]V: instt. 190-204= 15. A: instt. 190-97 = 8. B: instt. 198-202 = 5. GK: instt. 203-4 = 2 V[O-pro]: instt. 205-306 = 102. A: instt. 205-83 = 79. B: instt. 284-99 = 16. GT: instt. 300-3 = 4. GK: inst. 304-6 = 3. V[O+pro]: inst. 307=1. B: inst. 307 = 1. [O-pro]V: instt. 308-35 = 28. A: instt. 308-27 = 20. B: instt. 328-34 = 7. GK: inst. 335 = 1. [O+pro]V: instt. 336-54 = 19. A: instt. 336-46 =11. B: instt. 347-52 = 6. GT: inst. 353-54 = 2. V[S-pro][O-pro]: instt. 355-74 = 20. A: instt. 355-69 = 15. B: instt. 370-74 = 5. V[S+pro][O-pro]: instt. 375-89 = 15. A: instt. 375-85 = 11. B: instt. 386-89 = 4.

[S-pro]V[O-pro]: instt. 390-403 = 14. A: instt. 390-402=13. GK: inst. 403 = 1. [S+pro]V[O-pro]: instt. 404-34 = 31. A: instt. 404-22 = 19. B: instt. 423-32 = 10. GT: inst. 433 = 1. GA: inst. 434=1. V[O-pro][S-pro]: instt. 435-53 = 19. A: instt. 435-49 = 15. B: instt. 450-52 = 3. GA: inst. 453 = 1. S[O+pro]V: instt. 454-61 = 8. A: instt. 454-60 = 7. B: inst. 461 = 1. OV[S-pro]: instt. 462-75 = 14. A: instt. 462-68 = 7. B: instt. 469-75 = 7. Encilitics in compounds: instt. 476-81 = 6. A: instt. 476-77 = 2. B: instt. 478-81 = 4. Rel.: instt: 482-539 = 58. A: instt. 482-522 = 41. B: instt. 523-36 =14. GT: inst. 537 = 1. GA: inst. 538 = 1. GK: inst. 539 = 1. Misc.: instt. 540-65 = 26. A: instt. 540-54 = 15. B: instt. 555-63 = 9. GT: inst. 564 = 1. GK: inst. 565 = 1. ?: instt. 566-97 = 32. A: instt. 566-81 = 16. B: instt. 582-90 = 9. GT: inst. 591 = 1. GA: instt. 592-93 = 2. GK: instt. 594-97 = 4. Instt. 598-603 = 6: see note 15.

151 V instt. 1-100=100

A: instt. 1-62 = 62 1. BA 1.5-6, CA 11. 11-12, AI.11-12, GOSP 115, J 11. 21-22. Gwell gwneif a thi arwawtdy uoli better do.prs. Isg. with you.sg. on song your.sg. praise.vn. Ί do better with you, praising you in song.1 2. BA 1.10, CA 1. 21, A2.1, GOSP 116, J 1. 31. Kayawc kynhorawc men yd[ ]elhei brooched in-the-van where ptl. go.sbj.prs.3sg. "Wearing a brooch in the van wherever he would go.' 2. BA 1.12, CA 1. 25, A2.5, GOSP 116, J 1. 35. ny chilyei o gamhawn neg. retreat.ipf.3sg. from battle Ήε would not retreat from battle.' 4. BA 1.14-15, CA 1. 28, A2.8, GOSP 116, J 1. 38. rac pebyll madawc pan atcoryei before tent M. when return.ipf.3sg. 'Before the tent of Madawg, when he would return.' The non-explicit subject is Madawg. 5. BA 1.16-17, CA 1. 31, A3.2, GOSP 117, J 1. 41. ruthyr eryr en ebyr pan llithywyt rush eagle in estuaries when feed.prt.ips. The rush of an eagle in estuaries when it was fed.' For an alternative interpretation of ebyr, see note 9. 6. BA 1.18, CA 1. 34, A3.5, GOSP 117, J 1. 45. rac bedin ododin odechwyt before army G. flee.prt.ips. 'Before the army of the Gododdin there was flight' 7. BA 1.20-21, CA 11. 37-38, A3.7-8, GOSP 117, J 11. 47-48. ny ellir (...)rac ergyt catvannan catwyt neg. can.prs.ips. before blow C. guard.vn. 'From the blow of Cadfannan cannot be guarded.' For the omitted material, see next instance. Jarman (loc.ciL, 80) suggests reading elM (ipf.ips.) for ellir, on the grounds that, The emendation restores the internal rhyme in -id,' meaning the rhyme with the negative particle nid, appearing as -net in the text (next instance). But the usual spelling of the

152 negative particle in Middle Welsh is nyft), suggesting a different vowel from that of the ipf.ips. ending, -it. The line is in any case quite comprehensible as it stands. & BA. 1.20, CA 1. 37, A3.7, GOSP 117, J 1. 47. anet ry vaethpwyt int.-neg. ptl. nurture.prt.ips. 'Has he not been (well) nurtured?1 This is Williams' suggestion, CA 77. Jackson and Jarman both translate with a conditional clause: Jackson, 'unless one had been well nurtured,' Jarman, 'unless he were well nurtured.1 But OW ana is clearly interrogative, ami caber bit gl. num uescitur (Juv., VVB 41). The Old Welsh form of MW o 'if (realis) was hou. The sense of the lines is that Cadfannan has had all the martial and ethical training necessary to produce a good warrior, so his blows are naturally irresistable. 9. BA 2.6, CA 1. 61, A6.5, GOSP 118, J 1. 71. ket elwynt e lanneu e benydyaw though go.sbj.prs.3pl. to churches to do-penance.vn. Though they may go to churches to do penance.' 1Q. BA 3.1-2, CA 1. 72, A8.5, GOSP 119, J 1. 82. ket elwynt e lanneu e benydu though go.sbj.prs.3pl. to churches to do-penance.vn. Though they may go to churches to do penance.' Π.· BA 3.22-4.1, CA 1. 103, A12.7, GOSP 120, J 1. 113. rac bedin ododin pan vudyd before army G. when ?.prs.3sg. GOSP: "When he moved in front of the army of Gododdin.' J: "When he stood before the army of the Gododdin.' On vudyd, see CA 100-1, Jarman 85, Geirfa 83, GPC 346. Though the meaning is unclear, the structural context makes it apparent that the verb is intransitive, and that there is no explicit subject, so a classification is nonetheless possible. Contrast instt. 570, 592, where the meanings of the verbs are also uncertain, but in addition, the contexts do not allow a decision as to transitivity, and what roles various nominal constituents play in the sentences. 12. BA 4.11, CA 1. 123, A14.3, GOSP 121, J 1. 133. crei cyrchynt harsh attack.ipf.3pl. 'Harshly they would attack.' 12. BA 4.13, CA 1. 126, A14.6, GOSP 121, J 1. 136. a chethrei a chethrawr and pierce.ipf.3sg. with spears 'And he would pierce with spears.'

153 Oncethr, pi. cethrawr < Lat. centrum < Gk. κέντρον, Geirfa 138, GPC 471, VKG 198, IEW 567. Cp. inst. 192. The whole line reads efrwygei a chethrei a chethrawr 'He would rip and pierce with spears.' 14. BA 4.13-15, CA 11. 127-28, A14.7-8, GOSP 121, J 11. 137-38. od uch lied lladei a llavnawr en gystud heyrn dur ar bennawr above ? strike.ipf.3sg. with blades in conflict irons hard on heads 'Above the ?... he would strike with blades, in the conflict of hard iron, on heads.' On lied, cf. CA 108; GOSP has no translation of this word, Jarman has 'bloodshed', but 'uncertain* (87), GPC 2139 "blood, wound, scar'. Cf. note 2 below, s.v Uad. 15. BA 4.16-17, CA 1. 131, A15.1, GOSP 121, J 1. 141. Ο vreithyell gatraeth pan adrodir from land C. ptl. report.prs.ips. ' Of the land of Catraeth is related ..." OnbKithyell, cf. CA 76, 110,1. Williams (1932c). 1& ΒΑ 5.6-7, CA 1. 149, A16.9, GOSP 122, J 1. 159. byth hwyr e techei ever slow ptl. retreat.ipf.3sg. 'He was always slow to retreat.' 17. BA 6.11-12, CA 1. 220, A19.19, GOSP 124, J 1. 230. em blaen gwyned gwanei in front G. attack.ipf.3sg. 'At the front of the men of Gwynedd he would attack.' 18. BA 6.13, CA 1. 222, A20.2, GOSP 124, J 1. 232. can yfeis because drink.prt.lsg. Because I drank.' Jackson (GOSP loc.cit.) and Jarman loc.cit., 92 read yfeist 2sg. 12. BA 6.13-14, CA 1. 222, A20.2, GOSP 124, J 1. 232. disgynneis rann fin attack.prt.lsg. part border Ί attacked in the border region.' Again, the translators emend to a 2sg. verb. 2Q. BA 6.15, CA 1. 225, A20.5, GOSP 124, J 1. 238 (B-variant). gwerth na phechut value neg. sin.ipf.2sg. 'Because you would not sin.'

154 This instance is clearly an imperfect variant of inst. 73, 1. 233 (B-variant), with thechut 'you would not flee1, cf. CA 134, GOSP loc.cit., Jarman 93. Gwerth Value* here has a function as a conjunction, 'to the value of, in return for the fact that, because' CA 79, GPC 1647. 21. BA 7.2-3, CA 1. 245, A22.3, GOSP125, J 1. 250. ny ctidolit yng kynted o ved gwirawt neg. exclude.ipf.3sg. in hall from mead liquor Ήβ was not wont to be excluded in the hall from the mead-liquor.' 22. BA 7.13, CA 1. 282, A24.6, GOSP 126, J 1. 273. a chyn edewit en rydon and before leave.prLips. in ? 'And before he was left in ?....' On various possibilities for rydon, CA 146. Jackson and Jarman both translate 'fords' without comment The denotatum of the implicit object of the impersonal verb may be eryr tith tiryon 'the eagle of graceful motion' in 1. 283. 22. BA 8.3-4, CA 1. 304, A26.5, GOSP 127, J 1. 295. keny vaket am vym am borth though-neg. nurture.sbj.ipf.ips. ?... "Though he was not nurtured ?....' The phrases governed by the preposition am 'around, about, for' are obscure. On the interpretation of the line, see inst. 76,1. 314 (B-variant). 24. BA 8.6, CA 1. 309, A26.10, GOSP 128, J 1. 300. yt ladei a llauynvreith o()eithin ptl. kill.ipf.3sg. with blade speckled fierce *He used to kill with a blood-stained, fierce blade.'

25.. BA 9.1-2, CA 1. 350, A30.8, GOSP 129, J 1. 334. ο vrynn hydwn kynn caffat from hill H. before get.prt.ips. 'From Bryn Hyddwn, before they were taken.' The implicit objects of caffat are the names listed in 11. 347-48, ewein vab eulat, gwryen, gwynn, gwryat.

26. BA 9.8, CA 1. 361, A31.9, GOSP 129, J. 1. 345. a chet lledessynt and though kill.prt.pass.3pl. 'And though they were killed.'

155

27. ΒΑ 9.17, CA 1. 376, Α33.5, GOSP 130, J 1. 360. ragorei overtake. ipf.3sg. 1 'He would overtake. 28_. ΒΑ 9.19, CA 1. 376, A33.5, GOSP 130, J 1. 360. tyllei trwy vydinawr pierce.ipf.3sg. through armies 'He would pierce through armies.' 22. BA 10.1, CA 1. 383, A34.4, GOSP 130, J 1. 367. ny thraethei na wnelei kenon kelein neg. relate.ipf.3sg. neg. make.sbj.ipf.sg. C. corpse "He would not relate that Cynon would not make a corpse.1 The clause, na wnelei kenon kelein, inst 363, is the sentential complement of thraethei. 2Q. BA 10.16-17, CA 1. 418, A37.1, GOSP 132, J 1. 402. Disgynsit entrwm ygkessevin attack.prt.3sg.abs. in stress in van Ήβ attacked in battle-stress in the van.' 21. BA 10.20, CA 1. 423, A38.1, GOSP 132, J. L 411. Disgynsit en trwm yg kesseuin attack.prt.3sg.abs. in stress in van "He attacked in battle-stress in the van.' 22. BA 11.1, CA 1. 447, A39.1, GOSP 132, J 1. 425. Disgynsit entrwm rac alauoed [djwyrein attack.prt.3sg.abs. in stress before herds east "He attacked in battle-stress before the cattle-herds of the east' (ms. wyreiri) The exact form of the line must be emended for rhyme (CA 179, 181, Jarman 105), but this does not affect the present syntactic analysis. 22. BA 11.8, CA 1. 458, A 39.12, GOSP 133, J 1. 436. ny chynhennit na bei llu idaw llawr neg. dispute.ipf.ips. neg. be.sbj.ipf.sg. host to-him low 'It was not disputed that a host would be humble(d) before him.' The double negation can be reflected in a paraphrase such as, The objection was not raised that a host would would not be humbled before him.' The meaning of the line is quite clear, but Jackson seems unsure, 'It was not disputed(?) that an army would be humbled(?) before him.' Admittedly, a more expected order in the copula clause (sentential complement of chynhennit) would have been na bei llawr idaw llu, COP + PRED + SU. But the order attested in mis instance, COP + SU +

156 PRED is attested elsewhere in the corpus, e.g. 1. 347 ys vyn tyst ewein vab eulat 'My witness is Ywain son of Eulad.' 24. BA 13.7, CA 1. 584, A48.10, GOSP not trnsl.(cf. 136), J not trnsl. yn dyvu()wyt Ipl.O come.prt.ips. "There came to us.' The pronoun yn apparently gives the goal of the verb of motion (as such therefore not a direct object). The context is unclear here and for the next instance, and the passage is a confused version of the awdl of which Williams' LI B (= ΒΑ 34.6-10, B24) is the best text, cp. instt. 65, 186, 187, 11. 606, 603-6, 593-95. Cf. CA 216-17, Isaac (1993a: 7374). 25_. BA 13.7-8, CA 1. 584, A 48.10, GOSP not trnsl.(cf. 136), J not tmsl. yn dywovu Ipl.O come.prt.3sg. *It(?) came to us.' Cf. preceding instance. It is not clear from the context what the denotatum of the implicit subject might be. 26.. BA 13.8, CA 1. 585, A48.ll. GOSP not trnsl.(cf. 136), J not trnsl. dwys yd wodyn intense ptl. desire.ipf.3pl. 'Intensely they desired.' What 'they intensely desired' is presumably to fight well. This and the following instance show focal fronting of the adjectives in adverbial function. 27. BA 13.8, CA 1. 586, A48.11, GOSP not tmsl.(cf. 136), J not tmsl. llym yt wenyn severe ptl. attack.ipf.3pl. 'Severely would they attack.' 25. BA 13.19, CA 1. 627, A51.1, GOSP 137 (emended), J 1. 526 (emended). ym adawssut Isg.O prOmise.plpf.2sg. 'You had promised me.' The pronoun ym denotes the beneficiary as an indirect obfect See CA 227. The translators adopt the B-reading am adaussut, insL 555,1. 634. The syntactic context of the present Α-reading is unclear, but the structure of the instance itself seems clear. 22- BA 13.20, CA 1. 628, A51.2, GOSP 137, J 1. 527. lladut kill.ipf.2sg. 'You would kill.1

157

Cf. inst. 69,1. 635 (B-variant). 4Q. ΒΑ 13.20, CA 1. 628, A51.2, GOSP 137, J 1. 527. llosgut burn.ipf.2sg. 'You would burn.' The instance is active (antipassive) = 'You would bum (something)', not middle = 'You would be burned'. Cf. inst. 70,1. 635 (B-variant). 41. BA 13.20-21, CA 1. 629, A51.3, GOSP 137, J 1. 528. no moryen ny waeth wnelut than M. neg. worse do.sbj.ipf.2sg. 'You would do no worse than Morien.' The modal form may denote a volitional element in the content- 'You intended to do no worse than Morien'. The comparative waeth 'worse' stands adverbially before the inflected verb. 42. BA 14.17, CA 1. 668, A56.1, GOSP 139, J 1. 533. kyuaruuant meeLprt3pl. They met.' The denotatum of the implied subject are the ketwyr 'warriors' of the preceding expression, where the latter word is an explicit argument of the verbal noun kywyrein 'arise1. The whole line might be glossed: Kywyrein ketwyr kyuaruuant arise.vn. warriors meet.prt.3pl. "The warriors arose, they met (for battle).' 42. BA 14.20-21, CA 1. 669, A56.2, GOSP 139, J 1. 534. y gyt en vn vryt yt gyrchassant together in one mind ptl. attack.prt.3pl. ' Together, with one purpose did they attack.' The the denotatum of the unexpressed subject are the ketwyr 'warriors' of the previous line. The adverbial phrases y gyt en vn vryt are fronted in focus. 44. BA 15.7, CA 1. 415, A57.9, GOSP 139, J 1. 399. mal brwyn yt gwydynt rac y adaf like rushes ptl. fall.ipf.3pl. before his hand ' Like rushes would they fall before his hand.' The adverbial phrase mal brwyn is fronted in focus.

158 45_. ΒΑ 15.15, CA 1. 689, A59.1, GOSP 140, J 1. 539. 0 winveith a medweith yd aethant y genhyn from wine-feast and mead-feast ptl. go.prt.3pl. from with-us 1 From the wine-feast and mead-feast did the go (away) from us.' Jackson and Jarman both have, 'After wine-feast..." etc. It is not clear to me which function of the preposition ο is being translated as 'after' here, cf. GPC 2605-11. I understand it purely spatially, denoting the place of origin of motion. The denotatum of the implicit pronominal subject of aethant are the llurugogyon 'mailclad ones' mentioned in the next line. The phrase O winveith a medweith is fronted in focus. 46.. BA 15.19-20, CA 1. 695, A60.1, GOSP 140, J 1. 545. Ο winveith a medveith yt gryssyassant from wine-feast and mead-feast ptl. rush.prt.3pl. ' From the wine-feast and mead-feast did they rush forth.1 For the structure, cp. preceding instance. The denotatum of the implicit subject is here the gwyr reit.., etc. 'men of battle' of the next line. 47. BA15.21, CA 1. 697, A60.3, GOSP 140, J. 1. 547. gloew dull y am drull yt gytvaethant bright array from around bowl ptl. feed-together.prt.3pl. 1 As a bright array around the bowl did they feed together.' Though on the surface a nominal phrase, 'a bright array around the bowl', the phrase gloew dull y am drull, fronted in focus, here has adverbial force, in apposition to the implicit pronominal subject 'they', this role for the phrase being coded by the particle yt. Cp. inst 78,1. 697 (B-variant). 4S. BA 16.2, CA 1. 702, A60.8, GOSP 140, J 1. 552. tru namenvn gwrnyt atcorsant alas but one man neg retum.prt.3pl. 'Alas, but for one man, they did not return.' Jackson translates, 'Alas, none came back but for one man.' Jarman, 'Alas, save one man, none returned.' These translations do not capture the syntactic relations of the original as well as they might The plural morphology of the verbal form shows that the singular phrase rumen vn gwr cannot be the subject, and there is no justification in the structure of the original for a subject 'none'. Contrast instt. 178, 184,11. 29, 694 natnen vn gwr ο gant (e)ny delhei resp. namen vn gwr ny dyuu, where indeed the phrases beginning namen are indeed the subjects of the verbal forms. 49_. BA 17.13-14, CA 1. 832, A68.2, GOSP 143, J not tmsl. baran tan teryd ban gynneuet ferocity fire blazing when kindle.prt.ips. The ferocity of a blazing fire when it has been kindled.'

159 Jarman translates the B-varianL 5J). BA 17.19, CA 1. 841, A68.ll, GOSP 143, J 1. 660. rac pebyll madawc pan atcoryet before tent M. when return.sbj.ipf.ips. 'Before the tent of Madawg when there would be a return.' CA 11. 841-42 are clearly dependent on 11. 28-29 (instt. 4, 178), or vice versa. Simon Evans notes, on the imperfect subjunctive impersonal, There are a few examples of-(h)e? (GMW 129). This is in fact the only ending attested for this category in the present corpus (see note 1). 5J. BA 18.6, CA 1. 889, A71.1, GOSP, 144, J 1. 681. Disgynnwys en affwys dra phenn descend.prt.3sg. in depth over head '?He descended headlong into the depths.' The whole awdl is obscure as to content, cf. CA 284-85, GOSP loc.cit., Jarman 128. Jarman's observation that 'attack' is the more usual meaning ofdisgynnu in this poetry is significant: might affwys 'depth' be figurative for '(the depths of) battle'? Other instances do not support this suggestion, Geitfa 15, GPC 47, but is perhaps not therefore necessarily to be rejected as a nonce poetic use of the word. Despite these uncertainties in the interpretation, the syntax of this instance is quite clear. 52. BA 18.22-19.1, CA 1. 917, A73.6, GOSP 145, J 1. 703. ry benyt ar hyt yd attawr too penance on length ptl. allow.fut.ips. Too great a penance as long as there will be people.' On the future endings -(h)awr, -(h)awt, see note 5. I interpret the line somewhat differently from the previous translators. Jackson, great penance as long as they may be allowed [to survive]'; Jarman, Too great a penance while their presence was allowed.' In both translations, 'they/their' is taken as referring to the lloegrwys lliwedawr 'the hosts of the English' in the previous line. However, I interpret attawr as meaning 'anyone, people will be allowed [by God] to live', that is, 'there will be anyone at all alive'. This meaning of gadu is quite well attested, cf. CA 290, CL1H 197-98, Geirfa 517, GPC 1367. In this way, the force of the future ending is more clearly motivated The line means that the attack referred to in the awdl exacts a price too high, ry benyt, that cannot be justified ever, as long as there are people on the earth, ar hyt yd attawr. Jarman's note to the line (130) gives, 'As long as they may be allowed to remain', but in the main translation, he inexplicably translates with a preterite. Also, his Modern Welsh transcription ydd for yd in the manuscript cannot be right: the lenition of attawr indicates that the particle is fed/ not /3Ö/ (cf. GMW 171-72).

160 52. ΒΑ 19.14-15, CA 1. 950, Α76.2, GOSP 147, J 1. 730. mynawc am rann kwynhyator generous around share lament.prs.ips. 'Generous in sharing, he is lamented.' On the interpretation, CA 78-79, 298-99. The unexpressed object of the impersonal verbal form is the Mynawc gododin of the previous line, inst. 323,1. 949. 54. BA 19.15, CA 1. 951, A76.3, GOSP 147, J 1. 731. rac eidynaryal flam nyt atcor before E. ferocity flame neg. return.prs.3sg. The ferocity of a flame before Eidyn, he does not return.' The implicit subject is the anonymous hero (or Mynawc, cf. CA 298, GOSP 146, Jarman 132-33) of the awdl. Jackson translates, 'with the fury of flame, he did not retreat in front of Eidyn.' The preterite is inexplicable here, and since the hero is presumably a warrior from Eidyn (at least immediately, whatever his ultimate origins), it is unclear why he might 'retreat in front of Eidyn'. Jarman has, 'Before Eidyn the spirit of flame will not return', interpreting aryal flam as the explicit subject of atcor, a sort of "kenning1 for the hero. This still leaves the force of rac eidyn unclear I suggest therefore that rac here has a meaning similar to that in 1. 855 rac ystre, 1. 867 rac stre, 1. 870 rac fin "before the border', that is 'defending the border'. 'Before Eidyn,' that is 'defending Eidyn', and by implication, in his defense of the British territory as a whole, or specifically, the territory of the Gododdin. In his defense of Eidyn, the hero had the ferocity of a flame, but, alas, he will not return there. 55. BA 19.16-17, CA 1. 954, A76.6, GOSP 147, J 1. 734. en aryal ar dywal disgynnwys ptl. fierce on cruel attack.prt.3sg. "Fiercely he attacked a cruel (enemy).1 I follow the translators in treating en aryal as the adverbial particle en followed by an adjective. It might be considered whether aryal here is a noun in the meaning "battle", attested by its Old Irish cognate airgal DIL 26: en is then the preposition 'in'. 56. BA 19.17, CA 1. 955, A76.7, GOSP 147, J 1. 735. can llewes because devour.prt.3sg. "Because he devoured." 52. BA 19.22, CA 1. 964, A77.8, GOSP 147, J 1. 744. pan orvyd oe gat when conquer.prs.hab.3sg. to-his battle "When he conquers in his battle." The form oe "in his' contains the preposition y 'to', cf. CA 212, GMW 53.

161 58.. ΒΑ 21.4, CA 1. 1011, Α82.6, GOSP 149, J 1. 779. ac vei anwawsnyt e[n]dew[i]t and be.sbj.ipf.3sg. coarse neg. listen.ipf.ips. 'Were he coarse, he would not have been listened to.' (ms. edewyt) For the emendation, CA 310, Jarman 136. Jackson understands it as a form of adaw 'leave, allow'. On ac vei 'if he were', CA loc.ciL: the expression also occurs in BBC 18.4, 8, LIDC 7.5, 11, but Lewis, HGCr 126, suggests reading atvei there. The occurrence in CA suggests retaining the manuscript readings. 52- BA 21.9, CA 1. 1025, A83.12, GOSP 149, J 1. 974. pan adrode[t] when report.prt.ips. 'When it was reported.' (ms. adwder) The emendation, necessary for rhyme, is quite unproblematic. 6j}. BA 22.5, CA 1. 1065, A86.ll, GOSP not trnsl., J 1. 844. ban ry godhet when ptl. enrage.sbj.ipf.ips 'Whenever he would be enraged.1 6J.. BA 22.5, CA 1. 1066, A87.12, GOSP 100 (B-variant), J 1. 845. trwm en trin a llavyn yt lladei heavy in battle with blade ptl. kill.ipf.3sg. 'Severely in battle with a blade would he kill.' The expected lenition of the verb after yt is not shown. The preposition phrase a llavyn 'with a blade' is fronted with focus. Cp. the Bvariant, inst. 67,1. 1089. 62- BA 22.13, CA 1. 1104, A87.4, GOSP 151, J. 1. 1000. gochanwn sing.ipf.lsg. Ί used to sing.' B:instt. 63-97 = 35 £3_. BA 23.10, CA 1. 649, B2.1, GOSP 99, J 1. 1. Gododin gomynnaf oth blegyt G. ask.prs.lsg. on your.sg. behalf 'Gododdin, I claim on your behalf.' The -m- in the verbal form gomynnaf ( = gouynnaf, gofynnaf) Welsh exemplar at some point.

is evidence of an Old

162

GPC 2612 regards the first element of o-blegid On behalf of as the preposition o1 'from', GPC 2605-11, but it is not clear to me what function that could have here. I suspect that this ο is from go < OW guo < *wo < PIE *upo (cf. GPC 1416), in a usage similar to that given under Ilg for Olr./o in OIL 311, e.g. mani jochlither a bruth ... foa r rfodein 'Unless his ardour be attended ... according to his own pleasure' LU 8622, fiioratarfona man ί Mr tess jfliuacht 'They got it as they liked it, both heat and cold' BCC 396.23. Plegid is from Latin placitum 'pleasure', so that a primary meaning 'under pleasure (of)' for oblegid seems appropriate. Cf. one of Williams' examples (CA 231), Eweh odymma o blegyt Duw (Ite cum gratia Dei), S. J. Williams (1929: 37). £4. BA 23.12, CA 1. 652, B2.4, GOSP 99, J 1. 4. poet gno envn tynotreissyt be.ipv.3sg. manifest in one place conquer.prs.3sg.abs. "May it be manifest in the place where it conquers.' The implicit subject is the collection of poems 'Gododdin', the guarchan mob dwywei 'the song of Dwywei's son' (= Aneirin), or 'teaching, instruction', as Olr. forcetcd. The poet requests that the poetry be heard in the place where it conquers other poems, in the bardic contest. 65_. BA 23.19, CA L 606, B3.10, GOSP not trnsl, J not trnsl. en dyuuwyt Ipl.O come.prt.ips. There came to us.' Cf. inst. 34,1. 584 (A-variant). 66- BA 24.3, CA 1. 1088, B4.11, GOSP not trnsl., J 1. 844. pan ry godet when ptl. enrage.sbj.ipf.ips. 'Whenever he would be enraged.' Cf. inst 60,1. 1065 (A-variant). 67. BA 24.3, CA 1. 1089, B4.12, GOSP 100, J 1. 845. trwm en bin a llain yt ladei heavy in battle with spear ptl. kill.ipf.3sg. 'Severely in battle with a spear would he kill.' The preposition phrase a llain is fronted in focus. Cf. inst. 60,1. 1066 (A-variant). 68. BA 24.20-21, CA 1. 495, B6.10, GOSP 101, J not trnsl. prytwyf ny bei marw mor[y]e[n] sing.sbj.prs.lsg. neg. be.sbj.ipf.sg. dead M. Ί will sing that Morien would not have died.' (ms. morem) The subjunctive has a future meaning here, cf. GMW 113.

163 £9, ΒΑ 30.13, CA 1. 635, Β7.2, GOSP 137 (Α-variant), J1. 527. lladut kill.ipf.2sg. 'You would kill.1 Cf. inst 39,1. 628 (A-variant). Tfi. BA 30.13, CA 1. 635, B7.2, GOSP 137 (Α-variant), J 1. 527. llosgut burn.ipf.2sg. 'You would burn.1 Cf. inst. 40,1. 628 (A-variant). 71. BA 30.19-20, CA 1. 1131, B8.6, GOSP 101, J 1. 870. tru namen vn gur nyt engfhjyssant alas but one man neg. escape.prt.3pl. 'Alas, but for one man, they did not escape.' On the syntactic interpretation, cf. inst 48,1. 702. 72. BA 31.1, CA 1. 1136, B9.8, GOSP 102, J 1. 878. tru nyt atcorsant alas neg. return.prt.3pl. 'Alas, they did not return.1 On the construction of this whole awdl, see inst. 529. 73_. BA 31.16-17, CA 1. 233, B12.7, GOSP 102, J 1. 238. gwerth na thechut value neg. flee.ipf.2sg. 'Because you would not flee.1 Cf. inst 20,1. 225 (A-variant). 74. BA 31.18, CA 1. 250, B13.1, GOSP 103, J 1. 255. Pan gyrchei ygkywlat when attack.ipf.3sg. in borderland "When he used to attack in the borderland.' 75_. BA 32.16, CA 1. 1148, B17.1, GOSP 105, J 1. 887. ny ystyngei rac neb neg. bow.ipf.3sg. before no-one 'He would not bow before anyone.' On the verb ystyngei and the reading and interpretation of this passage, see note 12. 7j6_. BA 33.1, CA 1. 314, B18.3, GOSP 105, J 1. 295. keny vaccet am byrth amporth though-neg. nurture.sbj.ipf.ips. ? ... 'Though he was not nurtured ?....' Cf. insL 23,1. 304 (A-variant).

164 The line (in the A and B-variants) is obscure: Jackson, 'Though he were not reared by the gates, by the gate (?)' (following the Α-variant). Jarman, 'Although his rearing was not with burden, with suffering' (following the B-variant). The verbal form vaccet (A vaket) is regarded as preterite by Jarman (187, s.v. magu), though the provection of the -g- to -k-, -cc- < -g-h-, is formally consistent with a subjunctive stem. Functionally, the use of the subjunctive in a negative concessive clause is not unexpected. Jackson's Though he were not reared ..." indicates that he recognised the form as subjunctive. Koch (1985a: 455, 1987a: 275, 278) interprets the line, in its Α-variant, quite differently, and in my view quite wrongly, as 'Because he was nourished he performed feats before the stronghold's gate.' This translation is of Koch's 'restored' line, Cent-i macet, amm-birth amm porth. Cent-i is to be OW cam, MW, Modern gan 'with, because' followed by an infixed pronoun 3sg. -/. This combination of forms is otherwise unattested: in such a construction the infixed pronoun invariably takes the -s form in Welsh, kanys, canys. Amm-birth is to be the /-preterite of an unattested verb *ymfer < *ambi-ber- with which Koch compares a likewise unattested Olr. *imm-bert. This remarkable combination of unattested forms speaks against Koch's interpretation. Something such as Jackson's or Jarman's translations is probably correct. Further on these lines, CA 152-53, Jarman 97.

77. BA 33.10, CA 1. 703, B20.1, GOSP 106, J not trnsl. Ο osgord mynydawc pan gryssyassant from retinue M. ptl. rush.prt.3pl. ' From the retinue of Mynyddawg did they rush forth.' Jackson translates, "From the retinue of Mynyddawg they hastened forth.' Although I provisionally follow this for the translation of the preposition o, l wonder if it really makes sense. In what way should we understand that 'they (the heroes) rushed forth from the retinue of Mynyddawg1? They were the retinue. I suggest that ο here may be a reflex of OWgwo < *wo in a meaning 'amongst (basic meaning 'under'), comparable to a similar usage of O/MIr./o, DIL 310, meanin Id: cf. the examples fodlcad baill an mate ...fo naemaib hErenn 'distributes the limbs of the son amongst the saints of Ireland* BColm 92.13, co dosfosnaib tuathaib uili 'It was heard amongst all the tribes' TBC

1020. We might translate Ο osgord mynydawc pan gryssyassant as 'Amongst, as members of the retinue of Mynyddawg did they rush forth'. The preposition phrase is fronted in focus, increasing the prominence of the idea that the heroes did what they did in a conceited and highly honourable manner.

165 7§. ΒΑ 33.10-11, CA 1. 704, B20.2, GOSP 106, J 1. 547 (Α-variant). gloewdull e am drullyt gynuaethant bright array from around bowl ptl. nurture.prt3pl. ' As a bright array around the bowl they (?)nurtured.' Cf. inst. 47, 1. 697 (Α-variant). The Α-reading gytvaethant 'fed together' is undoubtedly better. 79. BA 33.14, CA 1. 708, B20.6, GOSP 106, J not tmsl. tru namenvn gur nyt anghassant alas but one man neg. escape.prt.3pl. 'Alas, but for one man they did not escape.' On the syntactic interpretation, cf. inst 48,1. 702. SQ. BA 33.22-34.1, CA 1. 1176, B22.3, GOSP 106, J 1. 915. pan ladhei ae lavnawr when kill.sbj.ipf.3sg. with-his blades 'Whenever he would kill with his blades.' Jackson inexplicably omits the possessive pronoun from the translation. SI. BA 34.1, CA 1. 1177, B22.4, GOSP 106, J 1. 916. men yt welet where ptl. see.prt.ips. 'Where he was seen.' S2- BA 34.4, CA 1. 1180, B23.1, GOSP 107, J 1. 919. Pan dei y [cyuranc] when come.ipf.3sg. to battle 'When he used to come to battle.1 (ms. cyuarchant) The emendation is suggested in CA 334, and has been adopted by the translators, and by Bromwich (1978a: 153). Williams proposes it explicity for reasons of rhyme [cyuranc]ldianc. But cyuarchant would form perfectly acceptable generic rhyme with dianc, cp. CL1H 10, Rowland (1990: 417) heneintldeintlieueinc, cf. Isaac (1992: 385ff), Donovan (1975: section 3). The real reason for favouring the emendation is syntactic. Dei may be taken as da- + -ι, that is the stem da- of dyuot 'come1 followed by the imperfect ending -/ (cf. Isaac 1994b: 201-2, and note 6 below). The verbal form cyuarchant is prs.3pl. It is just conceivable that y might contain a preverbal particle y + a 3sg. infixed object pronoun y. Pan dei y cyuarchant might then be translated literally, 'When he used to come, they greet him*. However, the sequence of tenses [temporal clause: ipfj + [main clause: prs.] does not appear to be well-formed: cp. 11. 206-7 pan gryssyei gydywal (kyf)dwyreei

awr

When Cydywal used to attack, a shout used to arise' [temporal clause: ipf] + [main clause: ipf.], 1.229 pan esgynnei baub ti disgynnvt "When everyone (else) used to flee, you used to attack* [temporal clause: ipf] + [main clause: ipf.]. This sequence is restored

166 in the present instance by reading 11. 1180-81 as Pan dei y [cyuranc] nyt oed hoedyl dianc "When he used to come to battle, he was not one to escape with his life1. Jarman and Bromwich (locc.citt) adopt Williams' reading of del, sbj.prs., for dei, but this is unnecessary. 83_. BA 34.19-20, CA 1. 1195, B26.10, GOSP 107, J 1. 930. hit pan grimbuiller length when praise.sbj.prs.ips. 'So that he may be praised.' See the notes, CA 336, Jarman 146. M- BA 34.20, CA 11. 1196-97, B26.11-12, GOSP 107, J 11. 931-32. prit pan aeth can wyr y gatraeth time when go.prt.3sg. with men to C. "When he went with the men to Catraeth.' On can, CA 336. Jackson translates, '... with a hundred men ...', which must be a slip of the typewriter, as it includes the form can twice, once as the preposition 'with' and once as the cardinal Ί001. The implied subject is the hero of the avwrf/,. apparently Heinif fab Nwython. S5_. BA 35.20, CA 1. 854, B29.12, GOSP 109, J 1. 660. rac pebyll madauc pan atcorhet before tent M. when retum.sbj.ipf.ips. "Before the tent of Madawg when there was a return.' 36. BA 35.21, CA 1. 434, B30.1, GOSP 109, J 1. 412. Disgynsit intrum in alauoed dwyre[in] attack.prt.3sg.abs. in stress in herds east Ήβ attacked in battle-stress into the cattle-herds of the east.' (ms. dwyrem) This form of the line is certainly not original, cf. inst 32, and the discussion in CA 178-79, Jarman 104. But the necessary emendations do not interfere with the present syntactic analysis. S7. BA 36.7, CA 1. 880, B32.1, GOSP 1 lO(A-variant), J1. 670( A-variant). Moch aruireith iQmeiti[n] early arise.prt.3sg. in morning 'He arose in the morning.' (ms. imeitit) The emendation is suggested by the Α-variant of the line (1. 869), with y meitin, and by the next instance, 1. 861, with ι more 'in the morning' in the position of imeitit. The reading with imeitit would be syntactically and metrically (generic rhyme) acceptable, with meitit in archaic orthography for meidyd 'mid-day', but the meaning, 'He arose

167 early at mid-day', is a little strange. So meitifn] < Lai. motutina is to be preferred (cf. Vendryes 1937a). The verbal form anareith is a r-preterite of the root *reg- 'extend, stretch' IEW 85457, in the compound *are-e\s-reg- (contrast Hamp 1963, who argues that the second preverb is *-ei- < PIE *epf). Cp. the examples of this root in Old Irish verbs, VKGII 593-96, L&P 388-89. This interpretation is mooted by Williams (CA 279), but the translators have followed the Α-variant, Moch dwyreawc y meitin, with adjectival dwyreawc 'arising'. Thus, Jackson, 'Arising early in the mom', Jarman, 'Rising early in the morning'. GPC 218 treats anareith (s.v. arwyraith) as a verbal adjective < *-reg-to-, but this and anareit from the next instance are the only citations. On the other hand, comparable f-preterites of this root are found, dyr(r)eith inst. 177, 1. 1371, L1H 20.24, BT 54.24, datwyreith BT 27.17. In this light it seems preferable to treat anareith also as a r-preterite. The subject is the upparently unnamed hero of the awdl. The appearance of the adjective dwyreawc in the Α-variant is a case of replacement of a poorly understood archaic form by one more immediately comprehensible. With the archaic adverb moch 'early', cp. Olr. mo, mos- DIL 465, Lat. max, Skt. ma*suIEW747.

8g. BA36.12, CA 1. 861, B33.1, GOSP HO(A-variant), J 1. 662(A-variant). Moch aruireit i more early arise.prt.3sg. in morning 'He arose early in the morning.' See the discussion of the preceding instance. Note that in CA, the B-awdlau beginning with these two lines are printed in reverse order to that of the manuscript, following the order of their A-variants. S9_. BA 36.17, CA 1. 1223, B34.3, GOSP 110, J 1. 957. rac goduryfy aessawr godechet before noise his shield flee-prtips. 'Before the clashing of his shield there was flight.' 2Q. BA 36.17-18, CA 1. 1224, B34.4, GOSP 110, J 1. 958. techin rac eidin vre [eujriuet flee.ipf.3pl. before E. hill innumerable They would flee before the hill of Eidyn in countless numbers.' (ms. uiriuet) On the emendation, CA 341, Isaac (1992: 345-46). Jackson omits uiriuet from his translation. Jarman translates the sense of the line with a participial construction, 'Innumerable men fleeing before the hill of Eidyn'. I understand [eujriuet adverbially, They would flee innumerably ...'

168 91. ΒΑ 36.20, CA 1. 1229, B34.9, GOSP 110, J 1. 961. pan grinie[t] when press, sbj.ipf.ips. 'When he was pressed.' (ms. griniec) The interpretation as sbj.ipf. is perhaps prefereble to an interpretation as a preterite because of the sequence of tenses formed by this instance and the next, pan grinieft] griniei "When he was pressed, he would press (in battle)'. In Evans (1908: 36), 1. 20 is wrongly labelled 1. 21. 92. BA 36.20, CA 1. 1229, B34.9, GOSP 110, J 1. 961. griniei press.ipf.3sg. 'He would press.' On the lineation in BA, see preceding instance. 91. BA 36.20-21, CA 1. 1230, B34.10, GOSP 110, J 1. 962. nit atwanei neg. re-strike.ipf.3sg. 'He would not strike again.' I.e. his blows were such that once he had struck his enemy once, he did not have to strike a second time. On the lineation of BA, see inst. 91. 24. BA 36.21, CA 1. 1231, B34.ll, GOSP 110, J 1. 962. ri wanei ptl. strike.ipf.3sg. 'He used to strike again and again.1 On the use of the particle ri, ry here with the imperfect to denote repeated action in the past, cf. GMW 168, VKGII 278, L&P 257. On the similar use of Ofr. TO, GOI 342, VKGII264. On the lineation of BA, see inst. 91. 25. BA 36. 21, CA 1. 1231, B34.ll, GOSP 110, J1. 962. ri wanet ptl. strike.prt.ips. 'He has been struck (dead).' On the lineation of BA, see inst 91. 26.. BA 37.13, CA 1. 935, B37.1, GOSP 111, J 1. 721. kyuaruu ac ac()ero meet.prt.3sg. with fierce "He met with a fierce (enemy).'

169 On the meaning of kyuaruot, cf. insL 42. Note how the etymological meaning gives greater vividness to the line: he did not just 'meet' his enemy, he was (bu) together (kyu-) with him, face to face (ar-), in battle. Jackson translates rather abstractly, Ήβ met with violence', but the concrete image of the hero facing his enemy in battle, preferred here, and by Jarman, and suggested by Williams (CA 292), is more in tune with the atmosphere of the poetry. The word acero, in modern orthography, agerw, Geirfa 15, GPC 47, is from Latin acerbus. 97. BA 37.20, CA 1. 1240, B38.4, GOSP 112, J 1. 970. godolei o heit meirche gayaf bestow.ipf.3sg. from herd horses in winter 'He used to bestow (gifts of) horses from the herd in winter.' Jackson translates, Ήβ gave out presents from his herds of horses in the winter*. Jarman, Ήβ gave gifts of horses from the herd in winter'. Although the translation seems clear enough, the significance of the line appears not to have been discussed. I presume it emphasises the generosity of the hero by saying that while others may have hoarded their wealth as security in the hard winter months, this hero was wont to give away valuable property, horses, as gifts during that time. I understand e gayaf as yng ngoeqf. Although Jarman translates 'in winter', he prints y goeaf'tiie winter' in his text While this phrase might easily be understood adverbially, the article is so rare in this poetry that it is unnecessary to see it here, where an alternative reading is quite unproblematic. Williams does not comment on the expression in his notes, and it is not one of his examples of the article in the discussion in the introduction (CA Ixvii-lxviii).

GA: insL 98 = 1 2S. BA 26.10, CA 1. 1325, GA.4, GOSP not tmsl., J not trnsl. pan ry dyngir when ptl. fate.prs.ips. "When it has been fated.' On this use of the particle ry, see notes, 8, 15. Cf. Loth (1908a: 58). In Evans (1908: 26), 1. 10 is wrongly labelled 1. 12.

GK: instL99-100 = 2 22- BA 26.20, CA 1. 1342, GK.6, GOSP not trnsl., J not trnsl. kyrchessit en avon attack.prt3sg.abs. in river 'He attacked in the river.'

170 100. ΒΑ 28.2-3, CA 1. 1405, GK.69, GOSP 156, J 1. 1114. namen tri nyt atcorsant but three neg. retum.prt.3pl. 'But for three men, they did not return.' On the syntactic analysis, cf. inst. 48.

V[S-pro] instt. 101-76 = 76 A:instt. 101-55 = 55 101. BA 1.12-13, CA 11. 25-26, A2.5-6, GOSP 116, J 1. 35. eny verei waet until flow.ipf.sg. blood 'Until blood would flow.1 102. BA 1.19-20, CA 1. 36, A3.7, GOSP 117, J 1. 46. ny nodi nac ysgeth nac ysgwyt neg. protectipf.sg. nor form nor shield "Neither fair form nor shield would protect.' On the uncertain word ysgeth, cf. CA 76-77,1. Williams (1932a: 113-14). The verb appears in the singular before the two coordinated subjects (no metrical motivation). 103. BA 2.3, CA 1. 43, A4.4, GOSP 117, J 1. 53. ket dyffei wyneda goglede rann though come.sbj.ipf.sg. G. and north to share Though (the men of) Gwynedd and the North came for a share.' The translators adopt differing suggestions of Williams (CA 80). Jackson has, Though the men of Gwynedd and the North should come against his land', understanding e roam as (in modem orthography) z ei ran, i'w ran, rann in the sense of rhandir 'apportioned land'. But it is not clear to me why the poet should speak of the men of Gwynedd and the North attacking the hero's land, in poetry which generally emphasises the cooperation of all the British territories against a common enemy. Jarman has, Though the men of Gwynedd and the northern part came', which makes better sense, but involves the emendation of gogled e rann to gogled rann, 'i gael llinell naw sillaf ('to get nine syllables') as Williams puts it (loc.ciL). The metrical argument is dubious (cf. Haycock 1988: 157-58), and I prefer to adopt the lectio difficilior, taking e rann as i ran, with rann in the meaning 'share'. The meaning of the whole passage, ket dyffei wyned a gogled e rann o gussyl mob ysgyrran ysgwydawr angkyman, is that even

171 though the men of Gwynedd and the North came for a share (in the battle), at the counsel of Ysgyrran's son, (the fighting was still so difficult that) shields were shattered. The verb appears in the singular before the two coordinated subjects. 104. BA 2.7, CA 1. 49, A5.4, GOSP 117, J 1. 59. kwydei pympymwntrac y lafnawr fall.ipf.sg. five fifty before his blades "Five fifties of men would fall before his blades.' Cf. inst. 125, 1. 377. On the possible significance of the 'five fifties', cf. Isaac (1993d). 105. BA 2.7-8, CA 11. 50-51, A5.5-6, GOSP 117, J 11. 60-61. ο wyr deivyr a brennych dychiawr from men D. and B. fall.prs.sg. vgein cant e[n] diuant en un awr twenty hundred in annihilation in one hour 'Of the men of Deifr and Brennych, twenty hundred fall into oblivion in one hour.' (ms. eu) 106. BA 3.5, CA 1. 77, A9.4, GOSP 119, J 1. 87. dwys dengyn ed emledyn aergwn dense grim ptl. fight.ipf.pl. battle-hounds ' In close ranks, grimly fought the battle-hounds.' The verb grees with the plural subject (metrical motivation in the rhyme pattern dengyniemledyri). The adjectives dwys dengyn are fronted in focus, in adverbial function. 107. BA 4.7, CA 1. 117, A13.13, GOSP 120, J 1. 127. perheit y wrhyt en wrvyd last-prs.sg.abs. his valour ptl. ?long 'His valour lasts long.' Gwrvyd is hapax legomenon, cf. CA 107, GPC 1708. The meaning of the line is presumably that the fame of his valour will last long. 108. BA 4.8.9, CA 1. 119, A13.15, GOSP 121, J 1. 124. pan dyvu dutvwlch dutQnerthyd when come.prt.sg. T. people-strengthener "When Tudfwlch the strengthener of the people came.' 109. BA 4.15, CA 1. 129, A14.9, GOSP 121, J 1. 139. e mordei ystyngei adyledawr in court bow.ipf.sg. reaper 'In court the reaper humbled himself.' On mordei 'court(s)' < mawr 'great'' id "houses', cf. CA 73-74, GPC 2487.

172

On adyledawr, CA 109-10, Geirfa not registered, GPC 21 (s.v. adladd). Hapax legomenon, but cp. athledhawr, inst. 505,1. 455, CA loc.cit., 184. Cf. note 2, s.v. llad, below. 110. BA 4.15-16, CA 1. 130, A14.10, GOSP121, J 1. 140. rac erthgi erthychei vydinawr before E. groan.ipf.sg. armies 'Before Erthgi armies would groan.1 Erthychei is probably to be read erthychi (still ipf.sg.) for rhyme with erthgi, CA 110. 111. BA 5.14, CA 1. 171, A17.22, GOSP 122, J 1. 181. kwydyn gyuoedyon fall.ipf.pl. companions 'Companions would fall.' The verb agrees with the plural subject (no metrical motivation). 112. BA 6.1-2, CA 1. 197, A 18.20, GOSP 123, J 1. 207. gogyuerchi ynhon inquire.ipf.sg. heathens 'The heathens would inquire' Ynhon is lenited from gynhon (after the ipf. ending, GMW 18), and is derived from LaL gentes > MW gym, with the addition of the plural ending -on , cf. CA 127. What the heathens used to ask is specified in 11.199-200, the next instance. The singular verb precedes the plural subject (no metrical motivation). 113. BA 6.2-3, CA 11. 199-200, A18.22-23, GOSP 123, J 11. 209-10. a dyvu ο vrythonwr well no chynon int. come.prt.sg. from Britons man better than C. 'Did a man better than Cynon come from the Britons?' This question is the sentential complement of the preceding instance. 114. BA 6.5-6, CA 1. 206, A19.5, GOSP 124, J 1. 216. pan gryssyei gydywal when rush.ipf.sg. C. 'When Cydywal would rush forth.' 115. BA 6.6, CA 11. 206-7, A19.5-6, GOSP 124, J 1- 216. (kyf)dwyreei awr arise.ipf.sg. shout Ά battle cry used to arise.' The ms. form kyfdwyreei cannot be right: either kyf- or -d- must go, but both emendations give the same meaning. Williams (CA 130) is probably correct in supposing that the addition of kyf- by the scribe to dwyreei is due to anticipation of kyui five words later.

173 116. ΒΑ 6.14-15, CA 1. 224, Α20.4, GOSP 124 (emended), J 1. 234 (emended). pan disgynnei bawb when attack.ipf.sg. everyone 'When everyone used to attack.' This Α-text line, pan disgynnei bawb ti disgynnvt 'When everyone used to attack, you (too) used to attack' is certainly meaningful, but the B-variant, 1. 229, pan esgynnei baub ti disgynnvt "When everyone (else) used to flee, you used to attack' has more eulogistic force, and is adopted by the translators, as by Williams (CA 133). 117. BA 6.15, CA 1. 225, A20.5, GOSP 124, J not tmsl. ys deupo gwaeanat 3.0 come.sbj.prs.sg. joy 'May he have joy.' The infixed (better here, prefixed) object pronoun ys has here dative/benefactive function, 'May there come to him ...'. On the distribution and development of such pronouns in early Welsh sources, cf. I. Williams (1925a: 283-84). Gwaeanat, apparently 'joy', is attested only here and at L1H 206, RBP col. 1192, 1. 12.

118. BA 6.20.21, CA 1. 240, A21.6, GOSP 125, J 1. 245. ny diengis namyn tri o wrhydri fossawt neg. escape.prt.sg. but three from prowess battle 'Only three escaped by dint of prowess in battle.' 119. BA 8.6-7, CA 1. 310, A26.ll, GOSP 128, J 1. 301. val pan vel medel ar vreithin like when strike-prs.sg. reaping-party on unsettled-weather 'As when a reaping-party strikes in unsettled weather.' Jarman translates medel as 'reaper'. However, this is not justified. While the instances in GPC 2391, meaning (b), show the word in metaphorical use for a warrior, the present line appears to employ the literal meaning of 'reaping-party' (cp. Olr. meithel DIL 458), with the line as a whole as a simile for the actions of the hero in battle. Breithin is hapax legomenon, but is apparently a transparent compound of broth 'speckled' (fern.) + hin 'weather'. The sense of the simile, completed by 1. 311, insL 361, e gwnaei varchlew waet()lin "Marchlew would make blood-shed', is presumably that Marchlew made bloodshed by cutting down the enemy in a rush, just as a reaping party cuts down the crop in a rush in unsettled weather, to get it in before wind or rain have a chance to damage it 120. BA 8.9, CA 1. 322, A27.5, GOSP 128, J 1. 307. men yt(h)glawd e offer where ptl. dig.prs.sg. his instruments 'Where his weapons dig.1

174 Jarman translates metaphorically, altering the tense, 'Where his weapons struck' (Acknowledging the literal translation in his note to the line, 98). Jackson is more literal, but still changes the tense, "Where his weapons gouged'. Jarman's suggestion (loc.cit.) that the hero of the awdl, Issac, might be the subject of glawd (lenited after the particle yf) is unlikely: the verb is clodyaw, GPC 504 (s.v. cloddiaf}, which, when transitive, takes a patient as object, that which is dug (either effectum, e.g. a hole, or affectum, e.g. the earth), not an instrument, that which is dug with, such as e offer here. This instance is apparently the only example of the use of the verb in this warlike, metaphorical sense of digging holes in the enemy.

121. BA 8.10-11, CA 1. 325, A27.8, GOSP, 128, J 1. 309. seinnyessyt e gledyf ym penn mameu resound.prt.sg.abs. his sword in head mothers 'His sword resounded in the heads of mothers.'

122. BA .13-14, CA 11. 331-32, A28.5-6, GOSP 128, J 11. 315-16. s deupo car kyrd kyvnot ywlat nef ?'May .O come.sbj.prs.sg. companion songs assignation in land heaven the companion of songs have an assignation in the land of heaven.' Cf. inst. 117,1.225. Williams (CA 161) moots that kyrd might be the plural of cord 'host, band1 (cp. Olr. coin?), but a form of cerd 'craft, song1 is generally interpreted, cf. Geirfa 133. The phrase car kyrd expresses the benefactive argument also coded proleptically by the pronoun ys. The absence of any prepositional coding of the role here (such as ...;y gar kyrd ...) is archaic (cp. insL 144, 11. 1004-5). The subject kyvnoi y wlat nef is extraposed to a position in the sentence after the beneficiary, producing a structure which is comparable to the VOS order sporadically attested in the poetry, where Ο is the direct object (see below).

123. BA 8.18-19, CA 1. 341, A29.9, GOSP, 129, J 1. 325. ys deupo kynnwys yg kyman 3.O come.sbj.prs.sg. welcome in host *May he have a welcome in the host.' Cf. inst. 117,1.225.

124. BA 8.19, CA 1. 343, A30.1, GOSP 129, J 1. 327. Pan gryssyei garadawcy gat when rush.ipf.sg. C. to battle 'When Caradawg would rush forth to battle.'

175 125. ΒΑ 9.19-20, CA 1. 377, Α33.6, GOSP 130, J 1. 361. kwydei bym pymwnt rac y lavnawr fall.ipf.sg. five fifty before his blades 'Five fifties of men would fall before his blades.' Cf. inst. 104,1. 49. 126. BA 10.2-3, CA 1. 385, A34.6, GOSP 131, J 1. 369. seinnyessit e gledyf em perm garthan resound.prt.sg.abs. his sword in head rampart 'His sword resounded on the top of the rampart.' On garthan 'rampart', cf. CA 169-70, Geirfa 523, Cf. inst. 121,1. 325. 127. BA 10.3-4, CA 1. 386, A34.7, GOSP 131, J 1. 370. noc [y ]c()escyc carrec (vyr) vawr y chyhadvan than ptl. budge.prs.sg. rock great her circumference Than a rock of great circumference budges.' (ms. noc ac escyc) On the reading of the line, cf. CA 170, Jarman 102. Cyhadvan is a hapax legomenon, cf. CA loc.cit., and on advan 'border', I. Williams (1928: 139). The meaning ofcyhadvan "circumference1 is conjectural. The comparison embodied in this line is completed in the following line, inst. 234. 128,. BA 10.6-7, CA 11. 390-91, A35.3-4, GOSP 131, J 11. 374-75. ny diengis en trwm e Iwrw mynawc neg. escape.prt.sg. in stress to path lord dywal dywalach no mab ferawc fierce fiercer than son F. 'No fierce man fiercer than the son of Fferawg escaped in the battle-stress, in the manner of a lord.' Williams (CA 171) emends to engis for the metre, followed by Jarman. But this is unnecessary. The subject of diengis is the substantivised adjective phrase dywal dywalach. ... On e Iwrw, cf. CA 120 (further below). While, or because, the translation and syntax seem quite clear enough, the significance of the lines appears not to have been discussed. Absolutely literally, they say that no fierce warrior fiercer than the son of Fferawg escaped in the thick of battle, as is the appropriate manner of a lord. But this seems to suggest that the son of Fferawg i) was not so fierce, and ii) did escape, neither of which can be right in this eulogistic poetry. Presumably we must understand the lines less literally, as meaning that neither did the fierce son of Fferawg escape, nor anyone fiercer than him (because there was no one fiercer than him), this being the manner of a warrior lord; anyone who did escape was certainly less fierce, a lesser man, than him.

176 The phrase e Iwrw mynawc 'in the manner of a lord' must be taken as referring to the statement that 'no fierce man ... escaped ...' etc.: i.e. not escaping is the correct behaviour for a lord. On the use of llwrw < *lorg-, cf. GPC 2236. It might be worth considering whether llwrw in these lines could not have a more concrete meaning such as the primary one 'track': 'in the track of the lord', i.e. 'in the following/troop of the lord', as in the meaning H(b) of the cognate Olr. lorg, DIL 441, 'troop, band, company': thus, 'the fierce warrior, the son of Fferawg, and no fiercer warrior than he, escaped in the thick of battle, in the troop of the lord.' 129. BA 10.15-16, CA 1. 406, A36.9, GOSP 131, J 1. 390. pan gryssyei gynon gan wyrd wawr when rush.ipf.sg. C. with green dawn 'When Cynon would rush forth with the green dawn.' 130. BA 11.1-2, CA 1. 448, A39.1, GOSP 132, J 1. 426. wyre llu llaes ysgwydawr arise.prs.sg. host slack shields Ά host arises with lowered shields.' On llaes < Lat. laxus, cf, CA 181, GPC 2081, Jarman 105. Jackson does not translate the word, but Jarman's suggestion is adopted here, 'shields held loosely, trailing', so 'lowered shields'. The implication perhaps is that the warriors were so intent on attack that they made no effort to defend themselves. 131. BA 11.19, CA 1. 485, A41.8, GOSP 133, J 1. 455. trenghis ... bratwen perish.prt.sg. B. 'Bradwen perished.' Bratwen functions as the subject of both this and the following instance. 132. BA 11.19, CA 1. 485, A41.8, GOSP 133, J 1. 455. ny diengis bratwen neg. escape.prt.sg. B. 'Bradwen did not escape.' See preceding instance. 133. BA 13.13, CA 1. 614, A49.7, GOSP 136, J 1. 513. ys bro[w]ysbrys treul(l)yawt rys en ri[o]dre[s] cop.prs.3sg. vigorous swift consume.fut.sg. R. in ostentation ' Vigorously and swiftly will Rhys consume in ostentation.' (ms. bro ys, riwdrec) Cf. CA 223-24. In my identification of the verb, I follow Williams' suggestion of a form of treulyaw 'consume, erode'. This is also adopted by Jackson and Jarman. Rowlands (1961: 239-40) reads trewyllyawt with a form of the verb *trewyllu. This would be the only instance of the verb. A noun trewyll occurs in Marwnad Cynddylan, 1.

177 60 (Gruffydd 1982: 24) which may mean 'destructive1 (Gruffydd 'difaol'), but the verb cited by Williams (CA loc.cit.), trewyllyei, is a ghost; cf. Bromwich (21978: 148, pace Rowlands loc. cit.), where it is shown that the correct reading of the latter form is treidei. While a translation 'charge' is very tempting in this line, the basis for it, a verb trewyllu, is very flimsy. So I retain the identification of the verb as treulyaw 'consume'. Williams, Jackson and Jarman have all identified the ending as prtJsg., -awd, modern -odd, with graphic t for /o/ in this instance. On this late, analogical ending, now generalised in standard Modern Welsh and most dialects, cf. GMW 125, and note 2 below, s.v. Had. This ending is very rare in CA: the present instance is one of only four possible instances in the Gododdin (plus one in the intractable Gwarchan Maeldderw, 1. 1477 dhisgynnyawd), though preterites as such are extremely common here, cf. D. S. Evans (1978: 74, 80), and note 1 below. To this fact may be added the observation that all the other verbs in the awdl are either present indicative, Erkryn 'fears', briwant 'shatter', or present subjunctive, gaffo

'may get1, anghwy 'may escape' (next instance),

wanwy 'may strike', where the last two in particular seem to require a future interpretation. In this light, it is rather strange to find a preterite in this instance, and at that, a preterite formation which is so rare in the corpus as a whole: apart from this instance, the other three instances of -awd are restricted to a single verb, lladawd. It has been supposed that the poems are uniformly in praise of heroes who died at the 'battle of Catraeth', but this view is really an assumption. As far as the ending of treul(l)yawt goes, I follow Rowlands (loc.ciL) in construing it as an example of the future ending -(h)awt (GMW 119, and below, note 5). The awdl is in praise of a living hero, hence the present and future tenses, indicative and subjunctive. IM- BA 13.14, CA 1. 616, A49.9, GOSP, 136, J 1. 515. nyt anghwy a wanwy odiwes neg. escape.sbj.prs.sg. rel. strike.sbj.prs.sg. overtake "Whom he happens to overtake will not escape.' The subjunctive of anghwy has a clear future meaning (GMW 113), indicating that the person praised in the awdl is to be understood as still active: this is not marwnad. The subject of anghwy is the relative clause a wanwy odiwes, on which see inst. 510. 135. BA 13.19, CA 1. 627, A51.1, GOSP not trnsl. J not tmsl. Da y doeth adonwy good ptl. come.prt.sg. A. 'Well did Addonwy come.1 Jackson (GOSP 137) and Jarman (1. 526) adopt the reading of the B-variant, 1. 634 da dyuot adonwy with dyuot emended to douot 'gift1 ('fortune' Jackson). This is undoubtedly the original reading (see inst. 555). Douot having been miscopied by some scribe as dyuot, the latter was then misinterpreted by scribe A (or the scribe of one of his exemplars) as the verbal noun dyuot 'come', and was replaced by the finite doeth 'came'.

178 Despite the obvious corruption in the Α-reading in comparison with the B-reading, the instance makes sense as it stands, so it would be inconsistent to exclude it. 136. BA 14.15, CA 1. 661, A54.6, GOSP not trnsl., J not trnsl. ny chemyd haed ud a gordin neg. reconcile.prs.sg. worthy lord with oppression Ά worthy lord does not reconcile himself to oppression.1 The line is gnomic, giving a universal conceptual context for the specific statement of the following line, ny phyrth mevyl moryal efn] dilin "Mortal does not bear shame in pursuit* (inst. 444). 137. BA 16.15, CA 1. 780, A63.7, GOSP 113 (B-variant), J 1. 616 (B-variant). nyt adrawd kibno neg. report.prs.sg. K. 'Cibno does not report.' The B-variant, inst. 475,1. 788, is better. 138. BA 17.1-2, CA 1. 801, A65.2, GOSP 142, J not trnsl. dylleinw auon flood.prs.sg. river Άriverfloods.' On this line, cf. CA 269-70, Jarman 122-23. There is reason to suspect scribal interference, in an attempt to clarify the sense of a line such as in the B-variant (1. 809), ac ein bu diuant dileit aeron 'And before the bolt of Aeron was lost.1 139. BA 17.4-5, CA 11. 806-7, A65.7-8, GOSP 142, Jarman not trnsl. ny doeth en diwarth ο barth vrython neg. come.prtsg. ptl. irreproachable from region Britons ododin wr bell well no chynon O. man far better than C. 'No man of the Gododdin came irreproachably from the region of the Britons far better than Cynon.' On the mutation of vrython, see note 9. 140. BA 17.8-9, CA 11. 823-24, A66.5-6, GOSP 142, J 11. 643-44. namen ene delei ο vyt hoffeint kyndilic aeron but when-ptl. come.sbj.prs.sg. from battle delight K. A. 'But when Cynddilig of Aeron would come from the delight of battle.' The syntactic context of these lines is not immediately clear, cf. the discussion in CA 273-74, Jarman 124. Jarman translates, 'But when he returned from battle blood-stained men praised Cynddilig Aeron', taking hoffeint as a verb, ipf.3pl., with wyr enouont (the words immediately following the instance as given above) as subject, producing a VOS pattern. But I prefer the identification of hoffeint as a substantive, GPC 1886 'praise, pride, delight'. This is adopted by Jackson, 'But when he should come from battle, the

179 praise of Cynddilig of Aeron by blood-stained men', where 'the praise of Cynddilig of Aeron by blood-stained men' is a further object of mynnei in 1. 821 (inst. 266). That the present instance is also dependent on mynnei is, I think, necessary, but I do not believe that Jackson's translation is an accurate reflection of the syntactic conditions of the original. My reading is to take wyr enouanr "blood-stained men' as the further object of mynnei, with kyndilig aeron as the explicit subject of delei, and hoffeint in the meaning 'delight'. The whole passage can be given as follows (with punctuation): hu mynnei, eng kylch byt, eidol anant yr eur a meirch mawr a med medweint, namen ene delei, o vyt hoffeint, kyndilic aeron, wyr enouanL Thus he desired, all over the world, praise of minstrels in return for gold and great horses and the intoxication of the mead; but when Cynddilig of Aeron came from the delight of battle, (he desired) blood-stained men (around him).' The form and syntax of the conjunction ene are discussed by Lewis (1921), where it is suggested that the first element is the cognate of Olr. & 'when' (GOI 299, 552). This view had already been expressed in VKGII183. The view that the Old Irish form cited is the neuter sg. of the article has been revised by the identification rather with CI iom (Botorrita), i.e. originally accusative masc. sg. of the relative pronoun, cp. Lat cum 'when' < OLat. quom (cf. Schmidt 1976b: 385). This revision can apply without modification to the analysis of MW ene. 141. BA 17.15, CA 1. 834, A68.4, GOSP 143, J not trnsl. diw merchyr perideint eu calch doet Wednesday prepare.prt.pass.pl. their lime protections 'On Wednesday was prepared their lime-white armour.' Diw merchyr is fronted in focus. Williams' objection, 'Nid addas yw iddynt ymarfogi ddydd Mawrth a chael arfau ddydd Mercher!' (CA 276, 'It is not appropriate for them to arm themselves on Tuesday and get weapons on Wednesday!'), is not really a justification for regarding catch doet as a gloss on gwrym dudet 'dark-blue armour' in the previous line (they are not 'cyfystyr1 'synonymous'). The style of the poetry is more concerned with creating impressive images of war in an artistically cumulative sequence, than with producing logical, chronologically plausible narrative. Diw is in old orthography for dyw, which reflects a dative PIE *diwei, or locative PIE *diwi, cp. Olr. indiu, W he-ddiw (MW he-dyw), ET. hiziu, Co. hethew. The usual word for 'day', W dydd, Br. deiz, Co. deth (> modern jeih) reflect the nominative sg. PIE *dyeus, cp. Lat .dies. The old plural dieu (still Modem Welsh in tri-diau 'three days') reflects nominative pi. *dyewes, cp. VKG II 92, L&P 171, WG 220, Hamp (1975a,

180 where, however, diw, dyw is analysed differently). Note the use in these lines of the etymologically oblique form diw in an oblique function, On Wednesday' etc. 142. BA 19.17-18, CA 11. 956-57, A76.8-9, GOSP 147, J 11. 736-37. o osgord vynydawc ny diangwys from retinue M. neg. escape.prt.sg. namen vn aryf amdiffry[t] amdiffwys but one weapon bear.vn. intense "From the retinue of Mynyddawg there only escaped one weapon-bearing, intense warrior.' (ms. amdiffryf) On the interpretation, cf. CA 300,186-87, Jarman 133, Geirfa 21. The verb (di-)eng(h)i 'escape' usually forms the prt.3sg. in -is in the early poetry, cf. the instances in Geirfa 324,450, and the inszances in BA, note 2 below, s.v. enghi. 143. BA 20.21-22, CA 11. 1004-5, A81.9-10, GOSP 148, J 11. 772-73. ys deupo eu heneit wy wedi trinet 3.O come.sbj.prs.sg. their soul they after battle kynnwys yg wlat nef welcome in land heaven 'May their soul, after the battle, have a welcome in the land of heaven.' Cf. instt. 117, 122, 123 (11. 255, 331-32, 341), on which the present instance seems to depend. Williams writes of eu, 'drll. yw, wy, "i'w", i gael synnwyr' (CA 308, 'Read yw, wy "to their", for sense'), followed by Jarman, who prints Ys deupo i'w henaid .... This is unnecessary however: cp. inst. 122 ys deupo car kyrd, where the beneficiary is likewise expressed without preposition. 144. BA 21.3, CA 1. 1010-11, A82.5-6, GOSP 149, J 1. 778. men na lleveri lliaws where neg. speak.ipf.3sg. many 'Where many would not speak.' Williams emends to prs.3sg.abs. lleverit for reasons of rhyme, with efnjdewfijt (CA 310, followed by Jarman). But that does not clarify the rhyme scheme of whole awdl very much (-ynn, -yn, -in, -in, -i, -fijt, -awr, -if), so one should perhaps hesitate before rejecting the manuscript reading here. Jackson translates, '... did not speak', as if the form were preterite. 145. BA 21.10, CA 11. 1028-29, A84.1-2, GOSP 150, J 11. 798-99. Pan ym dyvyd lliaws pryder when lsg.Ocomeprs.hab.3sg. many anxiety 'When a host of anxieties comes upon me.' The object pronoun ym expresses the indirect object in a dative function, GMW 57.

181 146. ΒΑ 22.3-4, CA 11. 1060-61, Α86.6-7, GOSP not tmsl., J not trnsl. (B-variant, 11. 839-40) eil weith g[wel]ideint am[ lalety(g)cat veircha seirchgreulet second battle see.prtpass.pl. around A. his battle horses and arms bloody 'In the second battle were seen on the banks of the river Aled his battle-horses and bloody armour.' (ms. gehvideint) Cp. the B-variant, inst. 58,11. 1083-84.

147. BA 22.6-8, CA 11. 1069-71, A86.15-17, GOSP 100 (B-variant), J 11. 848-50. ef gwenit a dan vab ervei... ptl. strike.prs.sg.abs. under son E. vn riein a morwyn a mynawc one lady and maiden and lord 'Any lady and maiden and lord approach the son of Erfai.' On the idiom, gwan(u) a dan 'approach1, cf, CA 319, PKM 170, Geirfa 613, GPC 1572. Williams (CA loc.cit.) and Jarman (140) suggest that gwenit could could be ipf.ips., but this would leave the expression vn riein a morwyn a mynawc without syntactic function, which it has if it is the subject of gwenit, which is thereby not impersonal, but prs.3sg.abs. (singular verb with three coordinated subjects). D. S. Evans (1978: 74) gives 'Igwenif as present indicative. Jackson translates 'could approach': it is unclear what morphological and syntactic analysis underlies this. The subjects of this gwenit also function as the subjects of the same form in the following line, see next instance.

148. BA 22.7-8, CA 11. 1170-71, A86.16-17, GOSP 100 (B-variant), J 11. 849-50. ef gwenit a dan dwrch trahawc ptl. strike.prs.3sg.abs. under boar arrogant vn riein a morwyn a mynawc one lady and maiden and lord 'Any lady and maiden and lord approach the arrogant boar.' Cf. preceding instance. The image is of the warrior violent in battle but gentle and popular in court.

149. BA 22.13-14, CA 1. 1104, A87.4, GOSP 151, J 1. 1000. gochenyn wythgeith sing.ipf.pl. eight-slaves "The eight slaves would sing.1 The verb agrees with the plural subject

182 150. ΒΑ 22.14, CA 1. 1105, Α87.5, GOSP 151, J 1. 1001. pan elei dy dat ty y helya when go.sbj.ipf.sg. your.sg. father you.sg. to hunt.vn. 'When your father would go to hunt.' On the force of the imperfect here, cf. Thomson (1978: 207), the subjunctive being selected after pan for a repeated past action without reference to specific occasions, cf. VKGII323, GMW112. 111. BA22.17, CA1. 1111, A87.ll, GOSP 151, J 1. 1007. pan elei dy dat ty e vynyd when go.sbj.ipf.sg. your.sg. father you.sg. to mountain "When your father would go to the mountain.' Cf. preceding instance. 152. BA 22.21, CA 1. 1117, A87.17, GOSP 151, J 1. 1013. nyt anghei oil ny uei oradein neg. escape.ipf.sg. all neg. be.sbj.prs.sg. winged 'All that was not winged would not escape.' In Evans (1908: 22), 1. 22 is wrongly labelled 1. 21. 153. BA 22.22, CA 1. 1118, A88.1, GOSP 152, J 1. 857. Neum dodyw angkyvwngo angkyuarch ptl.-lsg.O come.pf.sg. sorrow from secrecy Ά sorrow has come stealthily upon me.' On the interpretation, cf. CA 325. Angkyvwng is hapax legomenon, Geirfa 18, GPC not registered. On the sense of the line, cp. inst. 145,11. 1028-29, and the next instances. On the lineation of Evans (1908), see preceding instance. 154. BA 22.22-23.1, CA 1. 1119, A88.2, GOSP 152, J 1. 858. nymdaw ...a uo trymach neg. come.prs.sg. rel. be.sbj.prs.sg. heavier Ά heavier one does not come upon me.' The subject is the relative clause without explicit antecedent a uo trymach, which is simultaneously the subject in the next instance. The implicit antecedent is the angkyvwng of the previous line (preceding instance.) On the lineation of Evans (1908), see inst. 152. 155. BA 23.1, CA 1. 1119, A88.2, GOSP 152, J 1. 858. nym dyvyd a uo trymach neg.-lsg.O come.prs.hab.sg. rel. be.sbj.prs.sg. heavier Ά heavier one will not come upon me.' See preceding instance. On the lineation of Evans (1908), see insL 152.

183

B: instt. 156-71 = 16 156. ΒΑ 23.13, CA 1. 654, B2.6, GOSP 99, J 1. 9. er pan aeth daearar aneirin since when go.prt.sg. earth on A. 'Since the earth covered Aneirin.' 7. BA 23.13-14, CA 1. 655, B2.7, GOSP 99, J 1. 10. nu neutysgaras nat a gododin now ptl. part.prt.sg. song with G. 'Now song (has) parted from the Gododdin.' Williams proposes to emend ysgaras to ysgarat (CA 235). The emendation is based on two factors: i) the Α-variant of the line shows anysgana in this position; ii) the emndation gives internal rhyme with not. The emendation is adopted by Jarman and apparently by Jackson. Nevertheless, ysgaru 'separate, part' can be interpreted as intransitive, cp. Olr. scaiwd DIL 524, so I follow the manuscript text 15J. BA 24.1-2, CA 11. 1083-84, B4.6-7, GOSP 100, J I. 839-40. eil w[e]ith gwelydeint am[ ]alet second battle see.prt.pass.pl. around A. y gat veirch ae seirch greulet his battle horses and-his armour bloody 'In the second battle were seen on the banks of the river Aled his battle-horses and his bloody armour.' (ms. with) Cp. the Α-variant, inst 146,11. 1960-61. 159. BA 24.4-6, CA 11. 1092-94, B4.15-17, GOSP 100, J 11. 848-50. ef gvenit a dan vab uruei... ptl. strike.prs.sg.abs. under son U. vn rieina morwyna menavc one lady and maiden and lord 'Any lady and maiden and lord approach the son of Urfai.' Cp. inst. 147,11. 1069-71 (A-variant). 160. BA 24.5-6, CA 11. 1093-94, B4.16-17, GOSP 100, J11. 849-50. ef gvenit a dan dwrch trahawc ptl. strike.prs.sg.abs. under boar arrogant vn rieina morwyna menavc one lady and maiden and lord 'Any lady and maiden and lord approach the arrogant boar.' Cp. inst. 148,11. 1170-71 (A-variant).

184 161. ΒΑ 31.13, CA 1. 229, Β12.3, GOSP 102, J 1. 234. pan esgynnei baub when flee.ipf.sg. everyone 'When everyone would flee.' Cp. inst. 116,1. 224 (A-variant). On esgynnu (< Lat. ascendo) in the sense of 'flee', cf. CA 133, Jarman 92. GPC 1247 registers this meaning, but this line is the only instance given. Geirfa 491 does not register this meaning, but treats the present instance as one of the meaning 'mount (a horse)1. Presumably the meaning of the whole line pan esgynnei baub ti disgynnvt would then be, 'When everyone (else) would mount up (to ride away, flee), you would dismount (to attack)'. But I suspect the meanings of disgynnu, well attested in the sense of 'attack', and here esgynnu 'flee', operate at a more abstract metaphorical level than this. 162. BA 33.5, CA 1. 1168, B19.3, GOSP 105, j 1. 897. pan doethan deon ο dineidin parth when come.prt.pl. nobles from D. region 'When the nobles came from the region of Dineidyn.' The verb agrees with the plural subject. 163. BA 33.18, CA 1. 1169, B21.6, GOSP 106, J 1. 908. trossassei ysgwydawr roll.plpf.sg. shields 'Shields had rolled.' Williams hesitates about the meaning (CA 332). Jarman inexplicably translates with a passive, 'Shields had been toppled'. I prefer to follow Jackson,' Shields had gone rolling'. In view of trosi 'turn (over)', I interpret the line as an image of the chaos of battle, where shields lost by the men have rolled around and come to rest in disorder. The singular verb does not agree with the plural subject. 164. BA 33.18, CA 1. 1170, B21.7, GOSP 106, J 1. 909. kwydassei lafnavr fall.plpf.sg. blades 'Blades had fallen.' The line continues the imagery of the previous instance, and has the same lack of concord between subject and verb. 165. BA 34.16, CA 11. 1186-87, B26.1-2, GOSP not trnsl., J not trnsl. Dim()guo[ui] ed[iu] o adam [h]einim come.fut.sg. ? from hand H. '?... will come from the hand of Heinif.' (ms. Dim guoiu, ledui, neinim) Cf. CA 334-35 on these difficult lines.

185 The future dimguofuij is surprising, as all the other verbs in the awdl are preterite of imperfect (except for the well-motivated present subjunctive grimbuilier, inst. 83, 1. 1195). Perhaps the scribe forgot a minim in the form, and it was originally preterite, dimguouu. But we had better let the ms. reading stand from the point of view of the number of minims. Evans (1908: 34) prints edui, but the manuscript is quite ambiguous as the correct distribution of the minims in the form, and Williams is certainly right in printing ediu (for rhyme with fh/einim). The form is hapax legomenon, apart from the entry in Dr. Davies1 dictionary (1632), 'edif, idem quod Edifar', which is spurious. The adjective edif 'greedy, rapacious' L1H 350, is inappropriate, cf. Geirfa 438, GPC 1164. Williams suggests a connection with edifedd 'heir1, but this is a variant of etißdd < *ate-di-medGPC 1252, and it is not possible to segment a base edif from it (the segmentation is addi-medd). In Isaac (1991c), I suggested reading ediu »/eöiv/ < *adimä, as a cognate of Olr. aidem 'tool, instrument' (DIL 13-14), in a meaning 'weapon'. In any case, I do not see that ediu in the text can be anything but the subject in the sentence: the verb is intransitive, so no object can be sought, and there is nothing else in the sentence which is syntactically able to be a subject For the last word of this instance, Williams prints nei rum. It is true that in the manuscript, there is slightly wider gap between the first / and the second n than between the other letters, but it is nowhere near as wide as usually separates words in this part of the manuscript 166. BA 35.1-3, CA 11. 1204-8, B26.19.23, GOSP 107, J 11. 939-43. ny wisguis im[lim]luit... no [hein]im [m]ab nuithon neg. dress.prt.sg. in-battle than H. son N. gwr a uei well man rel. be.sbj.ipf.sg. better better man than Heinif son of Nwython did not dress for battle.'

(ms. imil imil hat, neim ab) On the emendations, cf. CA 337: the manuscript shows a rather extreme confusion of minims here. The intervening lines, 1205-6, contain an epithet of the hero. 167. BA 36.5-6, CA 11. 444-45, B31.6-7, GOSP 110, J 11. 422-23. mal taranjaur] nem tar(h)ei scuytaur like thunders heaven resound.ipf.sg. shields 'Like the thunder of heaven, shields would resound.' (ms. tanm) The plural subject does not trigger agreement of the verb, and there is no metrical motivation for the form of the verb.

186

The VS construction of tar(h)ei scuytaur is clear (the -A- in tarhei is, as far as I can see, unmotivated). However, the textual and etymological analysis is difficult, See the notes, CA 180-81. The emendation to tananfaur] is apparently necessary for rhyme with scuytaur, but there is no certain guarantee that this was the original metrical form. Williams suggests reading em for nem: this would be a particle, not the pronoun Tie'. Jarman adopts this emendation, printing eftarddei sgwydawr, but Jackson translates with nem "heaven*, 'shields resounded like the thunder of heaven'. Williams further connects tar(h)ei with tardei, seen in inst. 172,1. 1267 (also followed by Jarman), but I am not sure that the meanings are compatible: 'resound' here, but 'grow' (from 'leap, spring, issue forth1) in 1. 1267. The connection suggested by Williams with try-dor 'noise', tar-on 'thunder* appears more appropriate, in which case the verb tar(h)ei would be hapax legomenon, unconnected with tardei. The root would be PIE *tor- loud' IEW 1088-89, in the reduced grade *ijS presumably in a derivative formation *tr-y6-ti. 168. BA 36.7, CA 1. 881, B32.1, GOSP 110, J 1. 671. pan of i]s cinerein imidin when rush.prs.sg. warriors in-army "When warriors rush forth in the army.' (ms. cry) The verb does not agree with the plural subject, and there is no metrical motivation for the form. 169. BA 36.18-19, CA 11. 1225-26, B34.5-6, GOSP 110, J 1. 959. meint a gaffe[ ]i[ ]lau nyt atcoryei ohanau amount rel. get.sbj.ipf.sg. his hand neg. retum.ipf.sg. from-him 'As many as his hand could get hold of, none of them would return.' (ms. gajffeilau) Jackson translates, There did not come back as much of it as a hand could grip'. Jarman, "Not as much as his hand could grip of it would return'. Jarman explicitly (in a note, 148), and Jackson implicitly, follow Willams (CA 341) in their interpretations. The sentence is understood as an SV construction: meint a gqffe[]i[]lau is construed as the subject of atcoryei. The 3sg.m. form of the conjugated preposition ohanau is construed as referring to y aessawr "his shield* in 1. 1223. So not as much of his shield as he could hold would return, i.e. he was so ardent in battle that he fought on until his shield was in splinters. Two objections to this interpretation occur to me: i) The referentiality of ohanau to v aessawr is rather tenuous: another sentence, in which completely different content is presented, intervenes between the present sentence and the reference to "his shield1, viz. 1. 1224 techin roc eidin vre [eujriuet (insL 90) They

187 would flee before the hill of Eidyn in countless numbers'. It is not quite clear to me that 'his shield' is still accessible to be referred to anaphorically by a pronominal form in the present instance. ii) It is true that there is a comparable expression in Ystorya de Carola Magno (S.J. Williams 1930: 61.16.20), nor oed gantunt oc eu taryanev tymeint ac a gudyei ei dymeu 'that they did not have as much of their shields left as their fists would cover'. However, the present instance shows the verb atcor 'return', which in this poetry appears to take exclusively animate, human arguments as subject (cf. the citations in GPC 229, and other instances in the present work). It is not clear to me that atcor could take an inanimate argument as subject, such as 'as much of it (his shield) as he could hold'. For these reasons I propose the above interpretation, which has meint a gaff>[]i[]lau as a topicalised phrase (nominativus pendens), which is the referee of the resumptive prominal in ohanau. A form of this sentence without topicalisation would read, nyt atcoryei or meint a gaffe i lau 'Nothing of as much as his hand could get hold of would return', i.e. no one of those with whom he got to grips, once he had got to grips with his enemy, the latter had no chance to return. The heroic image is the same as in 1. 1230 (inst, 93) in the same awdl, nit atwanei "He would not strike a second time', because he did not have to. Although etymologically feminine, meint < *mantt (Olr. mot) is attested early as both feminine and masculine, cf. the instances in CA 71, GPC 2323. 170. BA 37.15-16, CA 1. 939. B37.5, GOSP 111, J 1. 725. gnissint gue[lui]onar e helo neigh.ipf.pl. pale.pl. opposite his possession 'Grey horses used to neigh under his protection, (ms. gueuilon) On the idiom are helo 'under his protection', cp. Olr. ars'eilb DIL 535, CL1H 69, GPC 1208. The Old Irish equivalent shows that the cr in question is not the reflex of OW guor < *wer On, over'< *uper(-i), but rather of *on? Opposite, facing' < *prh2-i. The verb agrees with the plural adjective guefluijon 'pale, (of horses) grey', substantival as subject Metrical motivation for the agreement is not discernible. 171. BA 37.16-17, CA 1. 941, B37.7, GOSP 111, J 1. 727. ni cilius tarotrin let un ero neg. retreat.prt.sg. bull battle width one 'acre' The bull of battle did not retreat the width of one acre.' The translation 'acre1 for ero, erw is anachronistic, cf. the note in CA 2%, GPC 1238. To the references given by Williams, add Uyfr lorwerth, Wiliam (1960: 53-54, 60, 96-97) and the note to that passage (op.ciL: 119): This was a unit of variable size according to the part of Wales in question', cf. also Jenkins (1986:99, 120-21, 199). As

188 an example of the sort of measurement involved, I give the following passage from Llyfr lorwerth (Wiliam op.ciL: 60): Ac yna e gunaethant messur er eru keureythyaul urth uessur e gronyn heyd: try hyt gnmyn heyd en e uotued; teyr motued en llet e palyf; try llet y palyf en e troetued; petwar troetued un e uerryeu; vyth en y ueyyeu; deudec en e kesseylyeu; un ar pymtheg en er hyryeu. Guyalen kyhyt a'r hyryeu en llau e geylwat, a'r escurr perued y'r yeu honno en e llau ydau, a hyt e kyrhaydo ef a honno gan hyt e ureych yu deu eyryonyn: sef yv henne, llet er eru keureythyaul: a dec ar ugeynt e honn yu yhyt. 'And then they made the measurement of the legal erw by the measurement of the grain of barley: three lengths of the grain of barley to the inch; three inches to the width of the palm; three widths of the palm to the foot; four feet to the short-yoke; eight to the midyoke; twelve to the three-quarter-yoke; sixteen to the long-yoke. A rod as long as the long-yoke in the hand of the ox-caller, and the middle bar of that yoke in his other hand, and as far as he can reach with that in his outstretched arm is two edges: that is the width of the legal erw, and thirty times that is its length.' According to McCone (1991b: 52), W erw '(measurement of) field1 < *arwi. This is irregular however, cp. teirw < *tarwl, beirdd < *bardl, etc.; e cannot reflect affected *a in this position, and must reflect originl *e, cp. Co. erw and O/MBr. eru 'furrow'. The direct connection with the root *hyrh3- 'plough1 IEW 62-63, is thus excluded, and the forms are perhaps rather to be derived, as Pokomy notes, from *hjer- 'earth' IEW 332, cp. Gk. εραζε = ερασ-δε 'to earth' (e.g. II. 12.156 νιφάδες δ' ως mirrov έραζε 'as snow-flakes fall to earth', 17.619 ήριιτε δ' εξ όχέων, κατά δ' ηνία χεΰεν εραζε 'He fell from the chariot, and dropped the reins to the earth'), OHG ero (hopox legomenon, Wessobrunner Gebet, 1. 2 Dot em ni uuos noh ufhimil 'that earth was not nor heaven above', Braune/Ebbinghaus 1979: 85), Go. aiipa etc., ON jyrvi 'sand, sand-bank', Arm. ertir. This last example is doubtful, see Schmitt (1972: 24-25), with extensive references, where the root given in IEW is not included as a possibility by Schmitt; see further Frisk (1954-72:1546-47; s.v. ερα), Lehmann (1986: 18) s.v. aiipa. W erwoin 'meadowsweet' may conceivably reflect the oblique nasal stem of the root *h2erh)- in its weak form ""h^hj-w-n-, if erwoin < *arwon-yo-, but ON jprvi is evidence of a derivative *er-w-on- of *hjer-, so erwain could just as well reflect proximate *erwon-yo-. There may have been a cross of these roots in Brittonic. GT: instt. 172-73 = 2 172. BA 25.5-6, CA1. 1267, GT.10, GOSP not trnsl., J not tmsl. ar hual tres tardei galled on chain attack spring.ipf.sg. bushes 'On the chain of attack were growing bushes.'

189 The interpretation is uncertain: on tres, cf. CA 308. Cp. Olr. 1 tress *battle' OIL 606, but LHA T-136 connects this with the adjective tren 'strong1, giving the etymology *treksu-. In that case, the only relationship possible between MW tres and Olr. tress would be that the former is a borrowing of the latter. Otherwise they cannot be related. I follow Williams (CA 351) in interpreting the line to mean that "bushes, vegetation were growing on (the grave of the hero who used to be) the chain of attack1. 'Chain of attack' perhaps indicates the hero's role as a constant factor, holding the whole attack together. Note the singular verb with the plural subject, though called might be construed as collective. There is no metrical motivation for the lack of concord.

173. BA 25.22-26.1, CA 11. 1306-7, GT.49-50, GOSP 155, J 11. 1051-53. yr med a chwryf yd aethan twryf droseu hawfin for mead and beer ptl. go.prt.pl. tumult over their border 'It was for mead and beer that the tumult went over their border.' The constructio ad sensum is noteworthy, with a plural verb with morphologically singular but semantically plural subject, ίwry/'tumult of the host* < Lat. turbo. Twryf might alternatively be considered appositionally, '... they went as a tumult ...' In any case there is metrical motivation for the form here, since the singular aeth would give a three-syllable line, not unheard of in this poem, but in any case unusual: splitting up the longer lines in Williams' edition into smaller metrical units, the whole poem consists of ninety-three such units ('lines'), of which eighty-two have four syllables, ten have three, and one has five (without considering any emendations). The preposition phrase yr med a chwryf is fronted in focus.

GA:instt. 174-75 = 2

174. BA 26.8-9, CA 1. 1322, GA.l, GOSP not trnsl., J not trnsl. Ny phell gwyd aval ο avail neg. far fall.prs.sg. apple from apple-tree 'An apple does not fall far from an apple-tree.'

175. BA 26.9, CA 1. 1323, GA.2, GOSP not trnsl., J not tmsl. ny chynnyd dyual a dyvall neg. prosper.prs.sg. fierce with ? Ά fierce warrior does not prosper ?...' On dyvall, CA 359.

190 GK: inst. 176 = l 176. ΒΑ 27.8, CA 1. 1371, GK.35, GOSP 156, J 1. 1086. Dyrreith gradQvoryon run.prt.sg. step-great.pl. 'Long-striding horses galloped.'

V[S+pro] no instt.

[S-pro]V instt. 177-89=13 A: instt. 177-84 = 8 177. BA 1.15, CA 1. 29, A2.9, GOSP 116, J 1. 39. namenvn gwr o gant (e)ny delhei but one man from hundred neg. come.sbj.ipf.sg. 'Only one man in a hundred would come (back).' On the reading, cf. CA 74, Jarman 79. 178. BA 4. 17, CA 1. 132, A15.2, GOSP 121, J 1. 142. maon dychiorant hosts fall.?prs.pl. 'Hosts ?fall.' On the verbal form, cf. note 2, s.v. Idychiawr. On maon, cf. CA 110, GPC 2351. Pokorny's identification with the root *meg- T)ig, great1 IEW 708-9, maon < *magones 'die Gro en', goes back to Loth (1906: 210-11; 1923: 342) 'les grands' (so also Hamp 1994: 36). This in turn rests partly on the testimony of mediaeval vocabularies which translate the word as 'argwyddi', 'lords', see the references in CA loc.ciL But this meaning is not supported by any unambiguous instances. The meaning is in fact 'subjects, people, men, hosts; loyal people, honest men; servants' (GPC loc.cit.). GPC probably correctly connects the word with W meudwy 'hermit' < *magus dewJ 'servant of God1, cp. Olr. mug 'slave, servant', and these are properly connected with PIE *ma^os, *maghu- 'boy, youth1 IEW 696, and with the hypochoristic formation, Celt *mag(g)-u-o- > *mak(k)wo- > Olr. mace, Gaul, mapo-, W/Co./Br. mob. For Gaulish instances oftnagu-, GPN 221-22. Maon < *magones probably reflects an inherited nasal stem, descriptively PIE *magp-o, *magh-on-, with individualising or hypochoristic function (cf. Hoffmann 1955, Leumann 1977, Schmidt 1992: 47-48).

191

17g. ΒΑ 5.3-4, CA 1. 144, A16.5, GOSP 122, J 1. 154. Blaen gwirawt vragawt [ae] dybydei B. drink bragget ptl.-3.O come.ipf.sg. 'Blaen, drink of bragget used to come to him.' (ms. eft The line is quite meaningless without the emendation. The object pronoun expresses a dative (benefactive) function. The construction is a classic nominativus pendens topicalisation, the role of the topicalised constituent Blaen being coded in the clause by a coreferential resumptive pronoun, in [ae]. 180. BA 6.21-22, CA 1. 241, A21.7, GOSP 125, J. 1. 246. deu gatki aeron a chenon dayrawt two battle-hound A. and C. come.prt.sg. The two battle-hounds of Aeron and Cynon came (back).' On dayrawt, see note 15, cf. Geirfa 287-88, Lloyd-Jones (1923), GPC 878 (s.v. daeredaj). Jackson translates the form adjectivally, 'stubborn?, but the reason is unclear to me. Cf.alsoCA 137. 181. BA 9.8-9, CA 1. 362, A31.10, GOSP 129, J 1. 346. neb y eu tymhyrnyt atcorsant no-one to their lands neg. return.prt.pl. 'No one returned to their lands.' The line apparently contains a construetio ad sensum, with the formally singular subject neb and plural morphology of the verb. The plural ending of the verb is of course required for rhyme. The line can perhaps be understood with a topicalisation, 'No-one, to their lands they did not return.' 182. BA 15.16-17, CA 1. 691, A59.3, GOSP 140, J 1. 541. kyn llwyded eu lleas dydaruu before greyness their death happen.prtsg. 'Before they became grey, their death occurred.' The preverbal subject is probably motivated by the contrast expressed in the line: 'greyness' = 'ripe old age' vs. 'their death1 = 'death in youth'. The construction should perhaps be considered as giving focus on eu lleas. 183. BA 15.18-19, CA 1. 694, A59.6, GOSP 140, J 1. 544. ο drychant namen vn gwr ny dyuu from three-hundred but one man neg. come,prLsg. 'Out of three hundred, only one man came (back).'

192 184. ΒΑ 16.18-19, CA 11. 794-95, Α64.5-6, GOSP 141, J 1. 622. gwr gwned divudyawc dimyng(ye)i y gat man war vigorous rush.ipf.sg. to battle 'The vigorous man of war would rush into battle.' On divudyawc 'tireless, vigorous', cf. Geirfa 340, GPC 987. The root is PIE *l^eudh- T)e awake, awaken' IEW 150-52, also seen in W rhybuddio 'warn' < causative *pm-bflowfl-e-ye-ti

'cause to be awake, aware' (cp. Russ. probudlf). The prefix di- in

divudyawc has intensive function, di-* GPC 943. The emendation of the verbal form to dimyngi is certain for reasons of rhyme. B: instL 185-87 = 3 185. BA 23.18-19, CA 11. 603-6, B3.7-10, GOSP not trnsl., J not trnsl. llu meithlyon ... o dindywyt yn dyouu host ravaging from D. Ipl.O come.prt.sg. Ά ravaging host came to usfromDindywyd.' The lines are garbled, cf. CA 216-17,218, the following instance, Isaac (1993a), and instt. 34, 65. 186. BA 34.9, CA 11. 593-95, B24.6-8, GOSP 107, J not trnsl. lu [m]eidlyaun letQlin lu ο dindywyt en dyowu host ravaging mongrel host from D. Ipl.O come prt.sg. Ά ravaging host, a mongrel host came to us from Dindywyd.' (ms. ineidlyaun) Cf. preceding instance. 187. BA 36.3, CA 1. 439, B31.1, GOSP 109, J 1. 417. Guir gormant aethant cennin men mighty go.prt.pl. from-us 'Mighty men went from us.' On gormant CA 137, Geirfa 570, GPC 1491. The etymological suggestion of GPC is plausible: PIE *uper-mb*(i)-/nnto- (sic: the significance of the bracketed /' is unclear to me), following Lloyd-Jones (1924: 109). But *gorymant < *gorymfant would be expected, and the syncope is not regular here. Contrast Hamp (1977/78: 12-13), where a Brittonic proto-form with double prefix *uor-uxs-mant- is posited (cf. Marmo 1995: 68). The root would be PIE *men- *be very high' IEW 726. G A: inst. 188 = 1 188. BA 26.10, CA 1. 1325, GA.4, GOSP not tmsl., J not trnsl. pawb pan ry dyngir yt ball everyone when ptl. fate.prs.ips. ptl. cease.prs.sg. 'Everyone ceases when it has been fated.'

193 I.e. 'Everyone dies at the appointed hour.1 Cf. Loth (1908a: 58), CA 359. In Evans (1908: 26), 1. 10 is wrongly labelled 1. 12. GK:inst. 189=1 189. BA 27.22-28.1, CA 11. 1401-2, GK.65-66, GOSP 156, J 11. 1111-12. try()wyr a thrivgeint a thrychant y vreithyell gatraeth yd aethant three-men and three score and three hundred to land C. ptl. go.prtpl. Three men and three score and three hundred went to the land of Catraeth.' The preposition phrase giving the goal of the verb of motion is fronted in focus.

[S+pro]V instt 190-204 = 15 A: 190-97 = 8 190. BA 2.19-20, CA 1. 66, A7.3, GOSP 118, J 1. 76. wy lledi a llavnawr they kill.ipf.sg. with blades 'They used to kill with blades.' 191. BA 4.13, CA 1. 126, A14.6, GOSP 121, J 1. 136. ef rwygei... a chethrawr he rip.ipf.sg. with spears 'He used to rip with spears.' On cethr, pi. cethrawr, cf. Geirfa 138, GPC 471, inst 13. The whole line reads ef rwygei a chethrei a chethrawr 'He used to rip and pierce with spears'. 192. BA 6.15, CA 1. 224, A20.4, GOSP 124, J 1. 239. ti disgynnvt you.sg. attack.ipf.2sg. 'You would attack.' Cf. instt. 116, 161. The syntactic chiasmus in the line, VSSV, pan disgynnei bawb ti disgynnvt, is clearly motivated by the topicalising function of the initial subject in the present instance, "When everyone used to attack, you (new topic) would attack'. Cp. insL 199,1. 229 (B-variant). 193. BA 9.8, CA 1. 361, A31.9, GOSP 129, J 1. 345. wy lladassan they kill.prt.pl. 'They killed.1 Cp. inst. 198,1. 1127.

194 194. ΒΑ 10.12-13, CA 1. 400, Α36.3, GOSP 131, J 1. 384. n[eu]t ef eistedei en tal lleithic ptl. he sit.ipf.sg. in end bench 'He used to sit at the end of the bench.' (ms. nyt) Cf. CA 173, GOSP 34 on the significance. In sitting at the end of the bench at the feast, the hero was possibly in the position closest to the high table of the lord. Initial nfeujt ef seems to have a strong focussing effect here: ΉΕ (no one else) used to sit at the end of the bench'.

ΒΑ 14.1, CA 1. 633, A51.7, GOSP 137, J 1. 532. wy lledin they kill.ipf.pl. 'They used to kill.1

196. BA 14.18, CA 1. 666, A55.3, GOSP not tmsl., J not trnsl. ef lladawd a chymawna llain he kill.prt.sg. with ?club and spear 'He used to kill with? club and spear.' Cymawn is hapax legomenon, cf. CA 239. Perhaps, with Geirfa 100 (GPC 755), cp. Olr. common "heavy stick' OIL 100 .

197. BA 18.2, CA 1. 872, A70.4, GOSP 144, J 1. 673. rac cant ef gwant gesseuin before hundred he attack.pit. sg. foremost 'Before a hundred he attacked foremost' Cp. inst. 201,1. 883 (B-variant), and see note 10.

B: instt. 198-202 = 5

198. BA 30.17-18, CA 1. 1127, B8.3, GOSP 101, J 1. 867. wy ladassant they kill.prt.pl. 'They killed.' Cp. inst. 193, 1. 361. 199. BA 31.13, CA 1. 229, B12.3, GOSP 102, J 1. 234. ti disgynnvt you.sg. attack.ipf.2sg. 'You would attack.' Cp. inst. 192,1. 224 (A-variant).

195 200. ΒΑ 35.8, CA 1. 1216, Β28.5, GOSP 108, J 1. 950. gnaut rac teulu deor em discin(h)ei usual before war-band D. he attackjpf. sg. 'It was usual that before the war-band of Deor he used to attack.1 The -A- of the subjunctive is apparently without motivation. 201. BA 36.8-9, CA 1. 883, B32.4, GOSP 110, J 1. 673. rac cant em gwant ceseuin before hundred he attack.prt.sg. foremost 'Before a hundred he attacked foremost' Cp. inst. 197,1. 872, and see note 10. 202. BA 37.18, CA 1. 1237, B38.1, GOSP 112, J 1. 967. efguant tra[ ]trigant echassaf he attack.prt.sg. beyond border splendid.sup. tie attacked most splendidly beyond the border.' Jackson translates,'He stabbed over three hundred of the finest'. Jarman, 'He charged before three hundred of the finest'. On the interpretation suggested here, see note 10. GK: instt. 203-4 = 2 203. BA 26.18, CA 1. 1337, GK.l, GOSP 155, J. not trnsl. Pei mi brytwn if.irr. I compose-poetry.sbj.ipf.lsg. 'If I were to compose poetry.' 204. BA 26.18, CA 1. 1338, GK.2, GOSP 155, J 1. not trnsl. pei mi ganwn if.irr. I sing.sbj.ipf. Isg. 'If I were to sing.'

V[O-pro] instt. 205-306 = 102 A: instt. 205-83 = 79 205. BA 1.11, CA 11. 23-24, A2.3-4, GOSP 116, J 1. 33. ene klywei awr when-ptl. hear.ipf.3sg. shout "Whenever he used to hear the battle-cry.' On ene, cf. inst. 140,1. 823.

196 2 f>. ΒΑ 1.11-12, CA 1. 24, A2.4, GOSP 116, J 1. 34. ny rodei nawd meint dilynei neg. give.ipf.3sg. protection amount follow.ipf.3sg. 'He would give no quarter however many he pursued.' The sense is that however many of the enemy he pursued in battle, he never felt the urge to be lenient but always cut them down. Jarman tanslates the sense of the line, obscuring the precise syntactic form, 'He gave no quarter to as many as he pursued'. Jackson translates, "He gave no quarter where he pursued1, adopting the emendation sugested by Lewis (1950b), see. inst. 482.

207. BA 1.13, CA 1. 26, A2.6, GOSP 116, J 1. 36. mal brwyn gomynei gwyr ny[ ]t()echei like rushes hew.ipf.3sg. men neg. flee.ipf.sg. 'He would cut down like rushes the men who would not flee.' (ms. nyt echei)

(. BA 3. 11-12, CA 1. 87, A 10.4, GOSP 119, J 1. 97. gwyartlyt gwynnodynt waewawr bloody stain.ipf.3pl. spears 'Bloodily they would stain spears.' Williams reads gwynnodyt for internal rhyme with gwyarllyt (CA 95), followed by Jarman, but such rhymes, while frequent, are not obligatory, and the line is meaningful as it stands. Jackson does not adopt the emendation.

209. BA 3.16-17, CA 11. 94-95, Al 1.5-6, GOSP 120, J 11. 104-5. na phurawr eu Hain neg. purify, futips. their spears 'May their spears not be cleansed.1 I.e. of the blood of their enemies, as a monument to their heroic deeds. The ending -awr is unattested as a pure imperative, but appears to be future, see note 5. The structure as translated here is an interesting mixture, the form of the negation, na, consistent with an imperative clause, expressing the hortative or optative sense, while the future verbal form makes the wish valid for all time to come, 'may their spears never be cleansed*. Alternatively, Koch (1992b:486) suggests that na < PIE *ne-kwe here has the original function as a conjunction 'and ... not' (cp. Lat neque, CI neCue. Thus, each eu cledyuawr na phurawr eu llain 'their swords were red and their spears will not be cleansed'. This is an ingenious suggestion and makes excellent sense of the line. Parallels from other early Welsh texts would be desirable. Puraw is deadjectival 'factitive' from pur < Lat. purus.

197 210.

3.19-20, CA 1. 98,

12.2, GOSP 120, J 1. 108.

neus goreu o gadeu gewilyd ptl.-3.O make.prt.3sg. from annies shame 'He made of armies shame.' The object pronoun -s is proleptic referring to the (extraposed) object gewilyd. On the sense, cf. CA 99: when the hero had finished his work, there was nothing left of the enemy armies but dishonour. 211. BA 4.11, CA 1. 123, A14.3, GOSP 121, J 1. 133. kynnullynt reiawr collect.ipf.3pl. booty 'They used to collect booty.' On reiawr, cf. Isaac (1994b: 200-1). 212. BA 4.17-18, CA 1. 133, A15.3, GOSP 121, J 1. 143. edym diedym amygyn dir splendid most-splendid defend.ipf.3pl. land "Wonderfully splendidly they used to defend the land.1 On edym < Lat. aetemus, aetemi, cf. CA 110, Geirfa 442, GPC 1169, Lloyd-Jones (1910:468-69). Diedym contains the same word as its base, preceded by the prefix diwhich I interpret as intensive here. On the idiom Adj. + gwayw (cp. OBr. guugoiuou gl. spiculis, DGVB 204). 266. BA 17.7, CA 1. 821, A66.3, GOSP 142, J 1. 641. hu mynnei eng kych byt eidol anant ptl. desire.ipf.3sg. in circle world praise minstrels Thus he desired, all over the world, the praise of minstrels.'

210 This is syntactically clear, but the precise meanings of the words is a matter of speculation. Jackson translates, "What he desired was the acclamation of bards all round the world'. The pseudo-cleft here is not motivated by the original. Jarman translates, 'He desired around him the praise of minstrels'. In a note to eng kylch byt, Jarman (124) remarks, 'Lit "around the circuit of the world1". What I presume this expression means is that he (Cynddilig of Aeron) desired that his praise by the minstrels be heard all over the world, that he be world-famous. Oneidol, cf. CA 241, Geirfa 453, GPC 1187 (s.v. eiddol1). The etymology is quite opaque. In any case, Williams' comparisons with idolte WB 159 (irur[i]dohe gl. in fanis) and Olr. idul-Taige (recte: indidaltaigce gl. fani, Sg. 66*19, Thes. II 120) are inappropriate: these are from Lat. idolum (< Gk. εΐδωλον) and have nothing to do with the eidol in question here, cf. BBC 11.2-3 eito(e)l rhyming with reol, L1DC 4.15 also Gruffydd (1975/76: 202, 207), and Gruffydd in J. E. C. Williams et cd. (1994: 29, 36). On anant, Geirfa 26, GPC 108. Loth (1920/2la: 56) suggests connection with Olr. anair, a kind of metre: Cormac Y41, anair.i. ainm n-airchetail, is έ dan dogni ctt 'that is, the name of a poem, that is, a song the ctt poet makes'. Vendryes connects this (LEIA A73) through W anaw, Olr. anae 'wealth1 with Skt. apnas 'wealth, possession'. The root would then be PIE *Hop- 'work, bring about' IEW 780. In that case, anant would be an old participial formation. Considering OE efiian, ON efiw 'effect, make, do', a formation such as *Hp-n-nt-es > W anant '(those) making, producing' would appear to be in the right semantic range to give a meaning such as 'poets, minstrels', remembering W prydydd< *pnt-yo- (< PIE *kwer- IEW 641-42), and Gk. ττοιητης. I wonder if a viable alternative to this suggestion would be to separate anant from anaw, and take it as a reflex of */J2en/ij-ni-, a participial formation of *h2enh]-ti (e.g. Skt. dniti) "breathes1, IEW 38-39. On the connection of poetry with breathing, blowing (cp. 'inspiration'), cp. W awen, Olr. d 'muse, inspiration' < PIE *hyuhi- 'blow' IEW 81-82; Lat. votes (as IEW 1113, a Celtic loan?), b. tith, W gwawd, ON o r < PIE *fi2woh]-t-, from the same root *h2euhj-; Olr. anamain 'a type of metre1 < ^hynhjmon-i-. On this theme, cf. Thieme (1954), Watkins (1963b: 215-17), Schn rt (1967: 302). Anant would then be 'those who breathe poetry, inspiration'.

267. BA 17.14, CA 1. 833, A68.3, GOSP 143, J 1. 652 (B-variant). duw mawrth gwisgyassant eu gwrym dudet Tuesday don.prt.3pl. their blue covering 'On Tuesday they donned their blue covering (armour).' Jarman translates, 'fine covering', adopting the reading of the B-variant cein du/djet, insL 295,1. 846.

211 268. ΒΑ 17.21-22, CA 1. 858, Α69.4, GOSP 143, J not trnsl. mal twrch y tywysseist vre like boar ptl. lead.prt.2sg. hill 'Like a boar you led to the hill.' Vre here is the goal of motion, without preposition, GMW 19. It replaces as locative direct object a suppressed patient object, who or what he led. 269. BA 18.15, CA 1. 904, A72.1, GOSP 144, J 1. 690. AerQdywys ry dywys ryvel battle-lead.vn. ptl. lead.prs.3sg. war Ά battle leader, he leads the war.1 Jackson translates, 'Leader in slaughter, he led to war', Jarman, 'Battle-leader, he led to war'. Both translators thus take ryvel as a goal of a verb of motion. But I see rather a simple direct object. The line expresses the same content twice over, once in a nominal clause with an endocentric compound, aerdywys "battle leader', as predicate, and once in a verbal clause, ry dywys ryvel 'he leads the war' (present tense, apparently historical). (He): Aerdywys (is a) battle-leader. ry dywys ryvel leads the war. By these means a chiasmus is achieved.

270. BA 19.1, CA 1. 918, A73.7, GOSP 145, J 1. 704 eillt wyned klywer e arderched subject G. hear.sbj.prs.ips. his excellence The man of Gwynedd, his excellence will be heard.' On eUh 'subject, vassal1, cf. Geirfa 549, GPC 73 < Celt. *al-t-yo-s < PIE 'grow, make grow, nourish' IEW 26-27. The formation of *altyo- turns up in Old Irish as the verbal adjective (preterite participle passive) of ailid 'rears, nourishes' DIL 16, cf. LEIA A-57, VKGII 455. The adjectival base implied by the substantival derivative afderched (suffixed *-ya) is not attested in Welsh, but cp. OBr. erderh gl. evidentissimis DGVB162, Olr. can rc 'well-known, obvious, renowned' DIL 24, LEIA A-41. The root is PIE *der£- 'see' (punctual) IEW 213, cp. Skt. dadar&t Ί have seen', perf. < *dedor£-e, Gk. δέρκομαι, Olr. od-con-dorc Ί have seen' (suppletive perfect), VKG II 487 (s.v. ci-), GOI 431, 470. The sentence shows a classic nominotivus pendens construction, with topicalised eilit wyned and resumptive pronoun, here the possessive e. The motivation for the lenition of wyned is unclear.

212 271. ΒΑ 19.17, CA 1. 955, Α76.7, GOSP 147, J 1. 735. porthes mawrbwys bear.prt.3sg. great-weight 'He bore great hardship.' 272. ΒΑ 19.19-20, CA 1. 959, A77.2, GOSP 147, J 1. 739. Dyfforthyn... llawr carry.ipf.3pl. hero "They used to support the hero.1 Who 'they' might be is quite unclear from the context: presumably the other heroes. Llawr is simultaneously the object of traethyennyn, next instance. 273. BA 19.19-20, CA 1. 959, A77.2, GOSP 147, J 1. 739. traeth()y()ennyn llawr praise.ipf.3pl. hero 'They used to praise the hero.' 274. BA 19.20, CA 1. 960, A77.3, GOSP 147, J 1. 740. ry due oe lovlen glas lavnawr ptl. carry.prL3sg. to-his glove blue blades 'He has carried in his glove blue blades.' On llovlen, cf. Loth (1908a:14), CA 192, GPC 2200. It is a compound of i) llov-, the compositional form of /tow 'hand' < */ m (O)Ir. lam(h) etc. < PIE * plh2-m-eh2 (cp. Gk. ιταλάμη, Lat. palma) < *pelfi2- 'broad, flat' IEW 805-6 (cp. Hitt. palhiS 'broad'), ii) llen(n) 'a piece of some thin, flat material' (generally with some covering function), GPC 2151-52 for the various applications; OBr. lenn gl. pallae, VVB 173, DGVB 240, Olr. lenn OIL 428. A connection with PIE * pelh]- 'cover, skin, piece of clothing' IEW 803 (e.g. Lat pellis, Eng. fell 'hide', OCS pelena < *pelhi-n-) seems likely but ablaut and suffix are unclear. Jackson translates, 'He was accustomed to carry the blue blades in his hand', with ry + preterite in the function noted by Loth (loc.cit), VKGII 277, L&P 256, GMW 168, of customary repeated action in the past. But the instances are ambiguous, and nothing proves that they are not simply resultative ('perfect') the familiar function of ry + preterite. The cognate Olr. ro shows no such customary meaning with the preterite, GOI 341, EIV 97-106. 275. BA 20.4, CA 1. 970, A78.5, GOSP 147, J 1. 995 (B-variant). gweleis gwyr dullyawr see.prt.lsg. men arrayed Ί saw men arrayed (for battle).' Cf. CA 303. Dulfyawr is hapax legomenon. The B-variant (inst. 388, 1. 976) reads tyllwvr (cp. tyllwas 'mighty1 CL1H 70), for which dullyawr might simply be in error. This idea is adopted by Lloyd-Jones, Geirfa 396, but GPC 1099 registers dulliawr

213 without comment Dullyawr is morphologically transparent, < dull 'array, mustering' + -(y)awr < Lat. -ans (dissimilatory variant of -aus), not -anus as suggested by GPC 242. 276. BA 20.11, CA 1. 988, A80.1, GOSP 148, J 1. 756. carasswn disgynnu yg catraeth gessevin love.plpf.lsg. attack.vn. in C. foremost Ί should have loved to attack foremost at Catraeth.' Both translators adopt an irrealis (conditional) meaning for the pluperfect tense of carasswn. Formally, it could be realis, Ί had loved ...', but this makes less sense. The poet is commenting on heroic actions in which he took no direct part (contrast next instance). On cessevin, cf. note 10. 277. BA 20.13, CA 1. 992, A80.5, GOSP 148, J 1. 760. carasswn eil clot love.plpf.lsg. son fame Ί had loved the son of fame.' On eil 'second1 < *afyo-, in the meaning 'son, descendent', cp. 11. 389, 726; cf. PKM 213, Geirfo 547 (meaning 3: Lloyd-Jones gives the present instance as an example of meaning 2, 'like', but that makes little sense), GPC 69. Both translators adopt the irrealis meaning of the pluperfect again (see preceding instance), but the realis interpretation seems more appropriate this time: the poet had loved the hero before the latter fell in battle. 278. BA 20.13, CA 1. 992, A80.5, GOSP 148. J 1. 760. dyfforthes gwaetlin cause.prt.3sg. bloodshed 'He caused bloodshed.' On gwaetlin, cf. CA 155, Geirfo 599-60, GPC 1546 (s.v. gwaedlin1). The second element is Hin3 'flow of blood' GPC 2181, cf. Loth (1929: 145) < *1ϊηο- < PIE *//Ajno- < *leihj- 'pour, flow' IEW 664-65. According to Pokomy, the root is also attested in Olr. do-lin 'flows' OIL 241 < *-li-nu-t, VKG II566, GOI357, McCone (1991a: 13, 18, 21, 34-35). But Pedersen suggested, and now McCone has strongly argued (locc.citt.) for connection of do-lin with PIE *plehj- 'pour, flow, fill' IEW 798-800. And this opens up the question whether W Hin itself could not be < *pleh]-no-, cp. Lat. plenus, or *plehj-nu-, cp. Olr. lin (α-stem) '(full) number, complement' DIL 434. Jackson translates this instance as a relative clause modifying eil clot, The man of famous race who endured bloody slaughter' (adopting a less heroic meaning of dyfforthi). This is possible, but in the absence of any unambiguous indication that it is relative, I prefer the interpretation as an independent clause.

214 279. ΒΑ 21.3, CA 1. 1010, Α82.5, GOSP 149, J 1. 778. menit y gynghor. await.ipf.ips. his counsel His counsel would be awaited.' Menit is obscure. Cf. CA 309-10 for various possibilities. Formally, the best possibility would appear to be prs.3sg.abs. of manu, men(n)u 'prevail (upon), effect, succeed' GPC 2338 (s.v. manaf), 2429 (s.v. menafif this is the same verb as manaf). Y gynghor would then be subject, 'His counsel prevails'. The context however does not favour a present tense, but rather an imperfect. The ending is then ipf.ips., with y gynghor as object. But then manu, men(n)u no longer suits the form and meaning of the sentence: men(n)u requires a prepositional argument as complement, mennu ar rywbeth. And what does "His counsel was affected' mean? For these reasons, a verb *mynu (hapax legomenonl cf. CA 310) < *men-, may be posited, with the same root as MW anmyned, modem amynedd 'patience' (Geirfa 30, GPC 104), Olr. ainmne 'patience' DIL 20, LEIA A-37, < *anmenya < PIE *n-me/iyet)2. The root is *men- 'stay, stand still' IEW 729 (e.g. Gk. μένω, LaL moneo). Jarman notes this argument (136) but translates 'heeded'. The semantic structure and development of * anmeny (with privative *on-) > W amynedd, Olr. ainmne may be conceived somewhat as follows: 'the quality of not standing still' > 'the quality of not resting' > 'perseverance' > 'patience'. 280. BA 21.9, CA 1. 1026-27, A83.13-14, GOSP 149, J 11. 795-96. torret ergyr ο veirch a gwyr break.prt.ips. assault from horses and men 'An assault of horses and men was broken.' On ergyr 'assault, host1, cf. Geirfa 486, Lloyd-Jones (1921a: 2), GPC 1234. The comparison with OBr. ercor gl. ictum VVB 122, DGVB 162, Olr. airchor 'extension, cast, shot' DIL 23, LEIA A40 (for Ί. Williams BBCS, I 2' read 'Lloyd-Jones BBCS, I 2'), is plausible, but not unproblematic: VKG II 5, 'er-gyr (y Umlaut von o)'. But what caused the affection? Olr. airchor < *are-kor-o- is an o-stem. W ergyr could be construed as a lexicalised plural, *are-Jtor-f 'thrusts, throws', collectively > 'assault, host'. 281. BA 21.20-21, CA 1. 1051, A85.5, GOSP 150, J 1. 830. ragorei veirch racvuan run-before.ipf.3sg. horses swift 'He used to run before swift horses.' 282. BA 22.1, CA 11. 1055-56, A86.1, GOSP not trnsl., J 11. 834-35. llanwet e hual fill.prt.ips. his chain 'His chain was filled.' Cf. inst 286,11. 1078-79 (B-variant). For context and interpretation, see note 11.

215 283. ΒΑ 23.1-2, CA 11. 1120-21, Α83.3-4, GOSP 152, J 1. 859. ny magwyt yn neuad a vei lewach noc ef neg. rear.prt.ips. in hall rel be.sbj.prs.sg. bolder than he 'No one bolder than he was reared in a hall.1 B:instt. 284-99=16 284. BA 23.12-13, CA 1. 653, B2.5, GOSP 99, J 1. 8. er pan want maws mvrtrin since when strike.prt.3sg. pleasant wall battle 'Since he struck down the pleasant one, the wall of battle.' The sentence is syntactically well-formed as it stands. However, since there is no constituent in the surrounding lines which could conceivably be coreferent with the implicit subject (the 'he* of the translation), the emendation to the impersonal wanet (CA 234) is probably justified. The syntactic pattern remains unaltered, V[O-pro]. On maws 'pleasant, delightful, tender', cf. CA loc.cit, GPC 2388, Loth (1906: 21213). 285. BA 23.16-17, CA 11. 599-600, B3.3-4, GOSP not trad., J not trnsl. ry due diwyll ο win bebyll ar lies tymyr ptl. bring.prt.3sg. luxury from wine tent on benefit country 'He has brought luxury from the wine-tent for the sake of the country.' On the function of ry here, cf. inst. 274,1. 960. 286. BA 23.21, CA 11. 1078-79, B4.1-2, GOSP not trnsl., J not trnsl. llanwet y h(a)ual fill.prtips. his chain 'His chain was filled.' Cf. insL 282,11. 1055-56 (Α-variant), note 11. 287. BA 24.9, CA 11. 1098-99, B4.21-22, GOSP 100, J not trad. dygyrchet y getae glot seek.prt.ips. his gift and-his fame 'His gift and his fame were sought' Jackson translates impressionistically, "His gifts and his fame brought visiting throngs'. The Α-text (adopted by Jarman) has the adjective digythrud 'assured* in the position of the verb dygyrchet, and the former appears to fit the metre better, from the point of view of rhyme. But the instance is quite meaningful as it stands. . BA 30.13-14, CA 1. 636, B7.3, GOSP 101, J 1. 529. ny chetweist naceithafna chynnor neg. keep.prt.2sg. nor wing nor van 'You kept neither the wing nor the van.' Cf. inst. 247,1. 630 (A-variant).

216 289. ΒΑ 30.14-15, CA 11. 638-39, Β7.5-6, GOSP not trnsl., J not trnsl. ny weleis ... marchauc a vei waeth no d ginne^. see.prt.lsg. horseman rel. be.sbj.ipf.sg. worse ?than ? ?man Ί did not see a horseman worse ?than ... ?' Cf. inst. 248,1. 632 (Α-variant), CA 230. Between weleis and marchauc appears the phrase or mor bwyr mor, ostensibly 'from the sea to the sea'. This this is almost certainly a garbled copy of something like emorchwyd in the A-variant. The final expression of this instance is quite obscure. Cf. Koch (1985/86: 57), where it is suggested that mor in or mor and nood reflect archaic orthographic renderings of mawr 'great* and nawd 'protection', < Proto-W *mpr, *ngd. 290. BA 31.4-5, CA 1. 1144, B10.5, GOSP 102, J 1. 883. er[y]vessit gwin gwydyr lestri llavn drink.prt3sg.abs. wine glass vessels full 'He drank the wine of full glass vessels.1 291. BA 32.18, CA 1. 1151, B17.4, GOSP 105, J 1. 890. gwyryavr nein gwaewawr kelin creudei battle sow.ipf.3pl. spears holly bloody In battle they would cast bloody holly spears.' Williams suggests that gwryavr is a gloss on the preceding kyndor our (CA 329), and Jarman omits it accordingly. This is possible, but Jackson retains the manuscript reading ('in strife ...'), as I do. On gwryavr, cf. CA 81-82, Geirfa 710, GPC 1709 (s.v. gwriawr). Creudei (Geirfa 174) is morphologically obscure, but must be a derivative of creu 'blood, gore'. Koch (1991: 7) treats gwryavr as the plural of an adjectival formation, with the meaning "heroic men1, functioning as subject ofhein. His justification (op.ciL: 39, n. 11) is weak, and until such a formation can be independently, internally motivated in Welsh, I retain the interpretation given above. 292. BA 32.18-19, CA 1. 1152, B17.5, GOSP 105, J I. 891. pan wanet yg kyueillt when strike.prt.ips. my comrade 'When my comrade was struck down.' On kyueillt 'friend, comrade', cf. Geirfa 206-7, GPC 675 (s.v. cyfaill). It is conceivable that in this archaic, heroic poetry, the more exact, etymological meaning of 'foster-brother* could be identified, as in the cognate Olr. comalta DIL 135, LEIA C-165 < *kom-al-t-yo- One reared together with someone1 < PIE *h2fl- 'rear, nourish' IEW 26-27.

217 293- ΒΑ 32.20-21, CA 1. 1155, Β17.8, GOSP 105, J 1. 894. pan dyduc cy(h)uran clotuan mordei when take.prt.3sg. portion renowned court 'When he took the portion of the renowned one of the court.' On the'champion's portion', cp. Olr. cuwdmirOlL 103 (s.v. caur), Jackson (1964: 21-22). Cyuran < *kom-ranna (Olr. comronn DIL 143, LEIA C-182) < *rasna < PIE *prh2-sneh2 < *perh2- 'divide, share1 IEW 817, cf. De Bernardo-Stempel (1987: 42, 135-36), cp. CI ?arsna.< (Botorrita), Eichner (1989: 44), contrast Meid (1993: 87). 294. ΒΑ 35.9-10, CA 1. 1218, B28.7, GOSP 108, J 1. 952. indeuit a lauarei heed.ipf.ips. rel say.ipf.3sg. 'What he would say would be heeded.' The relative clause without explicit antecedent α lauarei 'what he would say1 is the sentential object of the impersonal indeuit. The context of this clause shows a very interesting structure: gnaut mab golistan cen n(e)i bei guledic i tat usual son G. though neg. be.sbj.ipf.sg. lord his father indeuit a lauarei heed.ipf.ips. rel say.ipf.3sg. 'It was usual that thougt the son of Golystan would say would be heeded.' This whole assertion is a nominal predication without copula: gnaut 'usual1 is the predicate, the assertion indeuit a lauarei 'what he would say would be heeded1 is the subject, while the concessive clause cen n(e)i bei guledic i tat 'though his father was not a lord' modifies the main assertion. Mab golistan 'the son of Golystan1 is the main topic of the whole assertion, and appears topicalised as a nominativus pendens, left dislocated preceding the the two sub-clauses in which it plays a role. As such it is coreferent with two resumptive constituents, fulfilling two different syntactic functions, in the subsequent clauses: the possessive pronoun ; Tiis' in the concessive clause, and the zeroanaphoric subject of lauarei in the relative clause. The syntactic strands of this complex assertion may be illustrated schematically as in the following diagram. The arguments, Arg., are variously subject or object, depending on their grammatical properties This tree does not represent a syntactic phrase-structure diagram in the sense of generative grammatical theory, but illustrates the informational and propositional structure of the sentence. A high degree of correlation between syntactic structure and propositional structure can be assumed, but terminologically the two should be kept distinct.

218 Assertion 1 Pred.

neg.

Pred.

Arg.

Pred.

Arg. rel.

gnaut

cen

ni

bei guledic

Δ

i, tat

indem a

Assertion 4 Pred.

Arg.

lauarei

0j

Topic: mob golistani

295. BA 35.14, CA 1. 846, B29.4, GOSP 109, J 1. 652. diu maurth guisgassant eu ceindu[d]et Tuesday don.prt.3pl. their fine covering 'On Tuesday they donned their fine converings.' (ms. duhef) Jackson translates du[d]et as 'armour', but it need not be so specific: garments of some sort are meant, but perhaps only in general terms. 296. BA 36.16, CA 1. 1221, B34.1, GOSP 110, J 1. 955. Guelet e lauanaur en liwet see.prtips. his blades in host 'His blades were seen in the host* On lliwet, cf. CA 84, GPC 2193. Williams argues convincingly (CL1H 95-96, on kyuetliw) that lliwet is derived from lliw2 'accusation that someone is a thief seen with the stolen property' GPC loc.cit, perhaps by means of a semantic progression such as 'act of seeing someone in public' > 'the public, crowd itself. Pokomy (1EW 650) derives this lliw, in its derivative edliw 'scold, accuse' (< *aie-//w-), and its Old Irish cognate ttuf 'accuses1 (< *leyeti) from *la-, *le- 'some sort of noise'. But this seems inconclusive. In view of the importance of sight in the Celtic legal applications of lliw and

219 its Old Irish cognate, might it not be a highly specialised usage (in a way admittedly difficult to define) of Iliw 'colour1, Olr. U (IEW 965 s.v. *

297. BA 38.2-3, CA 11. 810-11, B39.3-4, GOSP 113, J 1. 630. riuesit i loflenar pen erir(h)on luit esteem.plpf.ips. his glove before head eagles grey ΉΪ5 glove (hand) had been esteemed before the grey eagles.' This interpretation differs from those of previous commentators. The motivation for the difference is the attempt to retain the manuscript reading. Following Williams' suggestion (CA 272), the translators have omitted or, and understood riuesit as prt.3sg.abs., with pen erir(h)on luit 'the heads (beaks) of the grey eagles' as subject and i loflen as object, giving a VOS structure. Jackson, The beaks of eagles praised his hand' (omitting luit without comment), Jarman, The beaks of grey eagles esteemed his hand'. In upholding the manuscript reading, maintaining or, the following considerations arise. The verb riuesit clearly requires an animate subject: i loflen is thus excluded as subject and must be the object Pen erir(h)on luit is not available as subject since it is governed by the preposition a: An interpretation with riuesit as 3sg. and the hero as implicit subject would make little sense, "He esteemed his (own) hand1. I therefore interpret riuesit as pluperfect impersonal, 'His hand had been esteemed1. The preposition cr (cp. Olr. of) < *are < PIE *prh2-i IEW 812, has the meaning 'facing, in front of, before', cp. or gatraeth 11. 814, 943, CA 272, J. E. C. Williams (1941), GMW 184, GPC 174 (s.v. cr2). The expression ar pen can be taken as a compound preposition equivalent to Olr. ar chiunn 'in front of, before, awaiting' DIL 107, GOI537 < *are kwennu. This syntagm is reflected, univerbated, but with regular reflex of dative sg., in Brittonic by W/Co. erbyn 'against, by (temporal)' < "are penn VKG II 72, L&P 162, cf. Hamp (1975b). In Cornish the elements are still separable, e.g. er owfyn 'against me' VKG 378. But in Welsh, erbyn is not segmentable. In the instance in question here, ar pen is a renewal of the syntagm within Welsh, after the elements of erbyn had become opaque. I presume this idiom survives in the Welsh expressions for On my own, his own' etc., i.e. arfy mhen fy hun(an), ar ei ben ei hun(an) etc. If ar pen erir(h)on luit "before grey eagles' may be understood as 'in the presence of grey eagles', then the expressions ar... pen... nun(an) can easily be understood as 'in ... own presence'. The sense of the present instance is thus that as a result of the slaughter which had brought the eagles together to feed on the corpses of the dead, the hand of the hero responsible for the slaughter had been the object of admiration. For the full context, see next instance.

220 298. ΒΑ 38.3-4, CA 1. 811, Β39.4, GOSP 113, J 1. 631. en anuit guoreu buit i sgliuon in fury make.prt.3sg. food for birds-of-prey 'In his fury he made food for the birds of prey.' I.e. by making corpses of his enemy: This instance may be interpreted as a causal clause in asyndeton with the preceding instance, explaining it The whole passage may be given as follows: ac ein bu diuant dileit aeron, riuesit i loflen ar pen erir(h)on luit; en anuit guoreu buit i sgliuon. 'And before the lock of Aeron was lost, his hand had been esteemed before the grey eagles, (because) in his fury he made food for the birds of prey.' 299. BA 38.11-12, CA 1. 785, B40.4, GOSP 113, J 1. 613. sengit guidgunet trample trees battle The trees of battle would be trampled.' The Α-variant (insL 264,1.777) reads sengi ipf.3sg., cf. the note in CA 265.

GT: instt. 300-3 = 4 300. BA 25.12-13, CA 1. 1284, GT.27, GOSP 154, J 1. 1029. Carat yreidvyw love.ipf.2sg. live-dangerously.vn. 'You loved to live dangerously.' On breidvyw, cf. CA 355, Geiifa 74, GPC 316. Lloyd-Jones is hesitant about the meaning, but the compound (breid adverb 'hardly, scarcely* + byw 'live, life') is well attested. 301. BA 25.13, CA 1. 1285, GT.28, GOSP 154, J 1. 1030. carwn dy vyw love.ipf.lsg. your.sg. live.vn. Ί loved your being alive.' Jackson translates modally, Ί should have liked you to be alive', but the contrast with the preceding line (preceding instance) favours a realis reading: 'You used to love living dangerously: I used to love your being alive at all* (the hero is now dead). 302. BA 25.14, CA 1. 1287, GT.30, GOSP 154, J 1. 1033. kwynaf dy varw lament.prs.lsg. your.sg. die.vn. Ί lament that you have died.'

221 303. ΒΑ 25.14, CA 1. 1287, GT.30, GOSP 154, J 1. 1034. carut dyhed love.ipf.2sg. war 'You loved war.'

GK: inst 304-6 = 3

304. BA 26.21, CA 1. 1344, GK.8, GOSP not trnsl., J not trnsl. tyllei gam gaffon pierce.ipf.3sg. C. G. 'He would pierce Cam Gaffon.' Cf. CA 364.

305. BA 29.19-20, CA 1. 1395, GK.59, GOSP 156, J 1. 1105. Etmygir e vab tecvann honour.prs.ips. his son T. 'His son Tegfan is honoured.' The translators omit e 'his', following the suggestion of Williams (CA 372), so, The son of Tegfan is honoured'. But the manuscript reading is quite meaningful: the son is honoured out of respect for the honour of the father. This matter does not affect the syntactic analysis.

306. BA 21.22, CA 11. 1398-99, GK. 62-63, GOSP 156, J 11. 1108-9. Pan vyrywyt arveu tros benn cat vleidyeu when hurl.prt.ips. arms over head battle wolves *When arms were hurled over the heads of battle-wolves.'

V[0+pro] inst 307 = 1 B: inst. 307 = 1 307. BA 30.17, CA 1. 1127, B8.3, GOSP 101, J 1. 867. ket ry[ ]lade[t] hwy though ptl. kill.prt.ips. they 'Though they have been killed.1 Cp. inst. 26,1. 361.

222

[O-pro]V instt. 308-35 = 28 A: instt. 308-27 = 20 308. BA 2.11-12, CA 1. 56, A5.ll, GOSP 118, J 1. 66. hyueidhir etmygir tra vo kerdawr H. tall honour.prs.ips. while be.sbj.prs.sg. minstrel 'Hyfaidd the tall will be admired as long as there is a minstrel.' 309. BA 3.6, CA 1. 78, A9.5, GOSP 119, J 1. 88. beych barnasswn burden judge.plpf.lsg. Ί should have judged it a burden.' Beych here is strictly a sub-predicate in apposition to an unexpressed object, but since configurationally the sub-predicate may be assumed to take the place of the object, it is classified here. 310. BA 3.6, CA 1. 79, A9.6, GOSP 119, J 1. 89. [e]ilyw dyn en vyw nys adawsswn shape man ptl. alive neg.-3.O leave.plpf.lsg. Ί should not have left even the shape a man alive.' (ms. dilyw) On the emendation, cf. CA 92. Only thus can the line be interpreted: dilyw 'flood' makes no sense. This is a classic nominativus pendens topicalisation. 311. BA 3.15, CA 1. 92, A11.3, GOSP 120, J 1. 102. med evynt melyn melys maglawr mead drink. ipf.3pl. yellow sweet ensnaring 'They used to drink sweet, yellow, ensnaring mead.' Cf. instt. 447, 448,11. 914-15, 914. Note that while the substantival object is preposed before the verb, its adjectival attributes follow the verb. 312. BA 5.17, CA 1. 178, A18.1, GOSP 123, J 1. 188. Teithi etmygant lawful-rights honour.prs.3pl. 'They honour lawful-rights.' On teitfa, cf. CA 124, GOSP loc.cit; in legal usage, Powell/Williams (21961: 64, 90, 92, 123; Llyfr Blegywryd), Wiliam (1960: 60, 123, 125, et passim; Llyfr lonverth). On the etymology, variously Loth (1910: 511), VKG 124, IEW 1068 (s.v. *tenk'firm, (become) thick'), LEIA T-41(s.v. techtae). W teithi and Olr. techtae appear to

223 reflect Celt *tarrxtyo- < PIE *tok-t-yo-. The Welsh form is formally plural, semantically collective. The expression seems to be a formulaic opening, cp. BT 40.21 Teithi

etmfygjynt,

ibid. 41.16 Teithi etmygant. 3_13_. BA 6.8, CA 1. 211, A19.10, GOSP 124, J 1. 221. blaen bragat briwei front army shatter.ipf.3sg. 'He would shatter the front of the army.1 On bragat, cf. Geirfa 71, GPC 306. 314. BA 6.10-11, CA 1. 217, A19.16, GOSP 124, J 1. 227. er amot aruot aruaethei for promise attack plan.ipf.3sg. To fulfil his promise, he was planning attack.' 315. BA 10.11, CA 1. 401, A36.4, GOSP 131, J 1. 385. e neb a wanei nyt atwenit the unspec. rel strike.ipf.3sg. neg. re-strike.ipf.ips. "Whoever he struck would not be struck a second time.1 I.e. the hero was so powerful that once he had struck his enemy once, no one needed to strike a second time. 316. BA 13.19-20, CA 1. 628, A51.2, GOSP 137, J 1. 527. [a wneli] bratwen gwnelut rel. do.sbj.ipf.sg. B. do.ipf.2sg. 'You would do what Bradwen would do.' (ms. wenn heli) The emendation is guaranteed by the reading of the B-variant, a wnelei vratwen gwnelut (inst. 330,1. 635). 317. BA 16.12, CA 1. 774, A63.1, GOSP 113 (B-variant), J 1. 610 (B-variant). Ardyledawc canu kyman caffat fitting song company getprt.ips. fitting song of the company was had.' Cp. inst. 333,1. 782 (B-variant). On the interpretation, cf. CA 263-64. 318. BA 16.16, CA 1. 776, A63.3, GOSP 113 (B-variant), J 1. 612. sathar sänget trampling trample.prt.ips. trampling was trampled.' The Welsh original does not contain aßgura etymologica. On sath(a)r, cf. VKG 72, 137, L&P 17, cp. OBr. saltrodon gl. graciles VVB 213 (etymology wrong), DGVB 301, Br. saatra 'soil, stain', Olr. satotid 'tramples' DIL 520, LEIA S-19. Pedersen connects these with PIE *sal- 'dirty grey' IEW 879, taking the meaning of Br. saatra as

224 primary, and regarding the meaning 'trample' as secondary, as a way of making something dirty. However, Pokorny (IEW 899) and Vendryes (LEIA loc.cit) take the meaning 'trample' as primary, and connect the forms with PIE *sel- 'spring1 (Gk. δλλομαι, LaL salio etc.). Particularly in view of the Old Breton adjectival derivative saltwcion, the latter view is probably correct: 'leaping' > 'nimble' > 'slender1 > 'feeble', though Fleuriot (DGVB loc.ciL) suggests an early confusion of these two roots, so that the Celtic instances would be portmanteau reflexes of both roots. On sangu, see below, note 2. 319. BA 16.21-22, CA 1. 798, A64.9, GOSP 142, J 1. 626. eur dylyi gold deserve.ipf.3sg. 'He deserved gold.' 320. BA 17.16, CA 1. 836, A68.6, GOSP not trnsl. J 1. not tmsl. diw gwener calaned amdyget friday corpses carry.prt. ips. 'On Friday the corpses were carried.1 In CA 277, Williams rejects amdyget in favour of the B-reading ciuriuet 'were counted', 1. 849 (inst. 562), on the grounds that it occurs two lines later (next instance), curiously he also rejects the latter instance of amdyget in favour of the B-reading of that line, atranhet 1. 850 (inst. 563). Both translators follow Williams in this instance. My approach is to let manuscript reading stand. 321. BA 17.17, CA 1. 838, A68.8, GOSP 124, J 1. 657. diw sul eu llavneu rud amdyget Sunday their blades red carry.prLips. 'On Sunday their red blades were carried.' Cf. preceding instance. 322. BA 17.17-18, CA 1. 839, A68.9, GOSP 124, J. 1. 658. diw llun hyt benn clun gwaetiun gwelet Monday length head thigh blood-image see.prt.ips. 'On Monday, up to the tops of the thighs, a bloody sight was seen.' Williams moots reading gwaettin 'flow of blood' for gwaetiun (CA 278), and this is adopted by the translators, Jackson, 'streams of blood', Jarman, gwaedlin 'blood flowing'. Indeed, gwaetiun is attested only here and in the B-variant of the line (inst. 332, 1. 852), cf. Geitfa 600, GPC 1546). But the very fact of its double occurrence in the A and B-texts, together with the rhyme pattern llunldunlgwaetlun, suggest that the manuscript reading should be upheld. Contrast Jarman (126) in his note to the line.

225 323. ΒΑ 19.14, CA 1. 949, Α76.1, GOSP 147, J 1. 729. Mynawc gododin traeth()e()annor M. G. praise.prs.ips. 'Mynawg of Gododdin will be praised.' Cf. CA 298. Jackson takes mynawc as a substantivised adjective, 'the courteous one1. Jarman on the other hand adopts the interpretation as a personal name. I have followed the same path, arbitrarily perhaps, but it makes no difference to the syntactic analysis. On the adjective mynawc, cf. CA 171, GPC 2538 (s.v. mynog). 324. BA 21.11, CA 1. 1032, A84.5, GOSP 150, J 1. 804. ku kystudywn deargrieve.ipf.lsg. Ί would grieve for a dear one.' 325. BA 21.11-12, CA 1. 1033, A84.6, GOSP 150, J I. 805. ku carasswn dearlove.plpf.lsg. Ί had loved a dear one.' Jackson translates with a relative clause, 'the dear one whom I had loved'. Jarman is ambiguous, The dear one I had loved'. I see no justification for a relative clause here. 326. BA 22.5-6, CA 1. 1067. A86.13, GOSP 100 (B-variant), J 1. 846 (B-variant). garw rybud ο gat dydygei harsh warning from battle bring.ipf.3sg. 'He would bring a harsh warning of battle.' The following line begins with the words conn colon. The equivalent line in the B-text reads gant; can yg colon. The B-text is certainly better here, with ganf at the beginning of the next line as subject of dydygei, producing an OVS structure (inst. 469,11. 1090-91). This is the reading adopted by both translators, cf. CA 319. Nevertheless, the A-reading is meaningful as it stands, and should be included on its own terms. 327. BA 22.12-13, CA 11. 1101-2, A87.1-2, GOSP 151, J 1. 997-98. Peis dinogate vreith vreith ο grwynbalaot ban wreith smock D. his speckled speckled from skins martens ptl. make.prt.lsg. 'Dinogad's smock, his speckled, speckled one, from the skins of martens did I make.'

B: instt. 328-34 = 7 328. BA 24.4, CA 1. 1091. B4.14, GOSP 100, J 1. 847. can ygcalan darmerthei song in Calend provide.ipf.3sg. 'He would provide a song at the New Year's feast.'

226 329. ΒΑ 30.12-13, CA 1. 635, Β7.2, GOSP 101, J 1. 527. a wnelei vratwen gwnelut rel. do.sbj.ipf.sg. B. do.ipf.2sg. 'You would do what Bradwen would do.' Cp. inst. 316.1. 628 (A-variant). 330. BA 31.3-4, CA 1. 1142, B10.3, GOSP 102, J 1. 881. yn dyd gwyth nyt ef weith gocheli in day anger neg. he battle avoid.ipf.3sg. 'On the day of anger he would not avoid the battle.1 On this construction, see note 8. 331. BA 35.15-16, CA 1. 848, B29.6, GOSP 109, J 1. 654. diuyeu cennadeu amodet Thursday envoys pledge.prt.ips. 'On Thursday envoys were pledged.' 332. BA 35.18-19, CA 1. 852. B29.10, GOSP 109, J I. 658. diu Hun hyt beim clun guaet()lun guelet Monday length head thigh blood-image see.prt.ips. 'On Monday, up to the tops of the thighs, a bloody sight was seen.' Cp. inst. 322,1. 839 (A-variant). 333. BA 38.9, CA 1. 782, B40.1, GOSP 113, J 1. 610. Erdiledaf canu ciman cafa[t] fitting, spl. song company get.prt.ips. A most fitting song of the company was had.' Cp. inst. 317,1. 774 (A-variant). 334. BA 38.11, CA 1. 784, B40.3, GOSP 113, J 1. 612. sathar s nget trampling trample.prt.ips. Ά trampling was trampled.1 Cp. inst. 318,1. 776 (A-variant).

GK: inst. 335 = 1

335. BA 27.18-19, CA 1. 1393, GK.57, GOSP 156, J. 1. 1103. e wayw drwn oreureit am rodes his spear ?fine gilded ptl.-lsg.Ogive.prt.3sg. 'He gave me his ?fine gilded spear.' On drwn, cf. CA 185-86, Roberts (1978: 318).

227 [O+pro]V instt. 336-54 = 19 A: instt. 336-46=11 336. BA 3.4, CA 1. 75, A9.2, GOSP 119, J 1. 85. oed cam nas kymhwyllwn be.ipf.sg. wrong neg.-3.O mention.ipf.lsg. 'It would be wrong were I not to mention them.' Though there is no formal mark of subjunctive mood in kymhwyllwn, it is probably to be considered syntactically subjunctive. Note the modal use of the formally indicative oed (see note 6). The object pronoun -5 is resumptive, referring to the topicalised Gwyr a aeth gatraeth (inst. 489,1. 74). 337. BA 5.14, CA 1. 174, A17.25, GOSP 123, J 1. 184. divevyl as talas faultless 3.O pay.prt.3sg. 'Faultlessly he avenged them.' The pronominal object refers to the gyuoedyon 'companions' of 1. 171, inst. 111. Jarman translates divevyl as 'without fail', which, while reflecting the content of the original conceptually, does not accurately capture the precise meaning, on divevyl, cf. Geirfa 338, GPC 979 (s.v. difefl). 338. BA 9.14-16, CA 11. 369-71, A32.7-9, GOSP 130, J11. 353-55. gwlyget gododin en erbyn fraeth ancwyn mynydawc G. G. for spirited feast M. enwawc e gwnaeth a phrit er prynu breithyell gatraeth famous ptl.3.O make.prt.3sg. and costly for buy.vn. land C. 'Gwlyged of Gododdin, for the spirited men, the feast of Mynyddawg, famously he made it, and costly, to buy the land of Catraeth.' This complex period contains the double topicalisation of subject, gwlyget gododin, and object, ancwyn mynydawc (resumptive pronoun in e), and focus of the preposition phrase, en erbyn fraeth (indicating the beneficiary), and of the modifier, enwawc, coordinated with prit which follows the verb. 339. BA 11.11, CA 1. 463, A40.5, GOSP 133. J 1. 441. gwyduc neus amuc ae w[ialen| wild ptl.-3.O defend.prt.3sg. with-his spear 'Wildly he defended them with his spear.' (ms. waiw) On gwyduc'wild, fierce', cf. Geirfa 735, GPC 1757.

228 Jackson's translation has gwyduc as an epithet of gwydien PN in the preceding line, and interprets the object pronoun -s as singular, 'he defended it with his spear'. It is not clear what 'it1 should be. Jackson is perhaps following Williams (CA 187), 'he defended if (emphasis original), but that was on the assumption that gwyduc was a local name. However, since Jackson, like Jarman and here, takes gwyduc as an adjective, in adverbial usage, the object 'it' is left stranded without reference. Jarman's translation makes clear the 3pl. interpretation, referring to the wyr 'men1 of the previous line, 'For his warriors, Gwyddien was an eagle; fiercely he protected them with his spear'. On the emendation to wialen, cf, CA loc.cit.: this is necessary for rhyme, and affects neither the syntax nor the meaning. 340. BA 12.4, CA 1. 530, A43.3, GOSP 134, J 1. 465. ο haedoten gelwit redyrch gwyrnot from merit ptl.-lpl.Ocall.prtips. renowned men fame 'Deservedly were we called renowned men of fame.' Haedot is hapax legomenon, but seems transparent < hoed 'successful, worthy' GPC 1803 + awd4 GPC 238 (forming a deadjectival abstract noun), apparently with orthographic -o- representing -aw-, which is either archaic (CA 166) or progressive. On redyrch, cf. CA 201, cp. Olr. rodarc 'sight, vision1 DIL 509, LEIA R-38. The preposition phrase o haedot is fronted with focus. 341. BA 12.16, CA 1. 557, A46.5, GOSP 135, J 1. 489. ο nerth e kledyfclaer e[m] hamuc from strength his sword shining ptl.-lsg.O defend.prt.3sg. ' With the strength of his shining sword did he defend me.' The emendation is motivated by the parallelism with the following line, next instance. The preposition phrase ο nerth e kledyfclaer is fronted with focus. 342. BA 12.16-17, CA 1. 558, A46.6, GOSP 135, J 1. 490. ο garchar anwar daear em due from prison cruel earth ptl.-lsg.O take.prt.3sg. ' From the cruel prison of the earth did he take me.1 Cf. preceding instance. The preposition phrase ο garchar anwar daear is fronted with focus. 343. BA 16.10-11, CA 11. 730-31, A62.14-15, GOSP 141, J 11. 574-75kywiryth elwir oth gywir weithret right ptl.-2sg.O call.prs.ips. from-your.sg. right deed [rector] rwyvyadur mur catuilet ruler leader wall battle-host ' Rightly are you called, by reason of your faithful deed, ruler, leader, wall of the battle-host.' (ms. mcta

229 Cp. instt. 348, 349, 350, 352,11. 759, 736, 737, 771 (B-variants). The emendation is supplied from the B-variants, and the syntax is not affected. Kywir is fronted with focus. 344. BA 16.11-12, CA 1. 732, A62.16, GOSP 141, J 1. 576. merin a[m] madyein mat yth anet M. son M. fortunate ptl.2sg.O bear.prt.ips. 'Merin son of Madiain, fortunately were you born.' Cp. inst. 351,1. 739. 345. BA 19.7-8, CA 1. 930, A74.7, GOSP 146, J 1. 716. an gelwir (m)or a chynn[o]r ymplymnwyt Ipl.O call.prs.ips. wing and van in-battle "We are called the wing and the van in battle.1 (ms. chynnwr) On the emendations, cf. CA 294. On or a chynnor 'wing and van', cp. instt. 247, 288, 11. 630, 636, nac eithaf na chynhor. Koch (1985/86: 57) attempts to read in mor an archaic spelling of mawr, but the collocation 'wing and van', attested elsewhere, indicates a more natural interpretation, the m in mor arising by scribal corruption from some source. On the other hand, Koch's suggestion (loc.cit.) that an gelwir contains not the Ipl. but a 3sg masc. object pronoun < *in, comparable with MCo. an gelwyr 'he is called1, is worth considering. It does not affect the syntactic classification though. 346. BA 19.13, CA 1. 947, A75.5, GOSP not trnsl., J not trnsl. an gelwir ny faw glaer fwyre Ipl.O call.prs.ips. we fame bright ? 'We are called bright of fame ?...' Faw < Lat. fama: fawglaer may be taken as a compound Of bright fame1, cp. BT 72.25. Fwyre is hapax legomenon: for various possibilities, cf. CA 298, Geirfo 515, GPC 1331.

The prefixed object pronoun an is reinforced by the suffixed ny. B: instt 347-52 = 6 347. BA 31.15-16, CA 11. 231-32, B12.5-6, GOSP 102, J 1. 237. en vavr yt uaer as gym()myrr()hut ptl. great to-your.sg. steward 3.O diminish.sbj.ipf.2sg. 'Vastly would you diminish it for your steward.' The object pronoun refers to the metaphorical wine in the concessive clause of 1. 230 (insL 532), ceuei gwin gwaet meirw meint a wanut Though the blood of the dead, as many as you would kill, were wine*.

230 The initial of the verbal form gym()myrr()hut does not show lenition, but apparently a graphic representation of the neutralisation of voice vs. voicelessness after the sibilant s. 348. BA 32.4-6, CA 11. 759-60, B14.9-10, GOSP 104, J 11. 595-96. enwiryt elwir oth gy wir weithret true ptl.2sg.O call.prs.ips. from-your.sg. right deed rector rwyfyadur mvr pob kyuyeith ruler leader wall every compatriot Truly are you called, by reason of your faithful deed, ruler, leader, wall of every compatriot.' Cp. instt. 343, 349, 350, 352, 11. 730, 736, 737, 771. Cf. CA 256-57, GOSP 103, Jarman 118, Koch (1988: 26-27) on the relationship of these lines. Jarman prints cywir instead of enwir, but the latter reading is almost certainly original. It is not en wir 'truly' (= modern yn wir), as at CA loc.ciL, Geirfa 481, but intensive, with prefix an-2 GPC 104 (cf. 1211 s.v. en-). This interpretation is favoured by Bromwich (1978a: 157-58), Koch (loc.cit). 349. BA 32.8-9, CA 1. 736, B15.4, GOSP 104, J not tmsl. enwiryt elwir oth gywir gvjrejit true ptl.2sg.O call.prs.ips. from-your.sg. right deed Truly are you called, by reason of your faithful deed,...' etc. (ms. gverit) Cp. instt. 343, 348, 350, 352,11. 730, 759, 737, 771. Jackson translates as if this and the following instance were deliberate incremental repetition, Tor your loyal deeds truly are you called - for your loyal deeds rightly are you called - the ruler, the prince, the rampart of every nation'. But the repetition is probably due to confusion in the manuscript tradition. 350. BA 32.9, CA 1. 737, B15.5, GOSP 104, J not tmsl. kewiryth elwir oth gywir weithret right ptl.2sg.O call.prs.ips. from-your.sg. right deed 'Rightly are you called, by reason of your faithful deed,..." etc. Cp. instt. 343, 348, 349, 352,11. 730, 759, 736, 771. 351. BA 32.10-11, CA 1. 739, B15.7, GOSP 104, J not trnsl. meryn mab madyeith mat yth anet M. son M. fortunate ptl.2sg.O bear.prt.ips. 'Merin son of Madiaith, fortunately were you bom.' Cp. inst. 345,1. 732. 352. BA 37.10-11, CA 1. 771, B36.10, GOSP 111, J 1. 607. enuir ith elwir od g[iui]r guereit true ptl.2sg.O call.prs.ips. from-your.sg. right deed Truly are you called, by reason of your faithful deed,..." etc. (ms. guur)

231 Cp. instt. 343, 348, 349, 350,11. 730, 759, 736, 737. Jarman prints cywir instead of enwir, cf. on inst. 348,1. 759. Williams prints giuir without comment in his edition, though Evans (1908: 37) had printed guur, and the manuscript clearly shows the latter reading, see the facsimile, Huws (1989). Palaeography apart, Williams' interpretation as giuir is, of course, quite correct. GT: instt. 353-54 = 2 353. BA 25.5, CA 1. 1266, GT.9, GOSP 153, J not trnsl. yd i gweles ptl. 3.O see.prt.3sg. 'He saw it.' Jackson translates, 'As long as I saw him', as if reading hyd y gweleis, following a suggestion of Williams (CA 269). However, the rhyme with tres in the following line forces the interpretation as 3sg.; and the occurence of yd y klywi in 1. 795 (inst. 515) suggests that yd i here be interpretated as above. An infixed pronoun of the form / may occur in the constructions with keui discussed in note 15. A regular phonological source for such a pronoun is provided by the accusative plural of the anaphoric pronoun, *is, possibly a syncretism of masculine PIE *i-ns and feminine PIE *l-ns. As a result of phonological reductions that rendered the form morphologically opaque with regard to gender and case, it may be that number too was neutralised, so that it was generalised as a third-person object pronoun. However, apart from regular phonological processes, idiosyncratic developments may have played a role in the development of the form, allowing us to posit a wider base for its generalisation. On the very reasonable assumption that the particle yd (/ad/) reflects PIE *i-d*e (IEW 284-85, and note 8, below), then MW yd i might be thought to reflect a Brittonic syntagm *ide in, with the masculine accusative singluar of the anaphoricum (< PIE *im, cp. Olr. infixed pronoun, masculine, Class A -a0-, GOI 283-84, VKG II 171, L&P 216, Szemeronyi 41990: 219, IEW 281-82). At an undefined time, but probably relatively late, the final vowel of *ide may have been elided in sandhi before the vowel of *in with compensatory lengthening of the latter, giving *id' in > yd i (cp. *to-aget > "taget > W daw 'comes', with idiosyncratic development of the initial consonant). A similar development would have affected the neuter *i(d). Together with the regular phonological development of *is, this would have favoured a generalised third-person object pronoun /'. It is possible that the object pronoun is proleptic here, referring to the following line (inst 172) as a sentential complement: yd i gweles or hutd tres tardei galled 'He saw (it) on the chain of battle bushes were growing'. From the context it is unclear who the subject of gweles might be.

232 354. ΒΑ 26.7, CA I. 1321, GT.64, GOSP 154, J 1. 1074. twryf neus kigleu uproar ptl.-3.O hear.prt.lsg. "The uproar, I heard it.' A canonical nominativus pendens topicalisation, with resumptive object pronoun -s.

V[S-pro]fO-pro] instt. 355-74 = 20 (There are no instances of V[S±pro][O+pro].) A: instt. 355-69 = 15 355. BA 3.8-9, CA 1. 82, A9.9, GOSP 119, J 1. 92. ny mennws gwrawl gwadawl chegrwn neg. demand.prt.sg. brave gift father-in-law "The brave man did not demand a father-in-law's dowry.' Jackson translates, The brave man wanted no father-in-law's dowry', Jarman, The hero desired no father-in-law's dowry'. However, the volitive sense of mynnu translated by 'wanted, desired1 is an ongoing and uncompleted action (durative), and is expressed with imperfective aspect, in the past, in the imperfect tense, in this poetry, e.g. inst. 266, 1. 821 hu mynnei eng kylch byt eidol anant Thus he desired, all over the world, the praise of minstrels'. The present instance, mennws, is in the preterite, and as such is coded with perfective aspect, and so cannot have a volitive sense, but must represent a punctual action. Thus, the meanings of mynnu which are relevant here are 'procure, claim, demand1. The sense of the line is that the hero did not make the father-in-law's customary claim for a dowry at the marriage of his son, for he did not live long enough to get that far. 356. BA 3.21-22, CA 11. 101-2, A12.5-6, GOSP 120, J 11. 111-12. goreu (yw) hwnn kyn kystlwn kerennyd make.prt.sg. this.m. before declare.vn. peace enneint creu ac angeu oe hennyd bath blood and death from-his enemy 'Rather than declaring peace, this man made a bath of blood and death of his enemy.' On kystlwn 'ally, join, claim, declare', cf. CA 100, Geitfa 269, GPC 817 (s.v. cystlynaft < *kom~stlond-, cp. OW istlinnit gl. profatur VVB 168, sdlinet 'lineage' (< 'what is declared, named') CA 1. 1189, Olr. sluindid 'expresses, declares, names' DIL 550, VKGII632, L&P 396-97, LEIA S-137-38. On hennyd 'companion, enemy' < *sent-yo-, cf. GPC 1852 (s.v. hynnydd), cp. Co. hynse, Br. hentez. It is a derivative of sentu- > W hynt, Br. hent, Olr. set etc. IEW 9089 'way'.

233 Hwnn is strictly pronominal, a demonstrative, but for this classification, I reserve the feature +pro for personal pronouns. 357. BA 7.7, CA 1. 261, A23.4, GOSP 126, J 1. 264. treiglessyd llawr lloegrwys giwet turn.prtsg.abs. lone Englishmen mob The lone hero overthrew the English mob.' Ciwet < LaL civitas: it has meanings in the range 'rabble, mob, people, nation, troop' in Welsh, cf. GPC 485. 258. BA 7.17, CA 1. 290, A24.14, GOSP 126, J 1. 281. gorgolches e greu y seirch wash-over.prt.sg. his blood his armour 'His blood washed over his armour.' An interesting instance where the modification the preverb makes to the meaning of verb is synchronically transparent: gor- < PIE *uper(-i) Over' + golchi 'wash1. 359. BA 8.2, CA 1. 301, A26.2, GOSP 127, J 1. 292. ny deliis meirch neb marchlew neg. catch.prt.sg. horses no-one M 'No one's horses caught up with Marchlew.' Cp. inst. 3701. 313 (B-variant). 360. BA 8.5.6, CA 1. 308, A26.9, GOSP 127, J 1. 299. yt rannei rygu e rywin ptl. share.ipf.sg. its.dear his its.wine The most dear one would share out his plentiful wine.' On the morphology of ry + adjective/noun, cf. WG 268, VKG II 10. On Olr. ro in this function, GOI528. W ry, Olr. ro < PIE *pro 'forwards, before' IEW 813-14. The intensive function in composition has developed transparently from the original adverbial meanings, e.g. *pro-koimo- 'dear before (all others)' > rygu 'most dear', cp. Gk. πρόκακος 'very bad'. 261. BA 8.6-7, CA 11. 310-11, A26.11-12, GOSP 127, J 11. 301-2. valpan vel medel ar vreithin as when strike.prs.sg. reaping-party on unsettled-weather e gwnaei varchlew waetQlin ptl. make.ipf.sg. M. blood-flow ' As when a reaping party strikes in unsettled weather would Marchlew make bloodshed.' The sentential simile valpan vel medel ar vreithin (on which see inst 119) is fronted under focus.

234 362. ΒΑ 9.2, CA 1. 352, Α30.10, GOSP 129, J 1. 336. ny weles vr[ ]un e dat neg. see.prt.sg. the one his father 'Not one saw his father (again).'

363. BA 10.1-2, CA 1. 383, A34.4, GOSP 130, J 1. 367. na wnelei kenon kelein neg. make.sbj.ipf.sg. K. corpse "That Cynon would not make a corpse.' This instance is the sentential complement of ny thraethei ... Ήβ would not relate ..." (inst. 29). Jarman prints galan for kelein. MW kelein, modem ceMn (GPC 455-56) is a reformed singular. The original singular *calan, pi. calaned, has been replaced under the influence ofgwreig 'woman', pi. gwraged, neidyr, pi. nadred, to give kelein, pi. calaned (cf. CA 169, LEIA C-156 s.v. colainn). *CaJan as opposed to Olr. colainn shows the same vocalic assimilation as ianm 'thunder' as opposed to Olr. tprann (De Bemardo-Stempel 1994: 21). Cp. Brit. Colania LN, PNRB 311-12.

364. BA 11.12, CA 11. 465-66, A40.7-8, GOSP 133, J 11. 443-44. amuc moryen gwenwawt mirdyn defend.prt.sg. M. fair-song M. "Morien defended the fair song of Myrddin.' On the significance of the line, cf. GOSP loc.cit, Jarman 107-8. Jackson translates gwenwawt as 'poetic inspiration'. In place of this rather flowery interpretation, Jarman's more sober 'fair song' is adopted here. On Myrddin, cf. Jarman (1937: particularly 20ff; 1959, 1960, 1975/76, 1978), Bromwich (1966: 32, 21978: 469-74), Tolstoy (1983/84).

365. BA 12.15, CA 11. 555-56, A46.3-4, GOSP 135, J 11. 487-88. nyt emduc mam mor eiryan gadarn neg. give-birth.prt.sg. mother so fair powerful "No mother gave birth to one so fair and powerful.' Jarman prints mam nid ymddug mor eirian gadam, with SVO order, following the suggestion of Williams (CA 209). The reason is that in the manuscript order of the words, mam ends the line, but does not form rhyme with the surrounding lines, gorucfessyllut etc. By reading mam nyt emduc at the end of 1.555, rhyme is restored. However, mam does form a rhyme in its position in the manuscript text after all. It rhymes with cadam/haeam in the middle of the following line. Cp. BT 69.20-21 calethvm rhyming with ascwrn (cf. Isaac 1992: 385ff.). The rhyme pattern of these lines is therefore as follows, marking the rhymes and :

235 llary vronn haeladon ny essyllut nyt emda daear nyt emduc mam mor eiryan gadarn haeam gaduc Cp. 11. 1143-45: baran baed oed bleidic mab eli er[y]vessit gwin gwydyr lestri llavn ac en dyd camavn camp a wne(e)i This is a case ofgaircyrch (Isaac op.cit.: 337ff.); for the use of gair cyrch in classical Welsh metrics, cf. Morris-Jones (1925: 275ff.). The manuscript reading is therefore to be upheld, and the instance is one of VSO. The phrase mor eiryan gadam is simultaneously also the subject of emda in inst. 442. 366. BA 13.10, CA 11. 608-9, A49.1-2, GOSP 136, J 11. 507-8. Er()kryn e alon ar(-)af ery brwydrin fear.prs.sg. his enemies weapon daring battle 'His enemies fear the weapon of the daring one of battle.1 On the form oraftwith svarabhakti) of arf< Lat. arma, cf. CA 220, GMW 12, and cp. araf(S. J. Williams 1930a: 207; Pererindod Siartymaen). Jackson translates, 'His enemies feared his weapon, the combative eagle', Jarman, "His enemies fear his weapon, the fierce eagle'. Both insert the second possessive pronoun in the translation, and both adopt the emendation suggested by Willams (CA 221) of ery to eryr 'eagle*. Both alterations to the manuscript text are quite unnecessary however. On ery 'daring' < CF + hy < *are-sego- (< PIE *segh- 'grip, defeat in battle' IEW 888-89), cf. Geirfa 486 (s.v. erhy2), GPC 1239 (s.v. ery2). The phrase arqfery brwydrin can be understood in a number of ways without emendation: i)

ery may be an adjectival attribute ofaraf, the whole phrase 'the daring weapon of battle' being a metaphor for the hero;

ii) or this phrase might be elliptical for '(him of) the daring weapon of battle'. iii) or, probably best, and adopted in the gloss above, ery is in substantival use, so 'the weapon of the daring (one) of battle'. Er kryn (sic ms.) 'fears' is compounded of or- < are Opposite, facing, before' and crynu 'tremble'. Literally then 'trembles before ...'. The etymological meaning is quite appropriate here: 'His enemies tremble before the weapon of the daring one of battle'. 367. BA 17.2, CA 1. 802, A65,3, GOSP 142, J not tmsl. dimcones [e] lovlen benn eryron llwyt satiate.prt.sg. his glove head eagles grey 'His glove/hand satiated the beaks of grey eagles.1 I.e. by making many corpses on which the birds fed.

236 Jackson translates, 'He glutted the grasp of grey eagles' beaks'. This is perhaps just possible, but the B-variant has / loflen in this position (inst. 297, 1. 810; this variant has, however, a completely different structure and interpretation), and the concrete use of lovlen for the 'glove/hand' of the hero seems more appropriate. Dimcones is certainly for digones, but the -m- here is difficult, cf. the comparisons with dymgo/u, dimpyner made in CA 271. Lloyd-Jones treats it as the Isg. infixed object pronoun (Geiifa 347, s.v. digoni2·), but what is the function? The word may be a fossilised form culled from early traditional texts or word lists, without consideration of the correct segmentation. If we were to insist on giving the infixed pronoun a function, we might think of something like a sympathetic dative: "Manus eius mihi ravarum capita aquilarum suffecit'. By this means, the poet might express his feeling of involvement in the situation mentioned, and indeed his approval of the behaviour of the hero, "His hand satiated the beaks of grey eagles, and that I find good'. Perhaps a similar analysis might be possible for inst. 165, 11. 1186-87, dim()guofui] ediu ο adorn [h/einim TTelum mihi ex Heinimi manu veniet'. But the point cannot be laboured. 368. BA 17.13, CA 1. 831, A68.1, GOSP 143, J not trnsl. Dyfforthes meiwyr molut nyuet support.prt.sg. cowards praise hall 'Cowards joined in the praising in the hall.' Cf. the discussion of the line by Jarman 125. Dyfforthi here seems to have a meaning in the range of 'support, assist' that is 'take part in1, cf. GPC 1114 (s.v. dyborthaf). The idea seems to be that although they were cowards, they too took part in the proceedings, so great was the cause. The first element of mei-wyr is < *medyo- < PIE ^me^-yo- 'middle1 IEW 706-7. The meaning is not 'middle-men' but 'mediocre men', cf. I. Williams (1921). On nyuet 'hall', cf. CA 274, PKM 298-99, cp. Gaul, νεμητον 'sanctuary, sacred grove/place', Lejeune (1985: 208), Olr. named 'sanctuary' OIL 475, LEI A N-9, IEW 763-64. The verb does not agree in number with the plural subject Cp. inst. 373,1. 843 (B-variant). 369. BA 22.16-17, CA 1. 1110, A87.10, GOSP 151, J 1. 1006. mal ban llad Hew llywywc like ptl. kill.prs.sg. lion ? 'As a lion kills a?...' Llywywc is hapax legomenon and unclear. Jackson does not translate the word, Jarman has, 'As a lion kills its prey1. Parry (1944: 19) gives, 'fel y lladd Hew anifail(?)', while Evans (1981: 13) translates with an antipassive construction, Tel y bydd Hew yn lladd1 (this translation is ascribed to Parry by Evans, op.cit: 114). GPC 2291 gives 'a

237 kind of animal, ?fox', suggesting, '?an error for Hywynog, a variant of llwynog'. This suggestion does not help since the rhyme is in -we, and llwynog is < MW llwynawc. The syntax is quite clear however. B: instt. 370-74 = 5 370. BA 32.22-33.1, CA 1. 313, B18.2, GOSP 105, J 1. 292. na deliis meirch neb marchlew neg. catch.prt.3sg. horses no-one M "That no one's horses caught up with Marchlew.' Cp. inst. 359,1. 301 (A-variant). 371. BA 33.8-9, CA 1. 1163, B19.8, GOSP 105, J 1. 902. ardwyei waetnerth e gerth or gat defend.ipf.sg. G. his right from-the battle 'Gwaednerth would defend his rights through battle.' certh < Lat. certus. The preposition ο here has its instrumental meaning, GPC 2601, meaning ll(b). Jackson translates loosely, 'in battle'. Jarman prints ο gad, omitting the article as suggested by Williams (CA 331: 'nid oes angen bannod'). But there is no reason to omit it, other than to tidy up a text in which the rarity of the article is considered to be a mark of archaism. 372. BA 34.18-19, CA 11. 1192-94, B26.7-9, GOSP 107, J 11. 927-28. ladaut map nuithon ο eur()dorchogyon cant o deymet kill.prt.sg. son N. from gold-torqued hundred from princes The son of Nwython killed a hundred gold-torqued princes.' 373. BA 35.12, CA 1. 843, B29.1, GOSP 109, J 1. 649. Ni forthint ueiri molutniuet neg. bear.ipf.pl. stewards praise hall The stewards could not bear the praise in the hall.' This at first sight strange observation probably means that the praise of the heroes which was declaimed in the hall was so loud (the heroes being so great), that the stewards whose task it was to serve at the feast were unable to bear the noise. Thus the translators: Jarman translates molut as 'clamour', but Jackson's translation is more transparent, 'the [shouts of] praise'. On molut, cf. CA 274, CL1H 155, GPC 2482. The use of molut figuratively for 'loud noise' is a particular poetic metaphor in CL1H 22.6a-b (Rowland 1990: 415), Yssydd Lanfawr dra gweilgi, y gwna mor molud wrthi There is a Llanforover the sea where the sea makes molud. However, in BT 43.10-12, Aduwyn goer yssyd ae gwna kyman media a molut ac adar bann, translated by Williams (1972: 165), There is a fine fortress, where a multitude makes loud revelry, and so do the

238 birds', the sense of '(loud) praise' is just as appropriate. In the present instance, in the specifically eulogistic context, it does not seem wise to divest nwlut of its meaning 'praise1 to regard it as a general term for 'clamour'. Cp. inst. 267,1. 831 (A-variant). The verb agrees in number with the plural subject. Evans (1910) prints muet, but the minims are really rather ambiguous. Only niuet is meaningful. 374. BA 37.1-2, CA 1. 1236, B34.16, GOSP 111, J 1. 966. dirlis(h)ei etari[ ]ued()iuet deserve.ipf.sg. E. his mead-drinking 'Edar deserved his mead-drinking.' The subjunctive -Λ- in dirlishei seems quite unmotivated. Jackson translates, 'Edar deserved to drink his mead1, Jarman, 'Edar deserved his drink of mead'. Ued()iuet in the text is a compound nomen actionis 'mead-drinking', in effect with incorporated direct object, making the whole intransitive, and thus allowing the possessive pronoun i to code the subject role, whereas for a transitive verb it would have to code the object role.

V[S+pro][O-pro] instt. 375-89 = 15 A:instt.375-85=ll 375. BA 4.2-3, CA 1. 106, A13.2, GOSP 120, J 1. 116. ne[u] llewes ef vedgwyn veiQnoethyd ptl. devour.prt.sg. he mead-feast midnights Ήε devoured the mead-feast at midnight(s).' Jarman's translation is somewhat inexact, 'Greedily he drank mead at night-time', though his note to the line (85) acknowledges the meaning 'lit. "midnight1". Jackson is more precise, "He drank up the mead-carousal at midnight'. The form veinoethyd is not quite clear. On meinoeth '(at) midnight1, cf. GPC 2412. Behind this is certainly a phrasal adverb *medyon noctan (accusative of time). The accusative case of the reconstruction is assured by the corresponding meinyd '(at) mid-day1, GPC 2411, < *medyon diyen (or *diyon, after an immediate Brittonic nominative sg. *diyos < PIE *dyeus). The alternative form meinddydd has the lenited initial of the word for day, dydd, analogically inserted. These old phrases must clearly have given primarily adverbs in Welsh, 'at midnight, at mid-day', cp. peunyd, beunydd, Br. bemdez < *papon diyen 'every day1. These adverbs must have been secondarily substantivised in Welsh, to give 'midnight,

239 mid-day', and from that base, meinoeth may have got a plural meinoethyd, attested (uniquely?) in this instance. Or meinoethyd could simply be a nonce formation for rhyme here. This formation of the plural cannot be inherited, but could the plural semantics be? Cp. Gk. με σα L νύκτες 'midnight'. 376. ΒΑ 6.4, CA 1. 201, Α19.1, GOSP 124, J 1. 212. Eveis y win a med e mordei drink.prt. Isg. I wine and mead in court Ί drank wine and mead in the court.' 377. BA 6.13, CA 1. 221, A20.1, GOSP 124, J 1. 231. Eveis y win a med e mordei drink.prt.lsg. I wine and mead in court Ί drank wine and mead in the court.' Both Jackson and Jarman read a 2sg. verb here. But it seems unnecessary to alter the structure of the line in this way, making the link with the first line of the previous awdl (preceding instance) less strong. 378. BA 12.9, CA 1. 539, A45.2, GOSP 135, J 1. 471. ny dialaf vy ordin neg. avenge.prs.lsg. I attack Ί do not avenge attack.' Jackson fills out the sense, Ί am unable to take vengence for provocation1. Jarman, Ί avenge no provocation'. Cf. GOSP 134, Owen (1978: 139). 379. BA 12.9-10, CA 1. 540, A45.3, GOSP 135, J 1. 472. ny chwardaf y chwerthin neg. laugh.prs.lsg. I laugh Ί do not laugh a laugh.' Thefigura etymologica is in the original. 380. BA 12.13, CA 1. 551, A45.14, GOSP 135, J 1. 483. neu cheinft] e ododin ptl. sing.prt.lsg. I G. Ί sang me Gododdin.' (ms. cheing)

381. BA 18.2-3, CA 1. 873. A70.5, GOSP 144, J 1. 674. oed garw y gwnaewch chwi waetlin be.ipf.sg. harsh ptl. make.ipf.2pl. you.pl. blood-flow 'It was harshly that you used to make a flow of blood.' The matrix clause is a cleft sentence.

240 382. ΒΑ 18.4, CA 1. 875, Α70. 7, GOSP 144, J 1. 676. oed llewy lladewch chwi dynin be.ipf.sg. easy ptl. kill.ipf.2pl. you.pl. corpse 'It was easily that you would cut down a corpse.' For the structure, cf. preceding instance. On/few'easy', cf. inst. 219, 1. 191. On dynin, cf. note 11. 383. BA 20.2, CA 1. 966, A78.1, GOSP 147, J 1. 991. Gweleis y dull o benn()tir a[ ]doyn see.prt.lsg. I array from headland rel. come.ipf.pl. Ί saw an array which was coming from the headland.1 Jackson does not translate with a relative clause, Ί saw an array, they came from the headland'; it is not clear precisely what sort of syntactic structure he is seeing here, cf. Koch (1988: 31-32). Jarman does translate with a relative clause, and translates penntir as a local name. In this he follows Williams (CA xli, I. Williams 1972: 79). 384. BA 20.10, CA 1. 986, A79.9, GOSP 148, J 1. 754. caraf vy vudig love.prs.lsg. I victorious Ί love the victorious one.' 385. BA 22.17-18, CA 1. 1112, A87.12, GOSP 151, J 1. 1008. dydygei ef penn ywrch bring.ipf.3sg. he head roe-buck 'He would bring (back) a roe-buck.' On the syntax of penn ywrch, cf. I. Williams (1926d: 135). The subsequent phrases, penn gwythwch 'a wild pig', etc., constitute further objects of dydygei. B: instt. 386-89 = 4 386. BA 23.6, CA 1. 972, Β 1.1, GOSP 98, J 1. 991. Gveleys y dull ο bentir a doyn see.prt.lsg. I array from headland rel. come.ipf.pl. Ί saw an array which was coming from the headland.' Cf. inst. 383,1. 966 (A-variant). 387. BA 23.7, CA 1. 974, Β 1.3, GOSP 99, J 1. 993. Gueleys y deu oc eu tre re ry gwydyn see.prt.lsg. I two from their town swift ptl. fall.ipf.pl. Ί saw two (arrays) which were descending swiftly from their town.1 The 'town' is presumably more likely an encampment or fort. Jackson translates the implication of this sentence, Ί saw a second one ...'.

241 388. ΒΑ 23.8-9, CA 1. 976, Β 1.5, GOSP 99, J 1. 995. Gueleys y wyr tylluavr gan wavr a doyn see.prt.lsg. I men sturdy with dawn rel. come.ipf.pl. Ί saw sturdy men who were coming with the dawn.1 On tyllvavr, cf. CA 303, CL1H 70. The Α-variant reads dulfyawr, inst. 275,1. 970. 389. BA 33.31, CA 1. 1174, B22.1, GOSP 106, J 1. 913. Neuteryueis y ued aryg kerdet ptl. drink.pit. 1 sg. I mead on my journey.vn. Ί drank mead (just) before I set out on my journey.' Jackson translates eryueis, Ί drank off mead at one draught ...'. I am not sure what motivates this rather expansive treatment of the verb. On eryuet 'drink, carouse', cf. Geirfa 488, GPC 1239 (s.v. eryfqft: the meanings do not justify Jackson's translation. Nor does the tense used: the preterite codes only that the action was completed at a past time (perfective aspect), not that it necessary took place in a single instant.

[S-pro]V[O-pro] instt. 390-403 = 14 (There are no instances of [S±proJV[O+prol.) A: instt. 390-402 = 13 390. BA 2.20, CA 1. 67, A7.4, GOSP 118, J 1. 77. colovyn glyw reithuyw rodi arwar pillar battle R. give.ipf.sg. delight The pillar of battle, Rheithfyw used to give delight1 Jackson translates, "... delighted in giving [gifts]', Jarman, "... delighted in giving'. In this they adopt Williams' second alternative (CA 88), while I adopt his first, also Koch (1988: 29). Cf. Jarman's note on nodi arwar (83), 'Probably a compound adjective, "delighting in giving", i.e. "generous"'. That interpretation has rodi as the verbal noun. But arwar is a substantive, 'pleasure, delight', apparently unattested adjectivally, cf. Geirfa 43, GPC 213. This implies that the finite interpretation of rodi is preferable. 391. BA 5.1-2, CA 11. 140-41, A16.1-2, GOSP 122, J 11. 150-51. Blaen echeching gaer glaer ew[y]gei B. ?spacious fort bright inspire.ipf.sg. gwyr gweiryd gwanar ae dilynei men faithful battle rel.-3.O follow.ipf.sg. 'Blaen of the ?spacious, splendid fort used to inspire the faithful men of battle who used to follow him.' Cf. Williams' notes (CA 113-15). Echeching is difficult Geirfa 435 and GPC 1160 register echeching as a hapax legomenon adjective in the meaning 'generous, liberal spacious', consisting of ech- + eching, echyng, the latter meaning 'captivity, restriction'

242 Geirfa 436 (GPC 1163 registers eching, echyng as the verbal noun of echyngaf 'confine, imprison1, as well as an adjectival use, 'narrow, confined', attested in dictionaries since the 16th century). Since the morphology of an adjectival echeching would be quite transparent, with the first ech in a privative sense (diechyng is also attested, with the meanings 'generous, spacious' etc., Geirfa 333, GPC 971: ech-eching, di-eching, cp. eh-ovyn, di-ovyn 'fearless') the manuscript reading can be maintained as it stands. It is perhaps a natter of taste whether one takes the adjective as modifying Blaen ('Blaen the generous'), or goer ('the spacious fort'), but the mutation of goer, consistent with the presence of a proposed adjective, favours the latter alternative. Koch's suggestion (1985/86: 55) that echeching consists of the preposition ech Out of and a dative case-form of ehang < *exs-ang- is perhaps not to be dismissed out of hand (cf. note 9), but the affection shown (assuming orthographic i for /i/), suggesting original o- or a-stem declension is not consistent with the presumably w-stem declension of this adjective (cp. Skt. awihu-, Go. aggwus, IEW 42). Various late analogical forces could be posited to account for alterations in the stem class, but we are hardly in a position to provide detailed motivation for this at present. Jarman also reads ech as the preposition, but emends eching to Eidyn (following Williams' afterthought, CA 114). This is a rather drastic emendation, and echeching is not, I think, so obscure as to justify it. In the interpretation of the verb (ewgei in the manuscript), I, like Jackson and Jarman, follow Lewis (1948: 23), who sees a denominative verb derived from ewyc 'eagerness' (GPC 1264) < *awiko- < PIE *h2euhj- "blow" IEW 81-82: ewygei would be factitive 'make eager, urge, inspire'. Lewis seems to be thinking of a phonological development ewygei > ewgei, but a simple copyist's error (caused perhaps by confusion of minims in something like euigei) is more plausible. Koch's argument (loc.cit.) for a verb gwgei is less attractive, requiring more radical emendation and the hypothesis of a seventh-century written version of the text in which the verb is to have appeared as *uuce. This gwgei is to have been the simple verb (correspondig to Olr. fichid 'fights' DIL 305) of which amygei, diwygei are compounds. The simple verb is otherwise unattested. 392. BA 5.2-3, CA 11. 142-43, A16.3-4, GOSP 122, J 11. 153-54. Blaen ar e bludue dygollou[e]i(t) B. on his feather-weave dispense.ipf.sg. vual en[ ]e vwynvawr vordei drinking-hom in his luxurious court 'Blaen on his feather cushion used to dispense the drinking horn in his luxurious court.' Onpludue, cf. CA 115. The first element is plu 'feathers', singulative plufen

< Lat.

pluma. In pludue, the spirant appears as a dental by dissimilation from the following labial (from OW */pluvwe/).

243 The second element is gwe 'woven article1 GPC 1067-68, < PIE *weg- 'weave' IEW

1117. The reading dygollouei for dygollouit in the manuscript is necessary for rhyme. 292. BA 5.5-6, CA 1. 148, A16.9, GOSP 122, J 1. 158. Blaen erwyre gawr B. raise.prs.sg. shout 'Blaen raises the battle-cry.' 394. BA 7.14-15, CA 1. 285, A24.9, GOSP 126, J 1. 276. beirdbyt bamant wyr ο gallon poets world judge.prs.pl. men from heart 'Poets of the world pronounce judgement on men of heart' Cf. note 9. Jackson translates, The bards of the world judged him to be of manly heart'. Jarman likewise, with 'poets' in place of "bards'. The historical present in this interpretation is by no means impossible, but suspicious in the circumstances: all other verbs in the awdl are preterite (bu χ 6, edewit, diebyrth, gorgolches) or imperfect (oed χ 2). Further, the emendation of plural wyr to singular wr seems quite arbitrary to me. I think Williams' first suggestion was right (CA 147), 'Gwaith y beirdd yw barnu'r dewr o galon, pwy sydd ddewr, a phwy ddewraf (The job of the poets is to judge the brave of heart, who is brave, and who is bravest'). The verb bamant thus has the present tense in a gnomic function. 395. BA 8.12, CA 11. 327-28, A28.1-2, GOSP 128, J 11. 317-18. kerediccaradwye glot achubei ...not K. lovable his fame seize.ipf.sg. renown 'Ceredig of the lovable fame used to seize renown.' Keredic and not (LaL nota) are subject and object respectively of the further verb gwarchatwei (next instance). 396. BA 8.12, CA 11. 327-28, A28.1-2, GOSP 128, J 11. 317-18. keredic caradwy e glot ... gwarchatwei not K. lovable his fame keep.ipf.sg. renown 'Ceredig of the lovable fame used to keep renown.' Cf. preceding instance. 397. BA 11.5, CA 1. 453, A39.7, GOSP 133, J 1. 431. twrchgoruc amot e mlaenystre ystrywyawr boar make.prt.sg. pledge in front border wily The boar made a pledge in bront of the border, wily.' Cf. CA 184. Jarman translates, Ά boar pledged himself before the treacherous border1, following Williams. But Jackson goes his own way, The boar, the wily one made a compact in front of the boundary', with ystrywyawr 'wily, artful* in apposition to

244 the subject twrch "boar1. I prefer this interpretation, with ystrywyawr referring to the animate, human hero, rather than to the inanimate, abstract "border' (as Williams and Jarman). 398. BA 13.11-12, CA 1. 611, A49.4, GOSP not trnsl., J 1. 510. byssed brych briwant banfingers speckled shatter.prs.pl. head 'Blood-spattered fingers shatter the head.' Jarman translates, The fingers of the speckled one shatter a head', following Williams (CA 222) in interpreting brych 'speckled1 in substantival function for the 'speckled hero'. In principle there is nothing against this, but I prefer the simpler adjectival interpretation suggested above, 'the (blood-) speckled fingers'. 399. BA 18.15-16, CA 1. 905, A72.2, GOSP 144-45, J 1. 691. gwlat()gord[ c]arei gwrduedel country-host love.ipf.sg. mighty-reaper The warband of the land loved the mighty reaper.1 (ms. gwlat gordgarei) On medel, cf. inst. 119,1. 310. The 'mighty reaper' is the warrior-hero, cutting down the enemy in battle. 400. BA 19.20-21, CA 1. 961, A77.4, GOSP 147, J 1. 741. peleidyr pwys p[er]iglyn benn periglawr spears weight endanger.ipf.pl. head endangerer "Weighty spears would endanger the chief endangerer.' (ms. preiglyn) Figura etymologica. Behind both periglyn and periglawr is Lat. periculum. The verb is denominative, periglawr is formed as if from Lat. **pericul-arius. Pwys < Lat. pensum. 401. BA 21.2, CA 11. 1008-9, A82.3-4, GOSP 149, J 1. 776. tavloywac ysgeth tavle[i] wydrina med T. with?form throw.ipf.sg. glass of mead Tafloyw with a ?shrug would knock back a glass of mead.' (ms. tavlet) Cf. CA 309. Williams suggests the emendation to give a cross-rhyme with tavlei in the next line. But the problem with tavlet is rather syntactic: the verbal form has two nominal arguments, tavloyw and wydrin a med, which an impersonal tavlet could not have, since the canonical subject argument, here the agent tavloyw, is by definition demoted from subject position, and usually faded out altogether, in impersonal syntax. The argument structure of this clause therefore requires personal morphology of the verb. On ysgeth, cf. inst. 102, 1. 36. Jackson has 'in style', Jarman 'with a flourish': my suggestion is a guess, to add to the possibilities.

245

Tavloyw is simultaneously subject of a second instance of tavlei, next instance.

402. BA 21.2-3, CA 11. 1008-9, A82.3-4, GOSP 149, J 11. 776-77. tavloyw ... rac teymed tavlei vedin T. before princes throw.ipf.3sg. army Tafloyw before princes would overthrow an army.' Cp. preceding instance. The passage shows tight syntactic parallelism, while playing on the meaning of tavlei: tavloyw:

ac ysgeth rac teymed prep. NP

tavlei wydrin a med tavlei vedin V NP

Note how the position of the subject proposed is essential for the full parallelism: if it had appeared in its canonical position immediately after the verb, the first clause would have read ac ysgeth tavlei tavloyw wydrin a med, the second, rac teymed tavlei vedin. The position of the subject tavloyw in the first clause would break up the parallelism with the second, where tavloyw would not be repeated.

GK: inst. 403 = 1

403. BA 27.17-18, CA 11. 1390-92, GK.54-56, GOSP 156, J 11. 1101-2. Gwarchan kynvelyn ar ododin G. K. on G. neus goruc o dyn dogyn gymhwylleit ptl. make.prt.sg. from man plenty mentions "The Gwarchan of Cynfelyn appended to the Gododdin made plenty of mentions of the man.' This seems to be a sort of commentary on the content of the poetry. The man' is presumably Cynfelyn himself. Note the use of dyn 'duine, Mensch, homo, ανφρωττος' for the hero as opposed to the usual gwr 'fear, Mann, vir, άνηρ'. This too suggests that these lines do not strictly belong to the poetry itself. The phrase Gwarchan kynvelyn ar ododin appears unambiguously topicalised Note the use of the proleptic object pronoun -s referring to the nominal object dogyn gymhwylleit separated from the verb (extraposed) by the preposition phrase o dyn.

246

[S+pro]V[O-pro] instt. 404-34 = 31 A: instt. 404-22 = 19 404. BA 2.2, CA 1. 42, A4.4, GOSP 117, J 1. 52. ef gwrthodes gwrys he repell.prt.sg. attack 'He repelled the attack.' On gwrys, cf. CA 80, Loth (1923: 370), Geirfa 722, GPC 1741. Gwrys perhaps < *wrt-to- < *wert- 'turn' IEW 1156-58, cp. Lat versus 'against' < vorsus < *wrt-to-; also Olr.yh', accented in composition frith-, W gwrth, wrth (cf. VKG 44, L&P 5, GOT 515, Hamp 1953; wrong in WG 406-7). Or gwrys might be < verbal abstract *wrt-tu-. 405. BA 3.20, CA 1. 99, A12.3, GOSP 120, J 1. 109. wy gwnaethant en geugant gelorwyd they make.prt.pl. ptl. certain biers 'They inevitably made biers.' That this was inevitable was due to the fact that they were so fierce and unstoppable in cutting down the enemy. 406. BA 4.7, CA 1. 116, A3.12, GOSP 120, J 1. 126. ef lladei saesson seithuet dyd he kill.ipf.sg. Saxons seventh day 'He used to kill Saxons on the seventh day.' Seithuet dyd 'seventh day' seems to be an idiomatic expression, generally interpreted as meaning 'continually, regularly', cf. CA 104, Jarman 86. Could the implication be 'even on the seventh day of the week'? That is, even though he had been killing Saxons for the previous six days of the week, he maintained his action on the seventh day as well. Thus the meaning 'continually' is upheld. Alternatively, there might be a reference to the Sabbath: even when others were resting, he would not let up. 407. BA 7.3, CA 1. 246, A22.4, GOSP 125, J 1. 251. ef gwneei arbeithing peithyng aruodyawc he make.ipf.sg. on destruction destruction blow-dealing Ήβ would make blow-dealing destruction upon destruction.' Aruodyawc 'blow-dealing' is hapax legomenon but transparent < aruot (< arf + -od GPC 194) + yawc Geirfa 39, GPC 195. Onpeithing, peithyng, cf. CA 139-40, Jarman 145, R ssel (1991: 108). ί. ΒΑ 8.21-22, CA 1. 345, A30.4, GOSP 129, J 1. 330. efllithyei wydgwn oe anghat he feed.ipf.sg. wild-dogs from-his hand 'He would feed wolves with his hand.'

247 The meaning is figurative: he gave the wolves food by leaving many corpses of the enemy on the battle-field. Ongwydgi, pi. gwydgwn, cf. Geirfa 731, GPC 1756. 409. BA 9.20, CA 1. 378, A33.7, GOSP 130, J 1. 362. ruuawnhir efrodei eur e allawr R. tall he give.ipf.sg. gold to altar 'Rhufawn the tall, he used to give gold for the altar.' 410. BA 10.17, CA 1. 419, A37.2, GOSP 132, J 1. 403. ef diodes gormes he drive-out.prt.sg. oppressor 'He drove out the oppressor.' 411. BA 10.17, CA 1. 419, A37.2, GOSP 132, J 1. 403 efdodes fin he place.prt.sg. border 'He set up the border.' I.e. by his heroic actions, he established the border across which the enemy knew it could not come. 412. BA 11.7, CA 1. 456, A39.10, GOSP 133, J 11. 433-34. e m[i]t efkrennit e gatQwaewawr in battle he shake.prs.sg.abs. his battle-spears 'In battle he brandishes his battle-spears.' (ms. myt) Bit < PIE *bkih2-ti- nomen actionis to ^eih?- 'strike1 IEW 117-18, of which Olr. benaid < PIE *b^i-ne-h2-ti is the nasal present: Olr. bith, cognate with MW bit, is the verbal noun of this. 412. BA 15.6-7, CA 1. 414, A57.8, GOSP 139, J 1. 398. eflladei oswyda llavynllymaf he kill.ipf.sg. enemy with blade sharpest "He would kill the enemy with the sharpest blade.' 414. BA 17.2-3, CA 1. 803, A65.4, GOSP 142, J 1. 613 (B-variant). ef gorev vwyt y ysg[ly[vyon he make.prt.sg. food to birds-of-prey 'He made food for birds of prey.' Cp. inst 298,1. 811 (B-variant). 415. BA 19.15-16, CA 1. 952, A76.4, GOSP 147, J 1. 732. ef dodes e dilis yg kynhor he place.prt.sg. his steadfast in van "He placed his steadfast men in the van.'

248 On dilis, dilys, cf. Geitfa 354-55, GPC 1016, Cp. Olr. dttes 'proper, permissable, faithful, reliable1 OIL 213. Both < Celt. *di-liss-o-l-a-. The root is attested uncompounded in W llys^ 'challenge to a witness or a juror because of some legal impedimement' GPC 2276. This is connected tentatively by GPC with Olr. les «-stem 'benefit, advantage' DIL 429, but the meaning is quite inappropriate, and the latter is transparently cognate with W lies 'benefit, advantage1 GPC 2156. 416. BA 19.16, CA 1. 953, A76.5, GOSP 147, J 1. 733. efdodes rac trin tewdor he place.prt.sg. before battle thick-door 'He set up a stronghold in the face of battle.' On the meaning of tewdor, cf. CA 299. 417. BA 20.13-14, CA 1. 993, A80.6, GOSP 148, J 1. 761. ef dodes e gledyf yg goeithin he place.prt.sg. his sword in fierce 'He struck with his sword fiercely.' On goeithin Tierce, cruel, bitter', cf. CA 154-55, Geirfa 544, GPC 1428. Jackson translates, 'in the fierce fight', but the interpretation adopted by Jarman, with yg goeithin adverbially is more likely. 418. BA 21.1, CA 1. 1006, A82.1, GOSP 149, J 1. 774. Efgwrthodes tres tra gwyar()llyn he repell.prt.sg. attack over blood-lake 'He repelled the attack over a lake of blood.' On tres, cf. insL 172,1. 1267. 419. BA 21.1-2, CA 1. 1007, A82.2, GOSP 149, J 1. 775. ef lladei val dewr dull ny[ JtQechyn he kill.ipf.sg. like brave array neg. flee.ipf.pl. 'He would kill like a brave hero an array which would not flee.' 420. BA 21.13, CA 11. 1036-37, A84.9-10, GOSP 150, J 11. 810-12 efda[ jdodes ar lluydpwys ar lies rieu he good place.prt.sg. on hosts weight on benefit kings "Well did he put pressure on hosts for the benefit of kings.' On the syntax and meaning of the adjective incorporated adverbially in the verb, cf. inst. 246,1. 619. 421. BA 22.15, CA 1. 1107, A87.7, GOSP 151, J 1. 1003. ef gelwi gwn gogyhwc he call.ipf.sg. dogs ?swift 'He would call his swift dogs.' On gogyhwc, cf. CA 239, Geirfa 551, GPC 1442. The meaning is a reasonable guess, but the form is etymologically opaque.

249 422. ΒΑ 22.16, CA 1. 1109, Α87.9, GOSP 151, J 1. 1005. ef lledi bysc yng corwc he kill.ipf.sg. fish in coracle 'He would kill a fish in a coracle.1 Pysc < Lat piscis is the old singular. In Modern Welsh, a singulative pysgodyn formed from the plural pysgod is used. B: instt. 423-32 = 10 423. BA 31.9-10, CA 11. 273-74, B11.6-7, GOSP 102, J 1. 266 (B-variant). ef gwnaei wyr llydw a gwraged gwydw he make.ipf.sg. men low.pl. and women widowed.pl. 'He would make men low and women widows.' The Α-variant hasgoruc wyrllfyfdw

... etc., inst. 223,1. 264.

424. BA 31.18-19, CA 1. 251, B13.2, GOSP 103, J 1. 256. efdilydei win gwr eurQdorchawc he deserve.ipf.sg. wine man gold-torqued 'He deserved wine, the gold-torqued man.' Without the initial pronoun, this would look like a VOS construction, with right dislocated subject. 425. BA 31.19-20, CA 1. 252, B13.3, GOSP 103, J 1. 257. efrodei gloyw dull glan y gwychauc he give.ipf.sg. shining array bright to brave 'He would give a bright, shining array to the brave.' The meaning seems to be that the hero contributed a fine troop of his own men to the already existing contingent of brave warriors, although Jackson takes y as the article ('the brave one', an epithet of the hero): this seems less attractive. An alternative would be to take y as a reflex of OW di 'from': Ήβ would give a bright, shining array of the brave' i.e. Of brave warriors'. 426. BA 32.18-19, CA 1. 1152, B17.5, GOSP 105, J 1. 891. pan wanet yg kyueillt ef gwanei ereill when strike.prt.ips. my comrade he strike.ipf.sg. others 'When my comrade was struck down, he would strike down others.' The topicalising role of the pronoun is obvious in this instance, promoting the argument from object in the first clause to subject in the second On the temporal clause, inst. 292. 427. BA 34.17.18, CA 11. 1190-91, B26.5-6, GOSP 107, J 11. 925-26. em ladaut lu maur i[ ]guert i adraut he kill.prt.sg. host great in worth his relate.vn. "He killed a great host to the value of his repute'.

250 I.e. 'to achieve his reputation'. 428. BA 37.18-19, CA 1. 1238, B38.2, GOSP 112, J 1. 968. efladhei a [pherjuetac eithaf he kill.sbj.ipf.sg. and centre and wing 'He would kill both the centre and the wing (of the enemy army).' (ms. a uet) The subjunctive -h- may be motivated by the sequence of tenses/moods in these lines: Ef guant tratrigant echassaf ef ladhei a [pher]uet ac eithaf

'He charged ...' (pit.) 'so that he might kill...' (sbj.ipf.)

I.e. the second clause may have a final sense, although the two clauses are strictly asyndetic. 429. BA 38.5, CA 1. 813, B39.6, GOSP 113, J 1. 633. em dodes i[ ]tu ar guaiu galon he place.prt.sg. his side facing spear enemies "He placed his side facing the spear of the enemy.' The preposition or < *are here clearly has the meaning 'facing, opposite, before', later taken over by roc when ar fell together with OW guor, guar On'. 430. BA 38.6-7, CA 1. 815, B39.8, GOSP 113, J. 1. 635. wy guenint ... seiuogion they strike.ipf.pl. opponents 'They would strike their opponents.' The pronoun wy is simultaneously subject of both asyndetic verbs guenint and lledint (next instance). Seiuogion is hapax legomenon but morphologically transparent < saf- 'stand' < PIE *steh2- with suffix *-m-) + adjectival -(y)awc + pi- -ion. Thus [[[seiu- Basel -ft)og- Adj.] -ton Subst.pl.] 'those standing', i.e. 'those standing to oppose' > Opponents'. 431. BA 38.6-7, CA 1. 815, B39.8, GOSP 113, J. 1. 635. wy ... lledint seiuogion they kill.ipf.pl. opponents "They would kill their opponents.' See preceding instance. 432. BA 38.16-17, CA 11. 1248-49, B41.4-5, GOSP 113, J 1. 978-79. efguisgus aur ig kinnor gaur he don.prt.sg. gold in van battle 'He wore gold in the van of battle.'

251

GT: inst. 433 = l 433. ΒΑ 25.16-18, CA 11. 1294-96, GT.37-39, GOSP 154, J 11. 1037-39. Efgwrthodes ...arlles pedyt petwarlliwet he repell.prt.sg. on benefit infantry four army 'He repelled four armies for the benefit of the infantry.'

Pedyt < Lat. pedites. There is some difficulty with the reading of the omitted phrase; ar ttwrw penes in Gwenogvryn Evans' text, and so printed by Williams; cf. CA 356. But this does not interfere with the present syntactic analysis. Jarman prints a conflation ar lies penes. GA: inst. 434 = 1 434. BA 26.10-11, CA 1. 1326, GA.5, GOSP not trnsl., J not trnsl. A garwn yefcarei anrheith(gar) rel. love.ipf.lsg. I he love.ipf.sg. spoil 'Whom I loved, he loved spoil.' In order to achieve rhyme with dwyweith, weith etc., -gar must be omitted, cf. CA

359. The whole sentence forms a nominativus pendens construction, with topicalised A garwn y and resumptive pronoun ef. In Evans (1908: 26), 1. 10 is wrongly labelled as 1. 12.

V[O-pro][S-pro] instt. 435-53 = 19 (There are no instances where either S or Ο is pronominal.) A: instt. 43549 = 15 435. BA 2.15-16, CA 1. 60, A6.4, GOSP 118, J 1. 70. mab botgat gwnaeth gwynnyeith gwreith e law son B. make.prt.sg. vengence deed his hand "The son of Botgad, the deed of his hand wrought vengence.' Gwnaeth (cf. note 2, s.v. gwneuthur) and gwreith < *wreg-tu- (old verbal noun) constitute an obscured gura etymologica. Gwynnyeith < Lat. vindicta. Note the topicalisation of the possessor mab botgat, resumed by the pronoun e 'his1. 436. BA 3.14-15, CA 11. 90-91, All.1-2, GOSP 120, J 11. 100-1. Gwyr a aeth gatraeth gan wawr dygymyrrws eu hoet eu hanyanawr men rel. go.prt.sg. C. with dawn shorten.prt.sg. their age their natures The men who went to Catraeth with the dawn, their mettle shortened their lives.'

252 On anyan 'nature, (inherent) quality1, cf. Geirfa 30, GPC 139 (s.v. aniari) < *andegano- < PIE *n*/ 'around') reciprocally, as in ymladd 'fight' < 'strike, cut each other'. Jackson apparently adopts the same suggestions,"... cast at each other'. In the first place, I prefer to retain the manuscript reading: the future (or really present here?) can have an enlivening effect on the situational content in a historic usage. Secondly, I think em- may have its etymological spatial meaning here of 'around, about': thus, Ά hundred thousand and three hundred will be thrown around', that is 'felled, routed' or the like.

278 544. ΒΑ 6.6-7, CA 1. 208, Α19.7, GOSP 124, J 1. 218. aessawr dellt anibellt a adawei shields splinters shattered ptl. leave.ipf.3sg. ' Shields as splinters and shattered would he leave.' On dellt 'laths, rods' figurative 'splinters', cf. Geitfa 312, GPC 912 < *del-to- < PIE *del(H)- 'split' IEW 194-96, cp. Gk. δελτος 'writing block' (< 'a split off piece of wood'), OEfdtf, Germ. Zelt 'tent' (from the laths used to support it). Onan(n)ibellt 'broke, shattered', cf. CA 130, CL1H 132-33 (note on pell), Geirfa 29 (but wrongly translated as '?grymus, cadarn'), GPC 139, and cf. 954 (s.v. dibellt 'unhurt') Evans (1908: 6) prints ambelk, but the manuscript clearly has anibellt. 545. BA 6.8-9, CA 11. 212-13, A19.11-12, GOSP 124, J 1. 222-23. tnab syvno ... a werthws e eneit son S. ptl. sell.prt.sg. his life 'The son of Sywno sold his life.' Cf. inst. 456. Apparently a canonical 'abnormal' sentence. 546. BA 9.12, CA 1. 366, A32.4, GOSP 130, J 1. 350. nyt mabmam ae maeth neg. son mother ptl.-3.O nurture.prt.sg. 'No mother's son nurtured them.' Jackson translates, "No mother's son succoured them', Jarman, *No mother's son has cared for them'. Williams is inconclusive as to the interpretation (CA 166). Jarman (100) notes, The meaning 'cared for' or 'succoured' ... is unusual'. There seems to be no reason to doubt that maeth here refers to the 'rearing' or 'adoption' by a lord of his warrior company, but the sense of this instance is elusive. It is tempting to think of mob mam as 'mummy's boy': it was no feeble mummy's boy that nurtured them, but a powerful and harsh lord. Koch's (1991: 21-22) translation, "No son that mother nurtured', implies a similar idea: it was no mother that nurtured them (him in Koch's interpretation). However, Koch's interpretation rests on the dubious analysis of mam ae maeth as a whole as a relative clause, i.e. that mam is not (part of) the antecedent, but is itself the focussed subject of the relative clause. Such a construction is not consistent with Welsh grammar. It is part of Koch's argument that it is an archaic construction, but for this to be acceptable, we must accept the whole of Koch's historical argumentation as to the origin and development of Welsh syntax, which is all very speculative and open to debate. I hesitate to fell back on emendation, but perhaps after all the original reading was nyt mam ae maeth "No mother nurtured them', miscopied by a scribe who did not subsequently notice his mistake.

279 547. ΒΑ 12.7-8, CA 1. 537, Α44.3, GOSP 134, J 1. 990. penn gwyr tal being a dely head men end bench ptl. deserve.prs.sg. The chief of men, he deserves the end of the bench.' On the significance of the end of the bench, cf. inst. 194,1. 400. Penn gwyr 'chief of men' is topicalised. 548. BA 14.22, CA 1. 671, A54.4, GOSP 139, J 1. 536. seith gymeinto loegrwys a ladassant seven as-many of Englishmen ptl. kul.prt.3pl. They killed seven times as many Englishmen.' 549. B A 14.22-15.1, CA 1. 672, A56.5, GOSP 139, J 1. 537. ο gyvryssed gwraged gwyt[w] a wnaethant from conflict women widowed.pl. ptl. make.prt.3pl. 'By conflict they made widowed women.' (ms. gwyth) The emendation is necessary for sense, cf. CA 240. 550. BA 15.21-22, CA 1. 698, A60.4, GOSP 140, J 1. 548. gwin a med a[ ]mall a amucsant wine and mead and malt p . contend.prt.3pl. 'It was wine and mead and malt that they contended for.' Jarman translates, They enjoyed ...', adopting Williams' suggestion (CA 249-50) that this is a verb meaning 'enjoy* (cf. I. Williams 1940), distinct from amwyn 'defend, fight for'. Cf. inst. 261,1. 712. But Jackson is dubious and translates, They set hands to ...'. The line makes perfect sense with amwyn in its familiar meanings: the line captures the dual image of the warriors contending for their drink in the feasting hall, and then contending on the battle-field to be worthy of what they had drunk. I suggest Williams' amwyn 'enjoy1 be regarded as a ghost (not registered in GPC). 551. BA 15.22-16.1, CA 1. 700, A60.6, GOSP 140, J 1. 550. a rwy(f) a golleis om gwir garant and too ptl. lose.prt.lsg. from-my true kinsmen 'And too many did I lose of my true kinsmen.' A scribe has added an -/to rwy, thinking of rwyf 'lord', but that obscures the syntax, and there is no doubt that rwy 'too much, too many' is the correct reading, in the light of the B-variant (inst. 560,1. 706). 552. BA 16.18, CA 1. 793, A64.4, GOSP 141, J 1. 621. ruduedel ryuel a eiduni red-reaper war ptl. desire.ipf.3sg. The red reaper, it was war that he desired.'

280 Ruduedel 'red reaper', a sort of kenning for the warrior hero, is topicalised.

553. BA 21.15-16, CA 1. 1042, A84.15, GOSP 150, J 1. 821. gereintrac deheugawr a dodet G. before south shout ptl. place.prt.ips. 'For Geraint, before the men of the South, a battle-cry was raised.' This is the only way that gereint can be incorporated into the sentence and its context This seems then to be an instance of a dative without preposition, a rather archaic phenomenon. Expected would be y ereint, or yr (les) gereint.

554. BA 22.6, CA 1. 1068, A86.14, GOSP not trnsl., J not trnsl. cann calan a darmerthei hundred Calends ptl. prepare.ipf.3sg. Ήε would make preparations for a hundred New Year's feasts.' The line is certainly a corrupt copy of a sentence like the B-variant (inst. 328, 1. 1091). This is why it is untranslated by Jackson and Jarman. However, as the line stands, it is meaningful, and can be thought to indicated the great wealth and generosity of the hero in question, who, though he would never see a hundred New Year's feasts, nevertheless would provide enough food and drink as if to last for a hundred years.

B:instt: 555-63 = 9 555. BA 30.12, CA 1. 634, B7.1, GOSP 101, J 1. 526. Da d[o]uot adonwy (adonwy) am adaussut good gift A. ptl.-Isg.O promise.plpf.2sg. 'It was a fine gift that you had promised me, Addonwy.' (ms. dyuot) Cp. instt. 135, 38,1. 627 (Α-variant). Cf. CA 227. On douot 'gift, prize, treasure', cf. Geirfa 425 (s.v. dyouot), GPC 1072 (s.v.

dofod),

PKM 153-54. Jackson translates, 'Good was the fortune, Addonwy, Addonwy, which you had promised me'. The translation of douot as 'gift' seems more appropriate to me in the context: what the gift was is specified in the following line, a wnelei vratwen gwnelut lladia llosgut 'What Bradwen would do you would do, you would kill, you would burn'. Jackson's syntactic interpretation (da as predicate in a nominal sentence, dfojuot

...am

adaussut as subject, am adaussut as relative clause with dfojuot as antecedent) is in principle plausible. It is hard to believe that the repetition of adonwy is part of the original text and not a scribal slip.

281 556. ΒΑ 30.16, CA 1. 1126, Β8.1, GOSP 101, J 1. 865. TryQcan eurdorch[auc] a gryssyassant three-hundred gold-torqued ptl. rush.prt.3pl. "Three hundred gold-torqued warriors rushed forth.' It is not inconceivable that the manuscript reading eurdorch could be construed as a bahuvnhi compound, One having a golden torque', as in red-head One having a red head', but such an instance would be unique in the corpus, and cp. the first line of the following awdl, Trycant eurdorchauc, so it seems necessary to read the adjectival form eurdorchauc here too. The line appears to be a canonical instance of the abnormal sentence. 5J7. BA 31.5, CA 1. 1145, B10.6, GOSP 102, J 1. 884. ac en dyd camavn camp a wne(e)i and in day battle feat ptl. do.ipf.3sg. 'And on the day of battle he would perform a feat (of strength).1 The verbal form must be read as wnei, disyllabic, to rhyme with eli, lestri etc. The conditions for focus arise from the contrast in imagery with the preceding line, erfyjvessit gwin gwydyr lestri llavn (insL 290), Ήε drank wine from full glass vessels (in court), and on the day of battle it was a feat of strength that he would do'. 5J58. BA 31.6-7, CA 1. 1147, B10.8, GOSP 102, J 1. 886. calaned cochwed a[ ]e()deui corpses red-hued ptl. leave.ipf.3sg. ' Blood-stained corpses would he leave.' 559. BA 32.22, CA 1. 312, B18.1, GOSP 105 (emended), J not trnsl. geu ath diwedus tutleo false ptl.-2sg.O say.prtsg. T. Ά falsehood Tudlew told you.' The A-variant (1. 300) reads Bu gwir mal y meud e gatlew, which is very corrupt and will not be included in the list of instances, though something like Jarman's Bu gwir mal amceuddai Gadlew probably lies behind the manuscript reading. Koch's (1985a: 453, 1987: 275) 'restored' reading Ennuir amm-coudanl canllou 'Most truly have the songs told ...' departs rather radically from the manuscript reading. 560. BA 33.12-13, CA 1. 706, B20.4, GOSP 106, J 1. 550 (A-variant). rwy (e) ry golleis y om gwir garant too ptl. lose.prt Isg. I from-my true kinsmen ' Too many have I lost of my true kinsmen.' Cf. CA 251 on the spurious e before ry: it represents scribal confusion over the form of the perfective particle ry, yr at a time when its use as such was moribund (possibly already extinct in the spoken language). Cf. GMW 62-63,169. Cp. inst 551,1. 700 (Avariant), where ry is replaced altogether by a.

282 561. ΒΑ 35.5, CA 1. 1211, B27.3, GOSP 108, J 1. 946. bubon a guoreu barQdeo B. ptl. do.prt.sg. fury-thick ' Bubon acted, full of fury.' The focus arises as a result of the need to specify the identity of Uiefraidus leo '?eager lion-hero' of the preceding line. 562. BA 35.16, CA 1. 849, B29.7, GOSP 109, J not trnsl. diu guener calanet a ciuriuet Friday corpses ptl. count.prt.ips. 'On Friday the corpses were counted.' Conditions for focus arise out of the contrast with the content of the previous line (inst. 331), diuyeu cennadeu amodet 'On Thursday envoys (topicalised) were pledged' (peace negotiations), but On Friday, it was the corpses that were counted' (the result of war). While it is certainly the case that the distinction between the topicalising construction of cennadeu amodet and the the focussing construction of calanet a ciuriuet is fine, and the manuscript reading not always 100% dependable, nevertheless, taking the reading as it stands, I believe the distinction is valid, and is part of the grammatical apparatus at the disposal of the poets. 563. BA 35.17-18, CA 1. 851, B29.9, GOSP 109, J 1. 657. diu sul laueneu nid a atran[n]et Sunday blades red ptl. share.prt.ips. 'On Sunday red blades were meted out' (ms. atranhef) The -A- in the manuscript reading atnmhet can have neither grammatical (there is no 'preterite subjunctive') nor etymological (the base is rann not **rant) significance. It therefore probably derives from a miscopying of a second -n-. GT: inst 564 = 1 564. BA 25.7-8, CA 11. 1272-73, GT.15-16, GOSP not tmsl., J not trnsl. ac eivyonyda gwynhei dyd kein edryssed and E. ptl. lament.sbj.ipf.sg. day fine excess 'And Eifionydd (LN) was to lament the day of fines excess.' The day of fine excess' may be interpreted as the day on which the hero performed his great feats in battle and on which he also fell, cf. CA 352-53. On edryssed, further, Geirfa 441, GPC 1168. Apparently an abnormal sentence, as no conditions for contrastive focus are discemable to me. The context is, however, quite obscure.

283

GK: inst. 565 = l 565. ΒΑ 28.4-5, CA 11. 1408-9, GK.72-73, GOSP 156, J 11. 117-18. Mab coel kerth vyg werth y a wnaethant son prophecy true my value I ptl. make.prt.3pl. ο eur pur a dur ac aryant from gold pure and steel and silver 'The true son of prophecy, my value did they make from pure gold and steel and silver.' This has been interpreted as referring to a story of the poet's ransoming from captivity, CA 373, GOSP 22-23, Jarman 156, Owen (1978: 137). But I suspect this may properly be a reference to the contract between the poet and his patron(s), in that the heroes paid for the eulogy they were to receive from the poet, in the form of valuable gifts (gold and silver) and their actions on the battle-field (with steel weapons).

instt. 566-97 = 32 A: instt 566-81 = 16 566. BA 3.10-11, CA 11. 84-85, A10.1-2, GOSP 119, J 11. 94-95. Gwyra aeth gatraethgan wawr trauodynt eu hed eu hovnawr men rel. go.prt.sg. C. with dawn move.ipf.pl. their abode their fears This is a surprisingly difficult instance, cf. CA 94-95. The meaning of trauet is 'move about, throw about', cp. Ny byd myssoglawc maen ο vynych trauot Ά rolling stone gathers no moss' (lit. Ά stone is not wont to be mossy by being frequently moved about") Phillimore (1886: 144), I. Williams (1926a: 24), Hay (1955: 182). In isolation one might take Gwyr a aeth gatraeth gan wawr as the topicalised subject of trauodynt (giving an SVO structure), They would shift their abode, (and) their fears (away from themselves)', in the sense perhaps of, They threw off their abode, their fears', no inappropriate comment on the actions of heroes fearlessly giving up everything to go off to battle. However, the superficial parallelism with 11. 90-91 (insL 436) would seem to force a VOS interpretation, with eu hovnawr as subject and Gwyr a aeth ... etc. as topicalised possessive phrase, resumed by the possessive pronouns eu. Williams notes (CA loc.cit) that the implied translation, Their fears disturbed their peace' is surprising in the heroic context, though Koch (1988: 34) adopts it. More heroic is the standard interpretation adopted by Jackson, Their fears shifted their abode (?)', and Jarman, Their fears departed from their dwelung place', following Williams. This interpretation

284 requires that the two possessives have different referata: the eu of eu bed refers to eu hovnawr, while the eu of eu honvawr refers to the Gwyr a aeth... etc.: [The men who went to Catraeth with the dawn];, [theiri fears]j would shift theirj abode. This is to imply that there was no place for fear in the hearts of the heroes. A problem arises here if the parallelism with 11. 90-91 is taken seriously, where both possessives refer to Gwyr a aeth... etc. But if the structural parallelism may be broken in that respect, then perhaps it could be broken to allow the alternative SVO interpretation suggested above. Or perhaps we overestimate or misjudge the heroic content of the poetry, so that the rejected Their fears disturbed their peace' is correct after all. I am not sufficiently confident of any of these possibilities to include the instance unambiguously in the classification.

567. BA 5.8, CA 11. 152-53, A17.3-4, GOSP 122, J 11. 162-63. gwledicgwd gy[n]gein nef enys brydein lord where contain.prs.sg heaven island Britain (ms. gyfgein) Jackson translates, "Where is the prince to be found, the lord of the island of Britain?', apparently emending ««/'heaven' to na/"'lord'. Jarman translates, 'Where can be found the lord of the heaven of the island of Britain?1, retaining nef. In his note (CA 119), Williams gives two possible translations, 'Pie mae arglwydd nef Ynys Brydain i'w gael?' (essentially the translation adopted by Jarman), and, 'Arglwydd, pie ceir nef ynys [sic] Brydain?' ('Lord, where is the heaven of the island of Britain found?'). All these suggestions are plausible, but which is correct is open to question, and is crucial for the syntactic analysis. Furthermore, on the assumption that the correct translation is one of those given, the significance of the line is obscure to me. I leave it unclassified therefore.

568. BA 6.7-8, CA 11. 209-10, A19.8-9, GOSP 124, J 11. 219-20. pareu rynn rwygyat dygymynei e gat spears straight ?slasher cuLipf.sg. to battle Jackson translates, 'Slashing at them, he cut off tough spears in battle', Jarman, Threatening spears the tearer cut down in battle'. From Jackson's translation it is unclear how he interprets the syntax, but Jarman's translation clearly reflects an interpretation as a structure, OSV, without any resumptive pronouns, unique in the corpus. This may be correct, but alternatively, rwygyat might be an adjectival formation from the verb nvygyaw 'rip, slash' with the rare suffix -ad, as in crwydrad 'wandering, stray' GPC 617). Pareu rynn rwygyat 'straight slashing spears' might be the object of dygymynei, giving an OV construction.

285 569. ΒΑ 7.9-10, CA 11. 266-67, Α23.9-10, GOSP 126, J 1. 267. greit uab hoewgi[ ]r()ac ysberi y beri creu G. son H. before spears to cause, vn blood 'Graid son of Hoywgi caused blood (to flow) before (his) spears.' My interpretation differs from that of previous commentators, though it is in principle implied by some of the suggestions in Williams' notes (CA 143). The difficulty is in the interpretation of the phrase y beri creu. Jackson translates, 'Graid son of Hoewgi used to form the battle-pen in the presence of spears', Jarman, 'Graid son of Hoywgi formed a battle-pen against the spears'. Both follow Williams in understanding beri as ipf.Ssg. of the stem par- 'cause' < PIE *kwr- < *kwer- 'make, form' IEW 641. The preceding y is regarded as a corrupt copy of the particle ry. And creu is interpreted as 'pen, fold' Geirfa 174 (s.v. creu1), GPC 582 (s.v. emu2). For the interpretation of beri as imperfect speaks the lenition of greu in the B-variant, 1. 276, but I do not have such confidence in the accuracy of the manuscript tradition with regard to mutations to depend on it here. Both A and B-versions have y beri (not ry), and I feel this is quite comprehensible as it stands, with beri (lenited) as the verbal noun of the same verb, preceded by the preposition y 'to' in the construction described by Evans (1950: 188-89): cp. Olr. ar is bes leosom indaim dothuarcain indarbe Tor it is the custom amongst them for the oxen to tread the com' Wb. 10^6; CA 1. 73 dadyl dieu agheu y eu treidu (non-finite main clause) Ά certain tryst of death met them'. In principle there is nothing to choose between "blood1 and "battle-pen1 for creu, except that the latter meaning depends on the use of the Irish cognate cro DIL 159, whereas creu "blood1 is quite unproblematically attested in Welsh, and extensively in the present corpus. I therefore prefer "blood". The relevant effect of this interpretation in the present context is to exclude it from this survey of finite syntax. 570. BA 10.18, CA 1. 421, A37.4, GOSP 132, J 1. 405. huteffyt y wrhyt e Iwry elfin ptl. ?.prs.3sg. his courage to path E. Jackson, '..., his courage like Elffin1, Jarman, He pressed forward with courage, like Elffin1. Effyt is hapax legomenon, and any translations are conjectural. For possibilities, cf. CA 178, Jarman 103-4, the best of which is perhaps to take the verb (unaffected * ffet) as a loan of Lat. appeto "try to reach, attack, desire". A possibility which does not appear to have been mentioned is a compound verb *qffat < * ad-pot-, with the same *pat- as seen in diaspat 'cry, scream", cf. I Williams (1950c, d, e), *pat- < Celt. *kw