The Third Grammatical Treatise

Vernacular Alphabets in the Auraicept na nÉces and The Third Grammatical Treatise

Nicolai Egjar Engesland

Masteroppgave i norrøn filologi Institutt for lingvistiske og nordiske studier UNIVERSITETET I OSLO

Høsten 2016

The main supervisor of this thesis has been Jan Erik Rekdal, professor of Irish language and literature, and the co-supervisor has been Jon Gunnar Jørgensen, professor in Old Norse philology, both at the Department of Linguistics and Scandinavian Studies, Faculty of Humanities, University of Oslo.

Vernacular Alphabets in the Auraicept na nÉces and The Third Grammatical Treatise

Nicolai Egjar Engesland

Preface I wish to thank Professor Karl Gunnar Johansson for feedback and constructive criticism during the initial stages of the project and Professor Jon Gunnar Jørgensen for efficient guidance through the final stages of the project. Heartfelt thanks go to my main supervisor Professor Jan Erik Rekdal. Not only for overseeing this project, but for having introduced me to Irish language and literature in the first place and for being a constant source of motivation and support. I also wish to thank Dr. Mikael Males for indispensable aid through the different phases of this project from its inception in Reykjavík 2015. Our colloquia have provided a platform in which to try out thoughts and ideas, though the number of participants has for most of the time been fairly limited due to our eclectic bouquet of texts. This eclecticism is to some extent represented in the present treatise. Finally I would like to thank Ayman Razek, who has had to delve into at least partly unfamiliar subject-matter in order to prevent the present thesis from violating English idiom.

R. XIII Trastevere, November 6th 2016.

TABLE OF CONTENTS

Preface 7

1. INTRODUCTION 11

1.1. Preamble 11

1.2. Historical Backdrop 12

1.3. On the Vernacular Alphabets 15

1.4. Institutiones grammaticae 18

1.5. Auraicept na nÉces 20

1.6. The First Grammatical Treatise 25

1.7. The Third Grammatical Treatise 27

1.8. Considerations on Method 31

1.9. Theoretical Stance and Aim of Thesis 34

2. ANALYSIS 39

2.1. Linguistic Theory: The Letter 40

2.2. Conceptualizing the Minimal Parts of Speech 40

2.3. The Divisions of the Alphabet 45

2.4. The Accidents of the Letter 50

2.5. Graphemic Inventory 55

2.6. Vowels 56

2.7. Consonants 63

2.8. De ordine litterarum 71

2.9. The Opposition between the Native Grammarian and the Latinist 74

2.10. Excursus: Mythography and the Origin of the Vernacular Alphabet 76

3. CONCLUSION 87

3.1. Practical and Theoretical Utility: Grammar as Genre 87

3.2. Texts as Part of a Curriculum 90

3.3. History of Linguistics and the Later Tradition 91

4. BIBLIOGRAPHICAL REFERENCES AND ABBREVIATIONS 95

4.1. Manuscripts 95

4.2. Editions and Translations 95

4.3. Monographs and Articles 97

5. APPENDICES 103

5.1. Keys to Alphabets 103

5.2. On Quotations from Editions and Translations 104

1. INTRODUCTION

1.1. Preamble This thesis will aim at an assessment of Auraicept na nÉces and The Third Grammatical Treatise from a comparative point of view. These are two vernacular treatises, from medieval Ireland and medieval Iceland respectively, dealing with the fundaments of grammar and poetics. What place these texts occupy in the history of linguistics will to some extent depend on one’s personal take on the turn of events. This might be experienced by browsing the scholarly discourse they have generated. The story to be told here grants them a certain distinction, without, in that particular appraisal, putting too much emphasis on cause and effect. The texts are rather comprehensive. A qualitative comparison between them could lead in several directions and deal with (from a modern perspective) widely different subject- matter. For this reason and others I have narrowed down the scope so that focus is to be put on those units which, at least in medieval thought, might not be further divided: The letters of the alphabet. Discourse based on such fundaments should be able to sustain a superstructure. A comparative approach is able, or so it is hoped, to further substantiate claims as to the place of these vernacular grammatical treatises within the history of linguistics and ideas. A juxtaposition of this kind has to the best of my knowledge, excepting the occasional cursory remark, not hitherto been attempted. The treatises are in some important respects deceptively basic productions and have on this account attracted some criticism of doubtful historical relevance. Some of these will still have to be dealt with in due course during the following argument. At any rate their apparent simplicity might account for a certain lack of serious scholarly attention and determination to take them at face-value. Their thematical affinity and their place within cultural history entitle them to a certain unity in treatment. The Latin grammatical tradition is the obvious tertium comparationis. For several reasons I have found it convenient to let Priscian represent this tradition. First of all he aimed at comprehensiveness, so that his work encompasses the different trends of the art of grammar and discusses different solutions to individual issues, thus allowing ease of reference. Secondly his influence was profound on linguistic thought and discourse after his rediscovery during the Carolingian Renaissance. The third reason is, however, the most important one: He is the chief source for the relevant parts of the texts. This will be further specified when these are properly introduced.

The present thesis is in its essence an attempt to follow a precept given by the Irish scholar Sedulius Scottus (fl. c. 850) concerning the art of definition: Sed notandum est, quod qui definitionem alicuius rei definit, ita debet definire, ut dicat ipsam partem, quae communis est ei cum altera, et dicat specialem, quam propriam et quam solam habet (Sedulius Scottus; cited from Law 1994: 105n45).1 This is in many ways a sound approach, but one which has been curiously absent in the modern reception of the relevant texts. Medieval precepts will not otherwise influence the argument, which will proceed in the following manner: A brief historical introduction to medieval grammatica precedes a discussion on vernacular alphabets, which in turn is followed by sections on the sources (including an outlook to the First Grammatical Treatise) and their material state. This raises some issues that will be developed in a separate sub-chapter on the theoretical stance of the present thesis. The main chapter or analysis will be roughly divided into two parts: One will proceed by a close reading of the relevant passages while an excursus is devoted to mythography and the origin of the alphabets. The conclusion will provide a brief summary of the central matters and direct attention towards issues arising from analysis. By the end it is to be hoped that adequate answer has been given to the following set of questions: 1) Why does each treatise deal with the vernacular alphabet? 2) How is the vernacular alphabet presented in the one as compared to the other? 3) Could the choices made in one of the treatises shed light on those made in the other? Thus a proper ‘definition’ of the texts in question might be allowed.

1.2. Historical Backdrop Linguistic thought in the medieval period was a continuation of tendencies which were found in Late Antiquity. As early as the third and second centuries BC, Stoic grammarians based at the school of Pergamum had developed a tripartite division (as in Donatus’ Ars maior) which would later become popular with the Latin grammarians (Law 1982: 12). Quite a few vernaculars of Western Europe were employed as written languages at different points of time during the Middle Ages. Both Icelandic and Irish were in fact equipped with writing systems which were in all probability derived from alphabetic systems (McManus 1991: 4-5; Pedersen 1909: §4; Spurkland 2001: 16-17). These are themselves the results of linguistic analysis and thus proof of a certain linguistic competence on behalf of their

1 “Note that anyone who gives a definition of anything should define it in such a way that he indicates what it has in common with other things, and singles out for mention its unique characteristics,” (Law 1994: 105). The remark is found in Sedulius’ commentary on Donatus.

12 composers (Ahlqvist 1982: 7ff.; McManus 1991: 5). There is no doubt that cultural contact provided the incentive for the framing of these alphabets (McCone 1996: 22ff.); the Irish habit of making epigraphs on sepulchral or other monuments may itself have been occasioned by contact with the late Roman Empire (Ó Cróinín 1995: 35), These and allied matters are further discussed in the next sub-chapter (1.3.). The adoption of Roman Christianity in Western Europe had far-reaching consequences for the cultural life of the subordinated areas.2 The linguistic situation in Western Europe was no less complex than that in the East, but the languages of the West had no established literary culture prior to the influx of learning brought by the Church. Latin had already been the language of administration in those areas of Western Europe which were part of the Roman Empire. The conversion of previously illiterate people (illiterate in a Latinate sense) necessitated education in literacy and written culture. Christianity followed in the wake of the Roman empire, but reached beyond its former borders. Neither Iceland nor Ireland had first-hand acquaintance with Roman administration. The new literary culture appears to have been dominated entirely by Latin. However, subsequently also the vernaculars were reduced to writing. Eastern Christianity was early furnished with translations of Scripture from Greek into a multitude of languages such as Coptic, Arameic and Syriac. Armenian, Georgian and Gothic followed suit in the 4th century (Law 2003: 124). Alphabetic scripts were developed in order to reduce these languages to writing. In the West the situation was different. The Latin alphabet was adapted to the vernacular languages, and the native alphabets were not employed as book-scripts. Old Irish was written with the Latin alphabet by 600 AD (Ó Croínín 1995: 189) or at least before the middle of the 7th century (McCone 1996: 31). There is little evidence to suggest that the native systems of writing had any profound role in the adaptation of the Latin alphabet to the vernacular language;3 unlike, for instance, British influence on the pronunciation of Latin (McCone 1996: 31). Old Irish was reduced to writing by the use of the Latin letters and their current pronunciation (Ó Cróinín 1995: 190), or, more eloquently, ‘The written standard of Irish was cast in the mould of written Latin,’ (Rekdal 2012: 169): No additional letters were coined and none were taken over from the Ogam alphabet. The Latin mould ensured a quality ‘similar to the classical written medium’ (ibid.).

2 A comprehensive account of these matters is given by Law (2003: 124ff.). 3 The opposite view is put forth for instance by Mac Neill (1931) and challenged by McManus (1986).

Given that acquaintance with Latin language in this way preceded the vernacular literary pursuits, it is sensible not to think of a written vernacular as essentially oral language penned down on parchment. Against this background it is interesting to observe that the language of the extant Old Irish corpus is more or less bereft of dialectal features (Thurn. Gramm.: §16; McManus 1991: 148-50) – which makes it reasonable to regard it as an artificial literary language on a par with Latin. This is unparalleled in early medieval Europe (Ó Cróinín 1995: 189), although a tendency towards a Schriftsprache is seen in early English literature as a result of increased West Saxon influence. The Irish reception of the late antique and early medieval Latin grammatical works played an important role in the further life of these. This with regard to more or less all aspects: ranging from layout on the manuscript page to commentary and gloss. Speakers of non-Romance languages were in need of language resources with different emphasis and with more thorough description. The movement towards such emphasis is seen in the the Hiberno- Latin grammatical works, which are the subject of Law 1982. From this tradition emerged the innovatory attempt at grammatical description of the vernacular language. Latin was learnt in the grammarians’ schools. It is only reasonable to assume that Old Irish would be learnt in the same way. Writing in the 1140s, Petrus Helias acknowledges the possibility of creating a grammar of the French language and clearly states that such a work was not yet in existence (Björn M. Ólsen 1884: iv; Law 2003: 172).

Est autem gramatica composita in lingua greca, latina, ebrea, et caldea. Et possunt huius artis species crescere, hoc est plures esse, ut si grammatica tractaretur in gallica lingua (quod posset fieri facile, si tantum nomina et figure proprie illius secundum illam linguam inveniretur), sive in alia aliqua in qua nondum tractata est (Thurot 1868: 127).4

This is written at about the time the First Grammatical Treatise (1.6.) was composed (i.e. 1150 ± 25 - see discussion in Hreinn Benediktsson 1972: 23ff.). The influence of Petrus Helias and his commentary on Priscian on FGT has been discussed in scholarly reception of the work (Hreinn Benediktsson 1972: 198f.). An opposition between the common idiom and the artificial one (see 2.3.) is also commented on by Henry of Crissey (14th century). For Petrus

4 “For grammar has been composed in the Greek, Latin, Hebrew and Chaldean languages. And the species of this art could grow, that is, become more, for instance if grammar was discussed in the French language (lingua gallica) (which could easily be done, if only its individual nouns and figures could be found according to that language) or in another language in which it has not been treated yet.”

Helias the vernacular grammar remained a theoretical possibility. When he was active vernacular grammatical literature had been a practical reality in Ireland for several hundred years and was about to establish itself as a genre in Iceland as well.

1.3. On the Vernacular Alphabets A few remarks on the vernacular alphabets themselves are requisite.5 The dearth of evidence concerning their origins opens up for speculation, but there is no reason to venture beyond the merely synoptic at this point, beginning with the Ogam:

L'alphabet ogamique n'est sans doute que l'une des formes de ces alphabets chiffrés (18). Il est issu d'une sorte de transposition, ou de camouflage, de l'alphabet latin ; les classes de caractères ogamiques semblent inspirées par la classification des lettres proposées par les grammairiens latins (19). La technique employée (traits et points) s'inspire des encoches ou entailles utilisées dans la comptabilité rustique (Lambert 1987: 18).

Lambert singles out precisely those details which are called for in the context of the present argument. The Ogam alphabet is derived from the Latin alphabet. This must be true as far as its relationship to language is concerned, but as a signary in which distinction is made by position-marking, Ogam is unrelated to the Latin alphabet and possibly of pastoral origin, as mentioned above (see also Ahlqvist 1982: 7; Thurneysen 1937: 197). In the words of McManus the Ogam is a result of ‘stimulus diffusion’ (1991: 3); its main theoretical basis, namely the alphabetic principle, was provided by the Latin alphabet, and was given an independent expression. The Ogam alphabet consists of four primary groups (DIL: aicme ‘family’, ‘class’) of five letters. Those letters which denote vowels are distinguished graphically from the consonants. A certain influence from the doctrine of the Roman grammarians, who divided the consonants into three groups, namely semivowels, mutes and ‘Greek letters’, is therefore likely (Ahlqvist 1982: 8), although these groups of consonants do not correspond to those of the Ogam. No distinction is made, for instance, between mutes and semivowels.

5 The expression ‘vernacular alphabet’ is admittedly not quite satisfactory, as both terms might convey the wrong connotations; ‘vernacular’ in the context of this treatise simply means ‘that which is not Latin’ (i.e. used irrespective of high vs. low register and certainly not used in a derogatory fashion), while ‘alphabet’ is used of all systems of writing based on the alphabetic principle, viz. the principle that there is a one-to-one correspondence, though in reality this is a rare achievement, between grapheme and phoneme.

Several theories have been issued on the niceties of the process of adaptation (e.g. Carney 1975; Ahlqvist 1982), but they resort to arbitrary reshuffling of the letters and are therefore questionable. McManus has, moreover, amply demonstrated (1986: 39f.) that the manuscript key to the letters and values of the Ogam alphabet is not trustworthy as far as the original values are concerned. The fact that the manuscript key postdates the origin of the alphabet by several centuries, in which drastic language change had taken place, is in itself a serious objection to these proposals.6 During the period between Primitive Irish and Old Irish tremendous changes affected the language. These are to some extent visible in the inscriptions, which are, however, seriously deficient in morphological information (McManus 1991: 83ff.). The visual appearance of the Ogam was not affected by language change and thus gives a misleading impression of uniformity. The values of the letters were not immune to sound-change, as these were closely related to the names of the letters, based as they were on the acrophonic principle. Names are, needless to say, susceptible to sound-change. The basic inventory of four times five signs was at some stage augmented with an additional fifth group of letters. These letters were probably a response to manuscript literacy, their creators being ‘out of touch with the objectives of the original framers of Ogam [...]’ (McManus 1991: 2), and will be discussed in more detail at a later stage. The Ogam alphabet is the first result of linguistic analysis ever made in Ireland. It was accomplished at least by the fifth century. Although this analysis was evidently rather advanced, and presumably dependent on circulation of Latinate grammatical doctrine in the Roman provinces, it furnished the later Old Irish linguists only with their object of study. The stage of the Irish language to which the Ogam was tailored (Primitive Irish) was more or less unknown to scholars in the Old Irish period, and the origin of Ogam clearly a great deal more obscure to them than it is to us. This does not imply that medieval scribes did not spend ink on the matter – obscurity rarely does. They ventured to elucidate the origin of the vernacular alphabet through mythography (2.10.), fitting it into the larger pseudo-historical account of the Irish cultural sphere. Hence they allowed their ignorance of the actual act of invention to

6 The earliest keys to the alphabet are found in the Continental manuscripts Codex Bernensis 207 (fol. 257r) from the eighth/ninth centuries and Bibliotheca Apostolica Vaticana MS. Reg. Lat. 1308 (fol. 62v) (McManus 1991: 135). The earliest relevant manuscripts of Irish provenance are from the 14th century; in addition to the Auraicept, they contain the texts Dúil feda na forḟid and Dúil feda ind Ogaim.

16 become apparent, but they also enabled the re-analysis of their alphabet by the application of that kind of grammatical doctrine, admittedly rather basic, as had enabled its design. The earliest attempt at such analysis is carried out in the grammatical treatise commonly referred to as Auraicept na nÉces (ed. Calder 1917; ed. Ahlqvist 1982).

The runic alphabet or fuþark was in use for a period of about 1300 years. Its origin is no less enigmatic than that of the Ogam, but consensus (Spurkland 2001; Barnes 2012) suggests that the Latin alphabet was the most important antecedent, although some graphical features are easier to account for if other impulses are acknowledged (see discussion in Spurkland 2001: 17). The fuþark is unlike the Ogam in that not only its underlying principle, but also its signs might be traced to the Mediterranean systems. The letters of the runic alphabet were organized into three groups (in later manuscripts these are called ættir (sg. ætt) (Barnes 2012: 17). The fuþark underwent two main periods of great change: The older fuþark contained 24 graphemes which were reduced to 16 graphemes in its younger variety.7 Probably several mechanisms were at work in the period of change. These are listed and discussed by Spurkland (2001: 92-93): ‘Graphical circumstances’ (i.e. graphical simplification), sound-changes and, as a consequence of these, changes of the names of individual runes. The runic names were likewise based on the acrophonic principle, and a change in their anlaut (e.g. loss of j- and breaking) would consequently have an impact on their value. Sound-changes not affecting the initial sound would naturally be of no consequence in this respect. Umlaut had led to the development of new allophones which were phonemicized as a result of syncopy. The result was a language with an increased inventory of vowel-phonemes (5 › 9, excluding distinction of quantity and later mergers). The 16-runes’ fuþark did not enable unambigious representation of these phonemes. Towards the High Middle Ages, however, new graphemes were derived from the principal ones by punctuation and distinction of the short-twig varieties from their long-twig counterparts (Spurkland 2001: 163ff.).8 The result was a set of additional characters. This second period of change was concluded by 1200 and was triggered by the presence (at least from the end of the 11th century or beginning of the 12th century) of a manuscript culture which was borne by the Latin alphabet.

7 This should not necessarily be understood as a ‘reform’ (cf. Barnes 2012). 8 E.g. t /t/ is distinguished from d /d/ and k /k/ from g /g/ by punctuation; o /o/ is kept apart from from ø /ø/ and ö/Ö /ɔ/ by establishing a distinction short-twig/long-twig (Spurkland 2001: 164-65).

Spurkland (2001: 166, 213; see also Barnes 2012: 123ff.) argues for the existence in Norway during the High Middle Ages of a literary culture in which the two alphabets were employed in complementary distribution, as it were. Evidence does not suggest that the runes were taken into use in a manuscript context, with the exception of common abbreviations, together with a few citations, and of course, the Danish compilation AM 28 8vo (Codex Runicus; c. 1300), which contains law texts written entirely in runes, but is otherwise adherent to MS conventions.9 Its only parallel is catalogued as SKB A 120, which is therefore quite possibly a product of the same school or scriptorium, located (on linguistic and thematic grounds) in Scania County (today in Sweden) (Svanhildur Óskarsdóttir 2013: 104). In any case, The Third Grammatical Treatise represents a violation of any complementary distribution,10 though composed in Iceland and not in Norway, by staging the entrance of the runes into the domain of the Latin alphabet.

1.4. Institutiones grammaticae The Institutiones grammaticae (‘The Institutes of Grammar’; henceforth Instt. gramm.) was among the grammatical works rediscovered during the Carolingian Renaissance. The popularity of the work during and following the Carolingian Renaissance is confirmed by the great number of manuscript witnesses which have survived (Law 2003: 143-47; Irvine 1994: 306). Among these we find the famous Codex Sangallensis 904, which has been crucial in the description of Old Irish grammar, as represented by Rudolf Thurneysen’s Handbuch des Alt- Irischen (Heidelberg 1909; translated to English in 1946), due to its rich collection of glosses, and which testifies to the solemnity with which the text was received in Irish circles in the middle of the 9th century. It is significant that the Irish interest in Priscian (and Donatus) might have predated the Carolingian Renaissance by as much as two hundred years (Ahlqvist 1982). For the present investigation, however, it is of equal importance that this manuscript is

9 A facsimile edition of this ms. might be found at https://handrit.is/en/manuscript/view/da/AM08- 028. Consult also Svanhildur Óskarsdóttir 2013 (104-105) for a discussion on the manuscript – “one of the most remarkable products of medieval literary culture” (ibid.). It is most probably a revivalist production with no genetical link to epigraphic script culture: “hafa menn giskað á að í rúnanotkuninni felist ákveðinn vilji til að endurlífga gamla norræna letrið eða hefja það upp á stall til hins latneska“ (ibid.). “Men have guessed that behind the use of the runes there is to be found a decided wish to revive the Old Norse alphabet or to elevate it to the standing of Latin”. 10 It is a possibility that traditional runic literacy had all but disappeared in Iceland at the time of composition, judging from abscence of evidence (see further Wills 2016: 127).

18 a contemporary document which testifies to the work done by Irish scribes in the reception of late antique grammatical writing, but also to current influence on Irish literate culture and grammatical meta-language (see Ahlqvist 1982: 14-15). St. Gallen, the monastery to which this manuscript, possibly of Irish provenance, was eventually brought, was evidently a centre of grammatical learning on a high level. This is evidenced by the presence of such persona as Notker (c. 840-912) (Irvine 1994: 310), who had a high command of both Latin and his vernacular tongue, and who created the first systematic orthography for the latter.

Priscian Caesariensis (fl. early sixth-century AD), the author of Instt. gramm., received his education in Constantinople and was active in the early sixth century. The context in which he worked, together with the character of the text, clearly shows that it was intended to serve native speakers of Greek, with its abundance of Greek examples and reference to dialectal characteristics. Among the features of the Institutes which made it expressly suitable for speakers of other languages was its inclusion of a discussion of pronunciation. Priscian’s influence extended well beyond the Carolingian Renaissance. The Instt. gramm. had a profound effect on the nature of linguistic inquiry because of its inclusion of more advanced discussions on linguistics, and this effect made an impact also on grammatical thought during the Italian Renaissance in the fifteenth century. Between these two renaissances the work was in steady use and served to fuel grammatical discussion. Due to its monumental size, it was commonly disseminated by means of extracts. An important example of this is the Excerptiones de Prisciano, which was adapted and translated by Ælfric (ca. 992- 1002)(ed. Zupitza 1880; see further Law 2003: 145). His work in turn became influential in Iceland (Gade 2007). Ælfric takes a markedly apologetic position when he introduces his Old English translation (Law 2003: 194). This is articulated in both languages.

1.5. Auraicept na nÉces The oldest extant attempt at adapting Latinate grammatical description on a vernacular language in Western Europe is the Irish treatise Auraicept na nÉces (ʻThe Scholars’ Primer’; henceforth Auraicept), which treats topics of linguistic interest. The oldest stratum of this text is traditionally dated to the early Old Irish period (Ahlqvist 1982). When the title of the treatise, which has medieval authority, was taken into use we do not know (cf. Burnyeat 2007: 191f.). Caution should therefore be issued in using this title as confirmation of the intention and aspirations of the author(/s). On its own, the term auraicept was compounded from the Latin borrowing aiccept/aiccecht and the preposition air,11 and used to denote such concepts as ‘primer’ or ‘primary lesson’. Hence the Auraicept na nÉces is commonly perceived as a « manuel de formation des (futurs) savants » (Lambert 1987: 17). 12 The genitival attribute na nÉces (g.pl. of éces ‘scholar’, ‘poet’) is generally taken to refer to the ‘scholars in spe’, but is ambiguous as it might also refer to the scholars who issued the ‘primer’.13 This ambiguity is not present in the reference to the Auraicept in the Mittelirische Verslehren II (henceforth MVII) (ed. Thurneysen 1891: 1-182), which has the reading na nÉicsine (‘of the students’, construed with the diminutive suffix) rather than na nÉces.

11 The latter part of the compound (aiccept), is derived from the supine stem (accept-) of the Latin verb accipiō. More general use of the term (i.e. with the meaning ʻlesson’) is documented for instance in the following passage, taken from the commentary to Félire Óengusso Céli Dé (Stokes 1905: 202): Is ann bói Ciarán ic frital a aicepta soisceli, 7 is e ní airithe ro labair .i. Omnia quaecunque uultis ut faciunt uobis homines sic facite iliss. This is rendered as: “Ciarán was then preparing his Gospel lesson, and this especially is what he said: ‘Whatsoever ye wish men to do unto you, that do ye unto them’.” (ibid.: 203). 12 Marstrander translates Auraicept na nÉicsíne as ʻstudenternes elementærbok’ (notes to Thurneysen 1891: 32; consult bibliography). Thurneysen: ʻVor-Aufgabe der Studenten’ (Thurneysen 1891: 115). Note that both translate ... na nÉicsine ʻof the students’ (DIL: éicsíne) and not ... na nÉces. (See further Burnyeat 2007: 189ff.). 13 Such a construction is paralleled in the text usually referred to as Audacht Morainn: Incipit Auraicept Morainn nó Tecosca Morainn for Feradach Findḟechtnach (slightly normalized from Thurneysen 1917: 80). Here the preposition for precedes the receiver of the teachings, while authorship is expressed by a genitive attribute. It should be noted, though, that the usage in MVII does seem to contradict this explanation, with its distinction between airacept and foacepta (Thurneysen 1891: 32, 34; see also Burnyeat 2007: 191).

Medieval usage seems to favour more specific reference to the core-text, but in a manuscript culture this core would invariably be surrounded by gloss and commentary. The make-up of this immanent kernel is not, however, immediately discernible. This is partly caused by the fact that the Auraicept na nÉces is traditionally divided into four different books which are all ascribed to pseudo-historical authors (see 2.10.). Due to its brevity, the text reconstructed by Ahlqvist could hardly be subdivided into four feasible books or chapters 14 - as stressed by the editor (1982: 19). Thus the division into four books must have been carried out only after a substantial amount of commentary was composed (see Bauer 2013: 11). MVII suggests that Auraicept na nÉcsíne was thought to consist of three parts: Prologue (brol(l)ach), main part and paradigms or inflections (réimenna) (Thurneysen 1891: 32, 115; Ahlqvist 1982: 17), and does not mention the fourfold division which is highlighted in the commentary to the Auraicept itself. It would seem that the commentators sought to augment the importance of their object by various means. This argumentation will be developed further. The Auraicept na nÉces was first edited by George Calder (1917). His edition is based on the recensions in RIA MS 23P12/536 (Book of Ballymote; henceforth BB), NLS MS Advocates’ 72.1.1 (E), TCD MS H 2.16 (Yellow Book of Lecan; YBL) and British Library MS Egerton 88 (Eg.). Together with the Old Irish text he has edited accompanying Middle Irish commentary and glosses. Calder recognized two families of text and commentary which go back to a common source, from which they presumably deviated during the eleventh century (Thurneysen 1928: 285; Ahlqvist 1982: 32). The text that is shared by both recensions is, as would be expected, often better transmitted in one of the families (Thurneysen 1928: 280). The first redaction is furnished with a somewhat problematic translation, which at times surrenders to Irish idiom to the degree that the reader unfamiliar with ‘standard Old Irish textbook prose’ (Charles-Edwards 1980: 147) in its Middle-Irish guise may be charmed into thinking that it is an acutely faithful rendition of the original. Arguably more serviceable to the researcher is Calder’s endeavour to portray the actual manuscript state of the text with its occasional imperfections – but his text is not a diplomatic transcript of any single witness. Information on his editing procedure is not included in the introduction to the edition. Calder did not always identify the correct source for the quotations in the Auraicept na nÉces, but a valuable contribution in this regard is Poppe’s paper “Latin Quotations in Auraicept na nÉces: Microtexts and their Transmission” (2002).

14 Note, for instance, that no statement of Amairgen’s is treated as canonical in BB. In YBL several of them are given this privilege.

The make-up of the source material is rather complex. The nature of the manuscript transmission of the Auraicept na nÉces and its commentary is to a very limited extent treated in the foreword by Calder (1917: xiii, xxiii). Calder’s edition is far from satisfactory on several points. Firstly, it may be noted that his edition “being the texts of the Ogham tracts from the Book of Ballymote and the Yellow Book of Lecan, and the text of the Trefhocul from the Book of Leinster, edited from eight manuscripts, with introduction, translation of the Ballymote text, notes and indices” (Calder 1917, title blade) does in fact contain two editions of the core text and its commentary (i.e. one of each family of manuscripts). Calder divided the manuscripts with which he worked into two families according to whether they contained a shorter version of the commentary (as do his B, L, M) or a longer one (as in Y, Eg.). Calder’s two editions of the text with commentary are edited from eight manuscripts altogether. In addition to this he includes treḟocal from LL (1917: ll. 5057-5414). Calder failed to provide accurate information on his editing procedures and principles of translation.15 The consequence of this and of his at times rather disorderly presentation of the material is that the use of his edition (not to mention the facing translation) as the basis of research requires a good deal of caution.16 The advantage of this is that a study of the text is almost forced to deal with, or at least take into account, the heterogeneity of the actual manuscript witnesses. Although her article is limited to two such witnesses, Deborah Hayden (2011) shows how taking into consideration the different organisation of the manuscripts’ content may assist our understanding of the text(s). The organisation and extent of the accreted commentary varies significantly in the manuscript witnesses (Hayden 2014: 24). With the facsimiles which are made available by the ISOS-project17 it is of course more convenient than before to do comparative examinations of the different witnesses to the text. Notwithstanding this ease of access, and the fact that the project provides updated catalogues of the items contained in the published manuscripts, thorough and reliable codicological reexamination would demand access to the manuscript themselves.

15 The somewhat forthright air is and is ben-si intan fognaithir dia sliasait (Calder 1917: l. 609) is translated with “for it is then she is a woman, quum femori ejus serviatur”. The use of Latin here for euphemistic purposes is rather confusing, as Calder does not in general translate Latin quotations (but compare for instance l. 214). He tidies them up and occasionally expands them (ll. 457-59). 16 For instance a sentence following i clar na croiche in BB is lacking from Calder’s text (1917: l. 165). This reads: Cid siu natised atuaid asin sgitia (BB: 368 a). Calder does not mark this omission. 17 http://www.isos.dias.ie

The so-called canonical text was later edited afresh by Anders Ahlqvist (1982). His edition is based on several manuscripts which were unknown to Calder. His catalogue-based research into the Auraicept-corpus yielded eleven fragments of the text in addition to eleven full copies, which were all divided into three families according to their content and their treatment of the canonical part of the text (Ahlqvist 1982: 22-27). All manuscripts are of Irish provenance (Poppe 2002: 312),18 with the earliest dating to the 14th century and the later paper copies to the 17th/18th centuries.19 As is the case with other works from the Old Irish period, many of the details surrounding its transmission will therefore remain unknown due to loss of evidence. The number of manuscript witnesses, the amount of gloss and commentary in them and the length of the period in which the Auraicept was read and was able to exert influence on other texts, is clear evidence that it was considered important. In scholarly discourse, Auraicept na nÉces is used to denote the text in its fully commented form, as it is represented by each of the two recensions in Calder’s edition (1917).20 This is complemented by more specific reference to the canonical text which became the locus for commentatory and glossatory activity, rather as it is delimited and edited by Anders Ahlqvist (1982). It is important to note that the perception of the text (including features as for instance layout) in the Middle Irish and Early Modern Irish periods has been of great importance also when reconstructing the Old Irish text of the archetype. Medieval “editorial technique” has, in a way, aided the modern researcher in moving past a point beyond which textual criticism may not reach.21 The accuracy of any reconstruction which aims to represent a text which is older than this point is dependent on the fallibility of medieval textual preservation and presentation. That differentiation of script is in itself no certain criterion when establishing a text is proven by Meroney (1945: 18-19; see also McLaughlin 2009: 2). The fact that only Ahlqvist’s group b preserves more or less systematic distinction between canonical text and commentary is worthy of note - how was the canonical text delimited in the common source? No stemmatic relationship of the different witnesses has yet been presented.

18 Against the notion that Auraicept na nÉces is an odd representative for Irish medieval literature, it might be noted that the rather more well-known heroic epos Táin Bó Cúailnge survives in eight mss. 19 The oldest extant fragment of material relating to the Auraicept proper is the so-called Ó Cianáin Miscellany (NLI G2; ca. 1345). 20 These recensions share several common mistakes which must by necessity have occurred before the textual transmission was split (Thurneysen 1928: 281; Meroney 1945: 18). 21 It should be mentioned that the tradition is contaminated (Ahlqvist 1982: 29), showing interplay of elements from different recensions.

The Auraicept in its commented form will be the main object of study for the present investigation. The conclusion must therefore be that Calder’s edition is the natural point of departure. The original Auraicept, if such a text existed, is irretrievably lost. Although Ahlqvist’s edition provides us with a probable archetype, it is not an authentic document.22 In spite of its many shortcomings, the first edition will have to serve. Ahlqvist’s edition, being a considerably more conscientious production, will be consulted when reference is made to the core-text. Below is an overview which might help the reader to find the relevant parts of the edition (given in lines). It is based on a summary which Thurneysen made in ZCP 17 (1928: 281ff.) of the contents of Calder’s edition. See also Calder’s harmony between the two families (1917: liv-lv).23

1. (I. 63-734; II. 2260 resp. 2616-3492) Book of Cenn Fáelad mac Ailella 2. (I. 735-1027; II. 3492-3984) Book of Ferchertne. 3. (I. 1028-1101; II. 3984-4101) Book of Amairgen Glúngel 4. (I. 1102-1636; II. 4136-4725) Book of Fénius, Íar and Goídel 5. (I. 1637-1758; II. 4726-4960) Coic fillte fichet i reim 6. (I. 1759-1879; II. 4961-5055) Do bhunadhaibh na remend 7. (I. 1880-1892) No title. 8. (I. 1893-1926) Do ernailibh in imchomairc 9. (I. 1927-2255) Treḟocul (1) 10. (II. 5056-5415) Treḟocul (LL) 11. (5417-5463) De duilib feda na forḟid (BB, LL) 12. (5465-6165) Ogam (BB)

22 For a discussion on the mechanisms involved in the preservation of canonical text surrounded by commentary, the reader is referred to Fergus Kelly’s article on the transmission of law-texts (2002) as well as his A Guide to Early Irish Law (1988: 225ff.). 23 I: first family of mss. II: second family of mss.

1.6. The First Grammatical Treatise The First Grammatical Treatise (henceforth FGT; ed. Hreinn Benediktsson 1972) provides us with a discourse on the adaptation of the Latin alphabet to Icelandic. The FGT is securely dated to the period 1125-75 (Hreinn Benediktsson 1972: 31), but is only preserved in the mid- 14th century Codex Wormianus. It is a unique document not only in an Icelandic context, but in the Medieval West. The FGT, being the second preserved attempt at the study of a non- sacred language in Western Europe, is occasionally mentioned in the same breath as the Auraicept. Apart from superficially dealing with cognate subject-matter, the texts have very little in common. The task the FG set himself was to present an improved alphabet for the Icelandic language. This necessitated reflection on the sounds of the language and of the ability of the letters to represent that which in his words might ʻalter the discourse’ (skipta máli). All his propositions are assessed and, in principle, supported by the use of a commutation test. The degree to which his propositions are backed up by the use of minimal pairs is quite remarkable for his time, as these were granted no central role in traditional grammatical literature (Hreinn Benediktsson 1972: 76-78). The FGT stems from a debate which must have taken place among men of letters, but which has not been elsewhere recorded. In other words, the FGT represents a kind of discourse which has left no trace in Irish sources: How should the vernacular be reduced to letters? Moreover, it shows that such discourse must have taken place over some time and offers a view of the kind of arguments that might have been used in the evaluation of different proposals. The FG did not venture to apply Latinate grammatical precepts on the vernacular or its alphabet - on the contrary his evaluation of the vernacular alphabet is less than enthusiastic: Eigi er þat rúnanna kostr þó at þú lesir vel eða ráðir vel at líkindum þar sem rúnar vísa óskýrt (normalized from Hreinn Benediktsson 1972: 214). 24 The most recent editor of the text, Hreinn Benediktsson, was of the opinion that rún was used as a ‘mere stylistic variant’ of stafr (1972: 42) and dismissed the reading of Björn M. Ólsen (1883: 103-104), who took rún to simply denote ‘runic letter’. This reading is preferable

24 “It is not the virtue of the runes if you are able to read well or give a likely interpretation where the runes are unclear.” Hreinn Benediktsson: “It is not the virtue of the letters if you can read well or make a good guess (in cases) where the letters are unclear,” (1972: 215).

– not least if style is invoked as a parameter.25 The polemic against the vernacular alphabet is lost in translation and replaced with polemic against unclear usage. At any rate, the perception of the vernacular alphabet at the time of the FGT, when Icelandic literacy was still in its infancy26 differed vastly from the one which comes to the fore in the ThGT about a century later.

25 The term rún does in fact only occur four times in the FGT; twice outside the quoted passage, both times as part of the ʻminimal pairʼ ru̇nar:rúnar (Ru̇nar heita gelltir enn rúnar málstafir “Male pigs are called r nar (‘boars’), but letters (are called) rúnar (‘runes’),” (Hreinn Benediksson 1972: 222/223). Hreinn offers no further support for his choice of translation, which might therefore be dismissed (see Males 2016: 266n13). Snorri uses the word occasionally, but rather to denote ‘secret’ or ‘enigma’ than ‘letter’ or ‘rune’ (ONP: rún) (Males 2015: 177n10.). 26 On the authority of the FGT it already encompassed laws, genealogy, exegesis and historiography (Hreinn Benediktsson 1972: 208/209).

1.7. The Third Grammatical Treatise The Third Grammatical Treatise (henceforth ThGT) is contained in four manuscripts: AM 748 Ia-b 4° (henceforth A; early 14th century), AM 242 fol. (W; mid-14th century), a fragment (w) now part of AM 757b 4°, and AM 757a 4° (B; c. 1390-1410) (B. M. Ólsen 1884: xlviii ff.).27 Sharing all the errors of W, the text in w appears to be a transcript of the former and is accordingly without text-critical value, but does nevertheless testify to the popularity of the text. The same is true for a number of later paper copies of the treatise. Although its manuscript record might seem modest compared to that of the Auraicept, with its eleven full and eleven fragmented copies (Ahlqvist 1982: 22ff.), the ThGT is the best represented of the medieval Icelandic grammatical treatises, with the important exception of Snorri’s Edda and especially the influential Skáldskaparmál (ed. Faulkes 1998), which, following Guðrún Nordal (2001), are counted as part of the grammatical corpus. In W, the so-called Codex Wormianus (on which see Svanhildur Óskarsdóttir 2013: 39ff.), the ThGT is found together with three other grammatical treatises and a prologue which introduces them as a thematic group. Unsurprisingly, none of these texts have survived as autographs (Johansson 1997: 57; Males 2013: 41n2), with the exception of the prologue, which is tailored to fit the particular ‘grammatical digest’ (Raschellà 2005: 359) contained within W (on its redactor, see Males 2013). W significantly includes a copy of the aforementioned Edda of Snorri Sturluson, and its wide range of works pertaining to the vernacular language and literature makes W the foremost representative of the Icelandic grammatical tradition of the Middle Ages. Apart from the texts contained within W, only minor texts are extant, as the fragment AM 921 III 4° (c. 1400) (ONPRegistre 1989: 464), which contains examples of Latin verbal conjugation with Icelandic translation, and AM 748 Ib 4to (c. 1300-25) (loc. cit.), which contains a diminutive portion of another grammatical treatise. Two of the grammatical texts containted in W are not extant elsewhere. The treatises are all anonymous and carry no name - hence their order of appearance in this manuscript provides a convenient way of reference to them.28 The other main witness to the ThGT, and

27 Consult also ONP Registre on http://onp.ku.dk/adgang_til_ordliste_etc/registre/ (last visited November 6th 2016) for updated catalogue information. The dates provided there correspond to those given by B. M. Ólsen (1884). 28 In this treatise they are referred to as First Grammatical Treatise (FGT), Second Grammatical Treatise (SGT), Third Grammatical Treatise (ThGT) and Fourth Grammatical Treatise (FoGT). Their

27 the one which is generally preferred by its editors, is A. B. M. Ólsen posits a single manuscript witness between the archetype and A and three witnesses between the archetype and W (B. M. Ólsen 1884: lxii). B derives from the same redaction of the text as that which is found in AW (op.cit.: lviii). Tarrin Wills (2001: 52ff.) has challenged B. M. Ólsen’s assessment of the stemmatic relationship between the manuscripts, and has made a fresh contribution in which he harmonizes the stemma of the ThGT and that which Finnur Jónsson made of those witnesses to Skáldskaparmál which are contained in the same manuscripts. The first part of the Third Grammatical Treatise (ThGT, c. 1250) is commonly referred to as Málfræðinnar grundvöllr (‘Foundation of Grammar’; henceforth Mg). As noted above, there is no medieval authority for this title, which first appeared in the editio princeps by Rasmus Rask (1818: 297) and has been used by later editors (Sveinbjörn Egilsson 1849; B. M. Ólsen 1884; Wills 2001).29 Mg is, though the bulk of its content is derivative of Priscian (or directly translated from him), the most original contribution of the ThGT (Finnur Jónsson 1927: 14f.; Raschellà 1994: 679). The very first passage is a deviation from traditional structure (represented by e.g. Donatus’ Ars maior) and have more in common with tracts on logic and dialectics; such as the 12th century Summulae logicales by Petrus Hispanus.30 This passage begins with the very general statement: Allt er hljóð, þat er um kvikvendis eyru má skilja (B. M. Ólsen 1884: 33). “Sound is everything which the ears of living creatures may comprehend.” The discussion of different sounds is gradually narrowed down, until it reaches the more humble level of abstraction where Priscian begins his discourse: Philosophi definiunt vocem esse aerem tenuissimum ictum vel suum sensibile aurum, id est quod proprie auribus accidit (GL vol. II: 5). This is rendered in the ThGT as: En Priscianus kallar rǫdd vera hit grannligsta loptsins hǫgg ok eiginliga eyrum skiljanligt (B. M. Ólsen 1884: 35). “But Priscian states that voice is the most subtle strike of air and intelligible to one’s own ears.” From this on Mg follows the lay-out of Priscian quite closely. The second part of the ThGT was named Fígúrur í ræðunni (ʻFigures in speech’) by Rasmus Rask (1818: 308). Later Sveinbjörn Egilsson called it Málskrúðsfræði (1849; authors are referred to by abbreviations on a par with ‘FG’ (leg. First Grammarian), as done by Hreinn Benediktsson (1972). 29 Finnur Jónsson (1927) and Thomas Krömmelbein (1998) both used Björn M. Ólsenʼs edition as their point of departure. The latter edition contains a convenient, though not quite satisfactory, German translation (see Jón Axel Harðarson 2003). 30 For a demonstration of the influence of this text on the ThGT the reader is referred to the commentary in Wills’ edition (2001).

28 henceforth M). It deals with rhetorical figures in skaldic poetry and includes a wealth of verses of anonymous and identified origin.31 M is, in terms of traditional Latin grammar as represented by e.g. Donatus’ Ars maior, the logical extension of Mg, as it follows the progression, beginning in Mg with the sound of speech and the syllable, to the larger units of textual composition. It is reasonable to believe that Óláfr’s treatise has been transmitted without significant losses or additions (Finnur Jónsson 1927: 16), although it should be said that certain systematic abridgements of the text are witnessed in B (Wills 2016: 119), some of which may have been made out of editorial expediency (ibid.). Although the editor of W has generally not refrained from editorial interference, the text of the ThGT is relatively untampered with (op.cit.: 43 §[4.2]). The ThGT has been edited and published several times. The first edition by Rasmus Rask (1818; 297ff.), solely based on W, has already been mentioned.32 Sveinbjörn Egilsson published an edition of the text from the manuscripts A, W and B, as part of his edition of SnE (1848-49; 173-200). Sveinbjörn chose A as the basis for his edition. This has been the standard since. The ThGT was published as part of the Arnamagnean edition of SnE (1848- 87). The standard edition of both the ThGT and of the FoGT was published by Björn M. Ólsen in 1884.33 Finnur Jónsson edited the ThGT in 1927. Tarrin Wills has edited Mg afresh as part of his doctoral thesis (2001).34 His edition contains transcripts from all four manuscript witnesses, together with an edited text accompanied by a text-critical apparatus and a translation. Wills has identified relevant passages in other texts that might have influenced the author of Mg. The authorship of Óláfr Þórðarson has medieval authority, as it is established in a passage in A (fol. 14v, ll. 6ff.): Hér er lykt þeim hlut bókarinnar er Óláfr Þórðarson hefir

31 A list of these verses are to be found at The Skaldic Project (ed. Clunies Ross et. al.): http://skaldic.abdn.ac.uk/db.php?id=32&if=default&table=text (last visited October 17th 2016). 32 A new diplomatic edition of W is available at the MENOTA project; http://www.menota.org 33 This edition of the FoGT has recently been replaced by Clunies Ross and Wellendorf (2014). Wills’ edition is a very valuable complement to B. M. Ólsen 1884. 34 Page numbers refer to the printed, but unpublished edition of the thesis, while the paragraph numbers might also aid in locating the relevant section in the digital edition.

29 samansett [...] (B. M. Ólsen 1884: 119).35 This statement, written in red ink (compare note in B. M. Ólsen 1884: 119 n.), is generally accepted (Clunies Ross 2005: 109, 186, etc.) and is furthermore confirmed by the FoGT, which contains one quotation which is attributed to Óláfr, though without the patronymic, together with explicit references to the terminology of the ThGT (Clunies and Wellendorf 2014: 12, 18-22). Óláfr hvítaskáld Þórðarson (d. 1259) was the son of Þórðr Sturluson, brother of the politician and author Snorri Sturluson. He was a member of the influential family of the Sturlungs, and ran a school at Stafaholt in Borgarfjörður. This school is the likely context for the production of the ThGT (B. M. Ólsen 1884: xxxvi; Finnur Jónsson 1927: 7). Óláfr was clearly well versed in Latin (‘en velstuderet klærk’) (Finnur Jónsson 1927: 3), judging by his output, which is limited to the ThGT and skaldic poetry (cf. Males 2015). Óláfr’s death, before the age of fifty, poses a certain terminus ante quem for the composition of the ThGT, which is therefore likely a production of the 1240s or 1250s (see a summary of biographical information based on the sagas in B. M. Ólsen 1884: xxxii-xxxvii; Finnur Jónsson 1927). Before the composition of the ThGT, Óláfr had been attending the court of King Valdemar the Conqueror (Valdemar Sejr). There is text-internal and external evidence (saga literature, e.g. Knytlinga saga) of this event. The Codex Runicus (see 1.3.), though postdating Óláfr’s death, might have been the result of similar influence, though this is neither confirmed nor precluded by the available evidence (Wills 2016: 125). The ThGT does not only offer the most relevant comparison with the Auraicept, but it is the only Old Norse/Icelandic (indeed the only Germanic) production which fully deserves to be referred to as ‘grammatical’ (as stated in Raschellà 1982: 4) – the only one to treat the full specter from the letter to rhetorics.

35 “Here that part of the book which Óláfr Þórðarson put together is ended.” As an aside it might be mentioned that the attribution of the Auraicept of Cenn Fáelad is not generally accepted, although this too might boast of medieval authority.

1.8. Considerations on Method The following discussion will direct attention to some of the issues that are germane to the methodological approach as laid out in the preamble. First of all, this study does not presuppose any direct connection between the Icelandic and the Irish grammatical traditions. A certain influence might be envisaged (see Bauer 2013), although this is inherently difficult to prove when the relevant confounders are ruled out. Insular influence on the Icelandic grammatical treatises would rather have been provided by the Hiberno-Latin grammatical commentaries (see Micillo 1999: 215). As a consequence, boundary between languages is avoided as an issue. The only relevant boundary is that between the vernacular language and Latin, and this, I think, might better be described as an interface in the context of bilingual textual communities. The Latin grammatical tradition, as a tertium comparationis, is the vantage point from which the vernacular traditions might be contrasted. The zenith of grammatical learning was the Instt. gramm. which provided many centuries with a huge collection of data and the means necessary to analyse them (Law 2003: 86). Chronology presents an issue in two ways. The first of these is the fact that the texts themselves date from different periods, a hitch, while the second issue pertains to the timespan of the Auraicept-tradition itself, and, though not unique to the present argument, is indeed a serious impediment, although it is frequently bypassed by scholars of Irish literature. Vivien Law proposes a fourfold division of the stages in the history of linguistic thought in Europe during the Middle Ages (2003: 112-15):

1) The early Middle Ages (500-800); defined by the endeavour to make descriptive grammars of Latin available and understandable to non-native speakers in Western Europe. 2) The central Middle Ages (800-1100); the period from the Carolingian Renaissance to the twelfth-century Renaissance; the rediscovery of Priscian’s Instt. gramm. together with Aristotle’s Categories and De interpretatione. 3) The later Middle Ages (1100-1350); development of Scholasticism and influence from Aristotelian writing on linguistic thought. 4) The end of the Middle Ages (1350-1500); heightened awareness of the vernaculars and attempts to present Latin grammar through the medium of the vernaculars. Some experiments with vernacular grammars.

It is difficult (and probably futile) to attempt to fit the relevant texts into the stages proposed by Law. The kernel of the Auraicept na nÉces would belong to the first stage and The Third Grammatical Treatise to the third.36 Instead of undermining her system this emphasises the singular character of these texts within the long term development of Western European linguistic thought until the end of the Middle Ages. It should be said here that the Instt. gramm. may have been unknown at the time of the initial composition of the Auraicept na nÉces. This is not quite the stumbling block it might seem at first glance, for two main reasons: The date of composition of the Auraicept is uncertain (see discussion in the following paragraph). Therefore its role in relation to the Carolingian Renaissance is not clear. 37 In this period Priscian was rediscovered, as rehearsed above. In addition to this the doctrines of Priscian provided posterity with an imposing collection of grammatical information which comprised the works of his predecessors. The definitions of Donatus’ Ars maior, for instance, do not explain the argument laid out by the author of the Auraicept, but Priscian provided definitions which do. Both the Auraicept and the ThGT mention Priscian in several instances and they make conscious use of him, if not always conscientious. The principal issue posed by chronology has at this point been anticipated. It is caused by the complexity of the Irish material, concerning its textual history (see 1.5.), and the difficulties that arise in the attempt at delimiting it. Choosing a particular chronological layer as the basis of study (as in Acken 2013)38 is at this stage a questionable approach, due to the uncertainty regarding the make-up of the text at any given time. There are two solutions to this problem: One is to let chronology prevail and accept doubt as to the actual contents of the

36 The actual manuscript witnesses would, for both texts, belong to the third and fourth stage. It is interesting to note that these manuscripts confirm the interest in vernacular languages towards the end of the Middle Ages. 37 It has traditionally been dated to the 7th century (1.5.) and thus to the period preceding the Carolingian Renaissance. In this connection it should be said that the latter would have been a probable period as far as cultural history is concerned. 38 The validity of a statement as the following, intriguing though it is, must therefore be questioned: “What the Auraicept received from Latin tradition it absorbed and returned with an enormous profit; its systems of interpretation finding their way into the Latin commentaries of the Carolingian Renaissance.” (Acken 2008: 159-60). Poppe’s view, namely that the Auraicept reflects a tradition of Hiberno-Latin learning (2002: 299), is more considerate of the limited evidence at our disposal and of the chronological difficulties pertaining to it.

32 text, while the other is to give priority to the factual make-up of the sources and accept uncertainty with regard to which period they represent with any degree of trustworthiness. Accepting such a principle of uncertainty is required for methodological stringency - not generally, of course - but due to dearth of evidence.39 The above discussion on chronology does not apply to The Third Grammatical Treatise, whose manuscript state (see 1.7.) is more straightforward and presents issues that rather belong to textual criticism, hence evading the scope of the present treatise. Some issues owing to the choice of method must be mentioned. One is that the inclusion of two texts in so limited space makes it impossible to include comprehensive discussion on all points of relevance. An obvious solution to this problem is expansion of scope, which remains a possibility for the future. Some discussions by necessity are more relevant to one of the texts than to the other. I have sought to acknowledge this and to avoid a narrative without secure basis in the sources. These two shortcomings are interrelated. Equal weight is difficult to achieve. Priscian is used to introduce the thematical units, while the Auraicept generally preceds the ThGT in the subsequent discussion. Only where the material does not fit easily into this structure is it abandoned - I have chosen to pick up threads even when no corresponding discussion is relevant to the other text. More often this is done in favour of the Auraicept, a text on which I think much remains to be said. This thesis was conceived of as a way to communicate some of those things.

39 A strictly synchronic study might have been possible within a different theoretical framework, but is not, I think, of particular relevance to a comparative undertaking such as this. Although this thesis will deal with texts from different periods, it is here assumed that they represent more or less the same ‘relative period’ in their respective traditions. They answer a need or several needs which arose at a certain juncture in a period of transition and adaptation. The actual manuscript witnesses postdate this moment.

1.9. Theoretical Stance and Aim of Thesis This thesis will aim at an assessment of the Auraicept na nÉces and The Third Grammatical Treatise from a comparative point of view. The texts will be compared not with the intention of proving influence of one tradition on the other (cf. Bauer 2013). Instead the virtue of such a comparison is that it reveals that nothing is done simply by default, and that reception under comparable circumstances may bring about widely differing results. The art of definition has already been introduced (see 1.1.): It should indicate what the thing to be defined has in common with other things, and what are its unique characteristics. When each of the tracts are seen as a part of a larger network of texts within a single tradition, that is, of texts which have either influenced them or been influenced by them, it is only natural that some of their hallmarks appear as self-explanatory in default of other ones. A comparative approach is not strictly needed in order to explain the facts themselves – but the overview which it is able to provide should prove helpful in answering questions introduced with ‘why’ rather than ‘how’. For such answers to come about, comparison with the Latin grammatical tradition and contextualisation of each text within its respective tradition does not suffice. The reception of each of the texts in scholarly discourse has arguably matured to the point where it becomes natural to assess them together. It has not, however, advanced to the point where such a task is simply one of gathering the available information at one place, although this would in itself be a useful undertaking. Reference to Priscian’s Instt. gramm., a seminal dispenser of grammatical doctrine, though it was often heavily condensed in order to aid digestion, is meant to further the investigation by narrowing its scope and at the same time to provide a common point of departure. The direct influence of Priscian on the two texts is, however, well demonstrated, and the reader is referred to Wills’ edition of the ThGT (2001) and to Poppe’s 2002 article for an overview of parallel passages and quotations respectively. More complex (or indirect) influence is implied on the Irish text by the fact that our knowledge of synchronically documented Old Irish usage is heavily dependent on commented copies of Priscian kept in libraries on the Continent (Thes. Pal. vol. II.). The Icelandic text, on the other hand, is influenced also by the Anglo-Saxon excerpts from Priscian, which must have reached Iceland in some form before its production (Gade 2007; cf. Kristján Árnason 2016: 207), although the details are far from limpid at this stage. We might conceive of each text as an important stepping stone towards full-blown literacy in the vernacular. It must by necessity have been written some time after the

34 vernacular was taken into service as a written language. Yet literacy, in a wider and Latinate sense, would not be fully achieved without theoretical discourse on language and writing. The texts which were supposed to serve as introduction to Latin literacy presupposed a reasonable grasp of written Latin. It is therefore of utmost importance to note the fact that they were serving to introduce something else than the mere ability to read and write, viz. the ability to engage with the language and its literature in a theoretical manner. It is equally important to note that the ability to read and write does not in itself bring about such skills. The grammatical tradition which emerged in the ancient library of Alexandria was first and foremost concerned with the formation and maintenance of a literary canon (Haugen 1990: 12 ff.; Irvine 1994: 39 ff.). Its definition was perhaps first put to writing by Dionysius Thrax (c.

170-90 BC) (cf. Law 2003: 55-56; cf. Luhtala 2005: x, 2) in his Τέχνη γραμματική (Tekhnē grammatikē): γραμματική ἐστιν ἐμπειρία τῶν παρὰ ποιηταῖς τε καὶ συγγραφεῦσιν ὡς ἐπὶ τὸ πολὺ λεγομένων (Uhlig 1883: 5).40 The introductory statement gives a good idea of the role the work was intended to play: To aid speakers of Hellenistic Koiné in the reception of the older poets, whose texts were preserved in an older form of the language. It would seem that the ideology which is expressed by the introduction to Τέχνη γραμματική remained largely intact (though not necessarily expressed) and accompanied the translation or adaptation of late antique grammars into the vernacular languages.41 For a broad survey of grammatica as a cultural project and ideological tool, the reader is directed towards Irvine’s 1994 study. The idea that ‘grammar’ and ‘poetic technique’ or ‘metrics’ are to be kept strictly apart from each other in a medieval context is erroneous, and any attempt at combining them should

40 “La grammaire est la connaissance empirique de ce qui se dit couramment chez les poètes et les prosateurs.” (translation by Jean Lallot cited in Seppänen 2014: 55). Discussion on an alternative version of this definition is provided by Seppänen (2014: 53-54). Consider also the following scholia to Dionysius Thrax: γινώσκειν δὲ χρὴ ὅτι τὸ παλαιὸν ἡ γραμματικὴ κριτικὴ ἐκαλεῖτο παρὰ τὸ κρίνειν τὰ ποιήματα· Ἀντίδωρος δὲ τις γραμματικὸς αὐτὴν ὠνόμασε παρὰ τὴν γνώσιν τῶν γραμματῶν. “It must be made known that in the old days grammar was called ‘criticism’ from ‘criticizing poetry’; a grammarian called Antidorus named it ‘grammar’, as the knowledge of literature,” (Seppänen 2014: 30).

41 In a similar fashion Τέχνη γραμματική was translated into Armenian and commented on by the scholar David (7th cent.) (Law 2003: 191). The general approach is reminiscent of the one we see in the Icelandic ThGT/FoGT (ibid.). The object language is described within the confines given by the framework of the translated text.

35 not automatically be assumed to be innovatory on the part of the author.42 The combination was there all along. An examination of the vernacular treatises in their entirety must take into account their relationship to poetry and poetic composition. This is indeed done in several studies (Acken 2008; Burnyeat 2008; Clunies Ross 2005; Hayden 2011; see also Tranter 1997), but it bears mention here, that, being the most authoritative form of literature, poetry is never far away; even the vices of common speech might well be illustrated by skaldic poetry.43 In the Icelandic grammatical treatises the poetic material is certainly given a very central role, and, according to Tranter (2000: 143), an importance which exceeds the need prompted by the framework of the Latin ars poetica. It is a bit intemperate, for instance, when definitions translated or adapted from Latin are exemplified with whole skaldic stanzas or even groups of stanzas, as opposed to a single line from Vergil. Such strategies reveal the skaldic sentiments of the author. The last sentence in Acken’s study of the Auraicept (2008) reads: “The Auraicept na nÉices, far from presenting an attempt to produce a vernacular grammar, provides instead exactly what its title claims: the fundamental principles of poetry,” (Acken 2008:161). Poetry is quite correctly an important concern of the commented form of the Auraicept, although the opposition produced by Acken is anachronistic and fails to set aside the modern notion of ‘grammar’. An Irish-speaking audience would know how to conjugate their verbs, whilst a non-Irish-speaking audience would be unable to understand a single word of the text. There is no evidence, to my knowledge, that anyone ever went from the study of the Auraicept to the

42 The following statement by Raschellà (1982: 5), on the association of the Icelandic grammatical tracts with the Edda of Snorri Sturluson is interesting in this respect: “At any rate, the idea of associating grammatical writings with works dealing with poetic technique and metrics no doubt derived from an analogous Latin practice, with the fundamental difference that, while metrics was subordinated to grammar in the classical and medieval Latin tradition, the Icelandic authors had inverted the relationship.” 43 The paradigms of the Auraicept (Calder 1917: 118 ff.; Ahlqvist 1982: 52-53) do for instance look like a rather left-handed application of the Latin case-system on Irish. If we see them as a list of variations which a word might undergo in poetic circumstances they immediately make more sense (see further Acken 2008: 36-40). Such a view finds support in the fact that some of the versions of the paradigms include forms arrived at by the use of poetic devices (see also Hayden 2011: 10).

36 production of Ogamic prose; to evaluate the writings of the grammarians solely by conjecture regarding their ability to teach basic reading and writing skills would be malapropos. 44 From the above considerations it will be apparent that providing the fundamental principles of poetry, together with a discussion on the origins of language and its authorities, is exactly how to produce a vernacular ‘grammar’, following the example of late antique Latin grammarians and commentators. Thus the Auraicept does in fact constitute such an attempt, and, importantly, it was an attempt which was valid not only in the time it was produced, but also subsequently, as evidenced by the manuscript record. We must surmise that its modus operandi made perfect sense in its own time – even though it stands in rather stark contrast to modern linguistic theories. Its mode of inquiry is derivative of Biblical exegesis, the methods of which were applied to grammatical commentary and introduced to Carolingian schools on the Continent by the scotti peregrini (Law 1994: 104). In other words, features as ‘crazy etymologies’ (Bergin 1938: 206-07) and ‘pretentious nonsense’ (Adams 1970: 157) that figure in the commentary to the Auraicept were very much in vogue. At any rate, they represented a valid form of inquiry and were recommended by the prestigious study of sacred writings.45 Exegesis was the zenith of the culture in early Christian Ireland (Charles-Edwards 1998: 75). The fact that the Irish and Norse spheres were deficient in written literature before the advent of the Church is not to be blamed on the lack of writing – they both had their own native alphabets centuries before the emergence of vernacular literature. What was absent was the Latinate notion of ‘literature’ itself. Such a literature would by definition have to be written as opposed to oral. The otherwise interesting interface between written and oral thus becomes negligible in the context of grammatica, in which ‘voice’ is that which may be written and

44 Consider for instance: “Whatever about its dating, however, there is nothing in the work that would stand comparison with the traditional late antique Latin tracts de orthographia; the Auraicept could not possibly teach anyone how to write Irish,” (Ó Cróinín 1995:194). Against such an approach Irvine issues the following caveat: “To study medieval grammatical treatises simply as a record of linguistic theory – with the inevitable march toward the clarity and precision of modern linguistics – is to miss the point: grammatica functioned as the pathway to textual knowledge and was always part of a system of social relations that were essentially ideological,” (Irvine 1994: 20).

45 “Deutlich sollte geworden sein, daß Auraicept na nÉces nicht an den Maßstäben neuzeitlicher Grammatikographie gemessen werden darf. Es steht in einem anderen geistesgeschichtlichen Zusammenhang innerhalb des mittelalterlichen Forschungsprogramms von grammatica und ist hier ein wichtiges und frühes Dokument der Anwendung ihrer Betrachtungsweisen auf die Muttersprache,” (Poppe 1996a: 70).

37 dissolved into letters. But which letters? This question is implicitly posed by our two texts. It might be anticipated here that the answer is twofold; one according to the word but another according to actual deed. The two texts which are the focus of the present thesis are two early manifestations of a development which eventually lead to full-scale vernacular literacy.

2. ANALYSIS This next chapter of the thesis is devoted to a close reading of those sections of the texts that deal with the most fundamental unit of medieval grammatical thought – the letter. The structure is as follows: The discussion of the presentation of the vernacular alphabets in the Auraicept na nÉces and The Third Grammatical Treatise is preceded by a general introduction to the teaching of the letter as presented by Priscian in his Instt. gramm.. An excursus on the medieval notion of the letter with its three accidents and an outline of Priscian’s doctrine de litteris will serve to introduce the subject-matter. It is to be hoped that this will prove an efficient way to assess the two vernacular tracts and the degree to which they are dependent on grammatical doctrine as represented by him and the degree to which they detach themselves from it. The objective of such an assessment is not to further substantiate the claim that these vernacular grammars were generated from intercourse with the writings of the Roman grammarians and their commentators, but rather to see on what points they diverge from their sources and from each other. The latter point, though pertinent, has not previously been satisfactorily treated. The partiality for the vernacular language and alphabet finds vivacious expression in the texts and the authority and primacy of the sacred languages, first and foremost Latin, is challenged implicitly (in both works) and explicitly (in the Auraicept). This is extraordinary in the medieval West which in itself recommends comparison. Presenting the rudiments of grammar, our texts may have been composed in order to be used within the confines of the medieval classroom (see B. M. Ólsen 1884: lviii). It has been noted that the title Auraicept na nÉces, appearing e.g. in its commentary and in MVII, suggests that this was at least how this work was received during the Middle Irish period (c. 900-1200). Similar evidence of the reception of the ThGT is not, to my knowledge, available. It is, on the other hand, important that both texts put emphasis on the vernacular alphabet. There is not much evidence to suggest that this was conditioned by the requirements posed by education in vernacular literacy, if we suppose that it would affect the outcome of literary culture, represented by e.g. saga literature. I am consequently inclined to believe that this feature of the texts was not a commonplace of education, but that it represented innovation. Several passages in the commentary to the Auraicept might support this view by the fact that they draw attention towards the dissonance between the implications of what the Auraicept says and what might reasonably be thought to have been the situation in the environment in which it was made and used. The following is a rather stark indication, though

39 indirect, that the role of the vernacular alphabet was less magnified than the commentator might have appreciated: [...] .i. in caingean forsin beithi-luis-nin in ogaim .i. inn oguamma no forsin bitheolus litterdha inn oghaim (Calder 1917: ll. 394-95). 46 An expression such as ‘the undying literary knowledge of the Ogam’ is, one would think, indicative of anything but general esteem.47 Already in the Codex Sangallensis 904 the sporadic use of Ogam is merely an alternative to writing with the Latin alphabet (McManus 1986: 12). The grammatical features of the language and the norms of spelling in these glosses are no different from those of the others. The scribes were not familiar with orthodox (as opposed to manuscript-influenced scholastic) spelling, nor with the grammatical system on which it was based (loc. cit.). This argument will not be further pursued at this point; suffice it to say that if the alphabets seem not to have been used to fulfil any practical purpose, then our texts require us to think about the reason for their inclusion. It is reasonable to think that this was done in order to strengthen their lines of argumentation and thus, I would suggest, to support the general undertaking of the treatises – their attempt at a transfer of grammatical doctrine onto the vernacular. The preoccupation with the vernacular alphabet is a marked feature even of the earlier strata of the Auraicept na nÉces. In Ahlqvist’s edition, the emergence of the Ogam is closely intertwined with the emergence of the Irish language itself. None of the two treatises propose any change in current orthography. Their originality does not lie in a reassessment of scribal practice or of the correlation between speech and writing. This will become apparent through close parallel reading of selected passages, to which the bulk of the present chapter is devoted. These passages must, however, be seen in the light of medieval linguistic theory.

2.1. Conceptualizing the Minimal Parts of Speech The ThGT is introduced by a section on the nature of sound and its subcategories (1.7.). For a discussion on the sources for this passage the reader is referred to the commentary section of Wills’ edition (2001: 108ff.). The role of the minimal parts of speech is articulated in the following passage:

Stafr er hinn minzti hlutr raddar samansettrar sem rita má. Ok er stafr kallaðr hinn minnsti hlutr eða úskiptiligr í því, sem heyrir allri samansetning stafligrar raddar, því at vér skiptim bœkr í capitula, en capitula í clausur eða vers, en clausur í málsgreinir,

46 “[...] the question on the Beite Luis Nin of the Ogham, that is, ind oguamma of the perfect alliteration, or on the undying literary knowledge of the Ogham” (Calder 1917: 31). 47 The same expression is also found in a passage on the semivowels (Calder 1917: l. 509).

málsgreinir í sagnir, sagnir í samstǫfur, samstǫfur í stafi. En þó eru eigi stafir náttúrulega úskiptiligir, því at stafr er rǫdd, en rǫdd er lopt eða af lopti formerat (normalized after B. M. Ólsen 1884: 37).48

There is nothing corresponding to this passage in the Instt. gramm., but in the Commentum in artem primam Donati, written by the Carolingian scholar Remigius of Auxerre (d. ca. 908) there is a passage strongly reminiscent of it, though it is not identical (see Hreinn Benediktsson 1972: 42n3). Also Ælfric has a passage which corresponds rather closely (ibid.), giving at least two possible channels of influence. The discussion on the nature of sound preceding the above quotation in the ThGT probably aided Óláfr in observing that the ‘letter’ (stafr) is not actually indivisible, for it is voice, which is generated from air. Instead of introducing a theoretically consistent division between ‘letter’ and ‘voice’, Óláfr picks up the indecisiveness of the Instt. gramm. by the equation stafr er rǫdd (‘a letter is voice’). Óláfr states, following Priscian, that the letters may have several ʻmeaningsʼ (merkingar). Though appropriated from the Instt. gramm., this statement is eminently true when applied to the 16- runes’ fuþark. The commentary to the Auraicept operates with a slightly differing hierarchy of linguistic and literary units, which are moreover given in the reverse order:

Dotuitead i litrib .i. tinntuit asinn aicniud sin i llitrib. Doaitnead .i. taitnit asna litrib sin i

foclaib. Doaidhbead .i. du-aispenait do eolchaib eistib .i. a cialla 7 cairechtaire .i. fuatha na

48 “The letter is the smallest part of connected vocal sound which can be written, and the letter is called either the smallest unit, or not divisible into that which can be heard in all composition of writeable voice, because we divide books into chapters, and chapters into paragraphs or verses, and paragraphs into sentences, sentences into words, words into syllables, syllables into letters. However, letters are not naturally indivisible, because the letter is voice and voice is air or formed from air,” (Wills 2001: 81 [§2.2]). Note: All manuscripts have the form formerat, which would agree with lopti, and not with rǫdd (as implied by Wills’ translation). Agreement with rǫdd would demand the form formeruð. Final -t could have been accounted for by hypercorrection due to loss of the distinction between final /t/ and /þ/ (realized as its voiced allophone [ð]), but this would leave the non-umlauted form unexplained. Thus, in order to avoid drastic normalization one would have to render the phrase as ‘from formed air’. This has the virtue both of corresponding more closely to Priscian’s syntagm aere icto (GL vol. II: 6), which might explain the odd syntax, and semantically to the composition lopts- formeran, which is used in the following sentence of the text.

liter. Dotiaghat asna foclaib sin i comighib 7 i sreathaibh roscaigh 7 fasaigh 7 airchetail (Calder 1917: ll. 333-37).49

Although these passages are relatively independent of the possible antecedent(s), and thus from each other, they belong to the same paradigm of linguistic thought. The genres mentioned in the passage in the Auraicept are consistent with its larger focus on poetic composition which is emphasized by Acken (2008). It is important to note that cáirechtaire is equated with fúatha na liter ‘the forms of the letters’ and thus, though this is nowhere stated, with figura.

2.2. Linguistic Theory: The Letter The concept of the letter in medieval linguistic theory is itself complex enough as to merit a few general remarks here. The origin of alphabetic writing by definition involved a reasonably accurate aural-based analysis of the constituents of speech. So does the adaptation of an existing alphabet for the purpose of recording a new language. It is only natural that the theoretical approach to this task has itself evaded the written records. Hence no alphabetical system of writing in Europe is accompanied by a clarifying discourse. The nature of alphabetic writing made it a convenient framework for the discussion of phonetics, and it significantly encouraged the study of written rather than spoken language. This was of great consequence to linguistic thought from Antiquity throughout the Middle Ages. Language study was the study of the written word. The letter (Lat. littera) was conceived of as a minimal unit of scriptible utterance. A clear distinction between the graphic level of representation and the phonic one is not invariably made. This is a natural consequence of the fact that the object to be handled by the grammarians was written literary output by earlier authors. In modern linguistics a distinction between the graphic level and phonic level is accomplished by the terms grapheme/phoneme.50 The phonemes are the

49 “They fall into letters, i.e. they are converted out of that primary nature into letters. They shine, i.e., out of these letters into words. They show to the learned out of them, to wit, their meanings and their characters, i.e., the forms of the letters. They come out of those words into texts, and series of proverb, commentary, and poetic composition.” (Calder 1917: 25).

50 The phoneme is the abstracted minimal sound-unit of language. It is established by performing a commutation test or establishing a minimal pair where each unit is distinguished from the other through contrast of a single phoneme. The allophone is the surface-form of the phoneme. The phoneme belongs to the system, while the allophone is dependant on phonetic context. Example: the

42 minimal distinctive units of spoken language. The graphemes are the minimal units of the written word. The relationship between the two is established solely by habit and is arbitrary. The medieval linguist had inherited an alphabet with several merely conventional features, such as the incidental ordering of the letters. But ‘arbitrary’ was not a relevant term to the medieval grammarian, who shared with his fellow scholars a concern for the origins of phenomena and their implications. Of considerable importance too, was the proper enumeration of quantitative facts, which themselves could be used to raise theoretical superstructures. Consequently much ink was spent on answering questions as ‘why are there five vowels?’ and ‘how should the vowels be arranged?’ (examples taken from Law 2003). Such features are notably present in exegetical Latin grammars (for which see Law 1982) and were carried over into the vernaculars. A distinction between sound value and its written representation is made by the early grammarians in some cases, however, by the use of the term elementum, current in Aristotelian and Stoic physics, to denote the actual sound value and littera to denote its written representation. The ‘element’ is said to be the minimal unit of speech, but ‘speech’ in this context is a language learnt from written texts and available only through letters. Hence the emphasis on written texts influenced the notion of the nature of the minimal unit and a certain circularity is inevitably part of the presentation. Latin littera corresponds to Greek γράμμα, while elementum corresponds to στοιχεῖον (as used in the Alexandrian school, notably by Dionysios Thrax). The latter term was used in physics before it entered grammatical terminology (Irvine 1994: 100), and later commentators, such as Sergius, would note that the letter resembled the atomos of the philosophers (GL vol. IV: 475).

Old Icelandic phoneme /þ/ was realized as the unvoiced dental fricative [þ] in word-initial position, while word-internally or in auslaut it was realized as the voiced dental fricative [ð]. The brackets denote allophones and the slashes denote phonemes. The grapheme is analogous to the phoneme. It is the minimal unit of a given writing system. Its surface forms are called allographs. Example: high ‘ſ’ as compared to ‘s’, which are two separate glyphs and could be used (for instance in Old Icelandic mss.) interchangeably to denote the grapheme ‹s›. The letter is duly introduced in its medieval context in 2.1. In modern usage it is equal to the grapheme; it is the minimal unit of the graphemic system. This is a bit unfortunate for the terminological consistency of the present thesis. I have deliberately avoided undue disambiguation by the use of modern terminology in cases where I try to represent medieval arguments.

Diomedes (fl. late 4th century) is quite sensitive to the distinction between littera and elementum and deals with this at some length. His definition of ‘letter’ reads: littera est pars minima vocis articulatae ab elemento incipiens una figura notabilis (GL vol. I: 421).51 Diomedes is frequently quoted by Priscian in his Instt. gramm., who to a certain extent included the discussion on the distinction between letter and element, but whose main definition of ‘letter’ is rather traditional: litera est pars minima vocis compositae [...] (GL vol. II: 6).52 This is then elaborated upon: possumus et sic definire: litera est vox, quae scribi potest individua (ibid.).53 The following extract proves that Priscian too was sensitive to the distinction:

Litera igitur est nota elementi et velut imago quaedam vocis literatae, quae cognoscitur ex qualitate et quantitate figurae linearum. hoc ergo interest inter elementa et literas, quod elementa proprie dicuntur ipsae pronuntiationes, notae autem earum literae. abusive tamen et elementa pro literis et literae pro elementis vocantur (GL vol. II: 6-7).54

Here the elementum (which is equalled with the ʻpronunciation itself’) is introduced as the actual minima pars of speech. The addition velut imago quaedam vocis literatae ‘as a certain image of writable speech’ makes use of a term which is derived from the term to be defined. Furthermore, defining the letter as an image or symbol of an actual atom of speech, rather than as the atom itself, withdraws from it its level of abstraction. The affinity between the letter and the sound it represented was too tight to be loosened during the Middle Ages. We might even suppose that any incentive to do so was absent; even in the vernacular grammar any thoughts on the arbitrariness of the signs remain unexpressed. As it may be, this is a key to understanding the inclusion of Ogamic or runic graphemes in grammatical discussion. The sounds of speech were closely interconnected with the letters. Priscian’s discussion of the element does not have much consequence for the discourse that follows, as he chooses to

51 “A letter is the smallest part of articulate utterance; it takes its beginning from an element [unit of sound] and can be marked in a figure [written character]” (Irvine 1994: 98). 52 “The letter is a minimal part of composite speech”. 53 “We might also define [it] thus: ‘the letter is a sound which may be written individually.” 54 “A letter, then, is a note for an element and as if a kind of an image of scriptible sound, which is understood from the quality and the number of the lines of the figure. This is the difference between the elements and the letters: that the pronunciations themselves are properly called ‘elements’, while their notes are called ʻletters’. Incorrectly, however, ʻelements’ are called ʻletters’ and ʻletters’ are called ʻelements’.”

44 stick with ʻimproper’ usage by applying the terms interchangeably and without any obvious regard to context.55 As a result of this ambiguity Priscian uses the term figura to specify when he is dealing with the actual letter-shapes: sunt igitur figurae literarum quibus nos utimur viginti tres, ipsae vero pronuntiationes earum multo ampliores [...] (GL vol. II: 7).56 In actual usage, then, Priscian is not very conscientious in the application of his most fundamental terminological units. This was inherited by the vernacular grammarians and in the Auraicept even led to the development of a major rhetorical topos (see 2.7.). The medieval letter had features in common both with the modern phoneme and with the grapheme – but significantly without the ʻintentional depth’57 of either, that is, it communicated neither the idea of a smallest functional part of speech nor of writing effectively. It was nevertheless used in grammatical discourse as if it did. It will be seen that the theory of the letter as outlined above permeates the vernacular texts.

2.3. The Divisions of the Alphabet The four divisions of the alphabet (see 1.3.) will be discussed in more depth here. The classification of letters found in the works of the Latin grammarians is in its origin an adaptation of the analysis made by the Greeks. The classification of the sounds of the Greek language into vowels and consonants was arguably codified by Dionysius Thrax (1.9.). The vocales are identified as A E I O U. The semivocales are F L M N R S X. The mutae are B C D G H K P Q T, while the last group, graecae, the more recent borrowings from the Greek alphabet, are Y Z. This is the system as it is laid out in the Instt. gramm., which is, however, far older; Quintilian (1st cent. AD) presents the same system, although without the addition of Y and Z (McManus 1991: 28). An anonymous commentary on

Donatus quotes Varro (2nd cent. BC): Varro dicit consonantes ab e debere incipere, quae

55 Consider for instance the following: cum enim dicimus non posse constare in eadem syllaba r ante p, non de literis dicimus, sed de pronuntiatione earum [...] (GL vol. II: 7). “For as we state that ‘r’ may not precede ‘p’ in a single syllable, we are not talking about letters, but of their pronunciation”. 56 “Thus the figures of letters which we use are 23, but their actual pronunciations are many more [...]”. 57 Translated from the Norwegian intensjonsdybde; a term used by the philosopher Arne Næss to denote the speaker’s own awareness of possible and reasonable interpretations of his statements (or vocabulary). Misunderstanding and ambiguity are reduced with greater depth of intention if it is shared by both parts. I think the concept is useful in reading medieval grammatical literature, working as an aid in understanding the grammarians on their own terms.

45 semivocales sunt, et in e debere desinere, quae mutae sunt (Ahlqvist 1982: 9).58 Thus the Greek division of the σύμφωνα (‘consonants’) into ἡμίφωνα (‘semivowels’) and ἄφωνα (‘mutes’) had most likely been adapted by Latin grammarians as early as the 2nd century BC. The consonantes were either semivocales or mutae. The formation of the name of each letter had a certain influence on its status; we will see how the potestas of the letters was closely related to it. The Instt. gramm. produces no accurate explanation of the rationale underlying its groups of letters. The excerpt below provides an attempt at a definition according to the pronunciation of the names of the letters:

Vocales igitur, ut dictum est, per se prolatae nomen suum ostendunt, semivocales vero ab e incipientes et in se terminantes, absque x, quae ab i incipit per anastrophen Graeci nominis ξῖ, quia necesse fuit, cum sit semivocalis, a vocali incipere et in se terminare, quae novissime a Latinis assumpta post omnes ponitur literas, quibus Latinae dictiones egent [...] mutae autem a se incipientes et in e vocalem desinentes, exceptis q et k, quarum altera in u, altera in a finitur, sua conficiunt nomina (GL vol. II: 8).59

This was a traditional classification already during Priscian’s time. The letters are divided into subgroups according to their traditional pronunciation when they are itemized: The vowels show their names when they are enumerated: a, e, i, o, u. The names of the mutes end in –e (be, ce, de ...), while the semivowels are those whose name begin with an e- and end in the sound they denote (el, em, en ...). This definition must have originated as a rule of thumb in the teaching of grammar; it is both imprecise and at odds with other definitions circulating in grammatical texts. More specifically, it is in disagreement with other definitions within Priscian’s text itself, which explains the remark concerning ξῖ.60 The defining feature of this

58 “Varro says that those consonants which are semivowels ought to begin in ‘e-’, and that those which are mutes ought to end in ‘-e’.” 59 “Vowels, then, as is said, reveal their names when pronounced through themselves , but semivowels when they begin in ʻe-ʼ and end in themselves, with the exception of ʻxʼ, which begins in ʻi-ʼ through anastrophe (inversion) of the Greek name ξῖ, wherefore it was a necessity, that it be a semivowel, beginning in a vowel and ending in itself, which, being the latest to be adopted by the Romans, is placed after all the letters which Latin words need. [...] Mutes, however, begin in themselves and end in the vowel ʻ-eʼ, with ʻqʼ and ʻkʼ excepted, of which the one ends in ʻ-uʼ, the other in ʻ-aʼ when they produce their names.”

60 Priscian’s inconsistencies, as we have seen regarding the letter and the element, owe a great deal to his use of varied sources (see further Luhtala 2005).

46 letter, i.e. the name (ksi), is altered (to iks) to make it fit the scheme (VC) – because, according to the Greeks, it was a semivowel. Priscian reveals a hierarchy of definitions. The Greek one is in turn aural-based, and does therefore not rely on the name of the letter. A different treatment is given to ʻfʼ ef, which Priscian considers to be a mute letter, despite the fact that its name would suggest otherwise and that he violates the opinion of ʻmost grammariansʼ (GL vol. II: 9, 11). Two additional definitions of the letter-groups are found in the first book of the Instt. gramm.. These are: 1) the letters are divided into three groups according to their relative amount of voice: hae ergo quantum vincuntur a vocalibus tantum superant mutas (GL vol. II: 9).61 2) The letters are grouped according to position: The vowels may stand alone and are furthermore required if the mutes are to be pronounced. The semivowels do not occur before ʻlʼ or ʻrʼ in one syllable. Unsurprisingly, all three definitions render the same three groups of letters that were supported by tradition, but which did have a solid basis in phonetic theory. All of Priscian’s mutae are stops (with the exception of ʻh’ – according to him this is only arguably a mute), whereas his semivowels roughly correspond to our continuants (fricatives, nasals and liquids). Accordingly, this term is not to be confused with the modern one, which denotes glide vowels. It hardly escapes notice that every attempt at defining a ʻletter’ invokes sound.62 Phonemic theory is, on the other hand, a relatively recent development within linguistics (its fundamentals being worked out by de Courtenay during the 1870s). Written representation is generally more conservative than spoken language, and is therefore grounded in tradition and not in synchronic accuracy. This was already an issue when Priscian wrote his Instt. gramm.. No conceptualization of allophonic variation existed in medieval language theory. When reading the medieval grammarians, therefore, it is important to keep in mind that several distinctions which are fundamental to modern language theory were not made in any systematic fashion in this period. The littera was nevertheless a sophisticated concept and formed the basis of the theoretical framework during the Middle Ages. A convenient discussion on the letter is offered by Hreinn Benediktsson in his edition of FGT (1972: 42ff.).

The Irish linguist divided his alphabet into vowels and (mute) consonants. This is a point of special pride to the Ogamist. The visual distinction of the groups of Ogam letters has been

61 “These [i.e. the semivowels] are surpassed by the vowels to the extent they supersede the mutes”. 62 E.g. Litera est pars minima vocis compositae [...] (GL vol. II: 6), possumus et sic definire: litera est vox [...] (loc.cit.).

47 mentioned in the introduction (1.3.). A statement such as the following is reiterated in the commentary: rolatha iaramh a fedha for leith 7 a tæbomna dano for leith, co fil cach æ dibh fo leth o’rlaile (Calder 1917: ll. 1058-59). “Therefore its vowels were placed apart and its consonants also apart so that every one of them stands apart from the other,” (ibid.). This visual separation is unparalleled in the Latin alphabet, and the fact that also the consonants are visually isolated in three groups is passed over in silence. The Auraicept argues that the rationale behind the twofold division of the Latin alphabet and that behind the twofold division of the Ogam was not the same. Both alphabets are divided into vowels and consonants. The Auraicept uses the borrowed Latin terminology to denote the Latin letters, while the letters of the Ogam are represented by the terms fid ‘vowel’, which might also denote the more general concept of ‘letter’, and taebomna ‘consonant’. The Ogam signary stands apart from its Latin counterpart as the vowels are separated from the consonants both visually and with regard to arrangement. Of the letter itself it is asked:

Cest, in gne no in cenel in fidh? Is cenel emh: 7 masa chenel, cadet a gnee? Ni ansa. Fid

særda 7 fid aicenta .i. fid særda fid inn ogaim; fid aicenta immorro fid na cailli. Et fid inn

ogaim, in gne no in cenel? Is cenel ecin, ar techtaid gne .i. fid airegda, 7 forfidh, 7 tæbomna. Is e insin in cenel cenelach gneach .i. fid (Calder 1917: ll. 853-58). 63

The hierarchy of species, themselves subordinated to ‘genus’, transcends the linguistic sphere. When it comes to the cause of the division of the letters of the Ogam, the commentary applies what might be its preferred dichotomy, namely that between the ‘natural’ (aicentae) and the ‘artificial’ (saerda). The Latin division, being made ‘according to the word’, is artificial, while the Irish division is ‘according to nature’. The background for this interpretation is the semantic ambiguity of fid and the redundancy of some of the letter-names. Consider the following:

At the time of the composition of the Bríatharogaim the meanings of the letter-names were still known, but several of them had become low-frequency words and were on the way to semantic redundancy. This, combined with the love of schematism and the fact that

63 “Query, is fidh, wood, a species or a genus? It is a genus certainly; and if it be a genus, what are its species? Not hard. Artificial wood and natural wood, to wit, artificial wood is the Ogham letter; natural wood, however, is wood of the forest. And as to wood, letter, of the Ogham, is it a species or a genus? It is a genus necessarily, for it has species, to wit, principal wood, vowels; cross wood, diphthongs; and side-woods, consonants. That is the genus generic and specific, i.e., wood,” (ibid.).

the largest single semantic category among the names was an arboreal one, paved the way for the ‘alphabet végétal’ fiction which one sees developing in the glosses and culminating in the tradition of the Bríatharogaim preserved in the commentary to the Auraicept na nÉces (McManus 1988: 167-68).64

The redundant names could therefore be identified with trees themselves (see also McManus 1991: 35). The arboreal taxonomy situates the letters of the Ogam firmly in the natural world and enlivens the terminology associated with them by allowing the use of a kind of imagery which was unavailable to ‘the Latinist’ (this is the term the Auraicept uses to denote its rhetorical opponent – see 2.9.).65 An aesthetical triumph over the Latinist, the arboreal fiction of the Irish is not, however, only relevant to schematism, as suggested by McManus, or aesthetics, as suggested here. I would like to propose that, to the medieval scribe or reader, it must have given primacy to the Irish vernacular alphabet, for nature derives from God.66 It is therefore a distinction which borders on theology, and as such it reverberates through other parts of the commentary. It further underscores the importance of exegesis as an interpretative tool in a grammatical context.

The groups of letters in the ThGT are those of traditional Latin grammar. They are given Icelandic names, influenced by the Anglo-Saxon tradition (see Gade 2007) but are not otherwise influenced by possible native theory. The division into four families of letters are given no attention. Neither does Óláfr dwell on the traditional runic names. They are named

64 The Bríatharogaim were edited together with the Auraicept in Calder 1917 (p. 284 ff.), and more recently in McManus 1988 (cited below under 5.3.). They contain kennings in relation to the Ogamic letter-names. 65 A terminology specifically suited to Irish, and distinct from Latin, might in any case have been regarded as beneficial in itself (discussion in Ahlqvist 1989). There are also traces of native terminology in the Auraicept (Ó Cuív 1983: 2); though to his argument I would like to add that a systematic distinction between native/foreign in the Auraicept is not evidence of the antiquity of terminological units, and that the native terminology (see 2.6. on consonants) could bee seen to harmonize well with Latinate theory; that such theory therefore provides no sound terminus ante quem. The Ogam is itself a product of Latinate influence. See also Russell (1996) on a few items of possible native origin related to grammatical gender. 66 A similar dichotomy is applied by Dante Alighieri in his De vulgari eloquentia, where the vulgaris locutio is naturalis, while the locutio secundaria, or Latin, is artificialis (Tavoni 2011: 1136-37); “Perché la natura deriva direttamente da Dio, mentre l’arte imita la natura,” (ibid.). Consider also Romans 1: 19- 20.

49 after the acrostic principle (see 1.3.), as are the letters of the Ogam, but whereas this is a maior topos in the Irish treatise (see 2.7. on the semivowels), that is not the case in the ThGT. The cause might be that Óláfr aimed to follow the text of Priscian more closely. Óláfr’s text is nevertheless far from being an accurate rendering of Priscian in Icelandic. This is therefore not a fully satisfactory explanation. The contrast with the Auraicept (2.7.) makes it clear that he in fact could have dwelt on the traditional grouping of the letters and that his argumentation could have benefitted from this. One might be justified in thinking that Óláfr chooses to soft-pedal the differences between the runic alphabet and the Latin alphabet when they are not blatantly obvious, as is the case with their total number (2.5.). His starting point is the unity of the traditions (2.10.) and so reference to external authority, such as Greek practice (2.5.), is made when the scheme and the facts do not go too well together.

2.4. The Accidents of the Letter The idea that a letter possessed three accidents or attributes was a fundamental one in medieval grammatical thought. We have seen above (2.1.) that such a threefold characterisation could provide no precise analytical device. The letter could both figure as an abstract entity (representing the elementum/minima pars) or as a concrete one (itself being the minima pars). What consequences might follow from the choice of level of abstraction was not further explored. Hence littera remained the fundamental analytical concept, despite the fact that Priscian and a grammarian such as his predecessor Diomedes seemed to regard elementum as the proper minima pars (2.1.). The three accidents are presented in the Instt. gramm. in the following manner:

Accidit igitur literae nomen, figura, potestas. Nomen, velut a, b. [...] Figurae accidunt quas videmus in singulis literis. Potestas autem ipsa pronuntiatio, propter quam et figurae et nomina facta sunt. quidam addunt etiam ordinem, sed pars est potestatis literarum (GL vol. II: 7, 9).67

67 “To the letter belongs name, figure and value. Name, such as ʻaʻ, ʻb’. [...] To figure belong those [forms] that we see in the individual letters. Value, however, is the pronunciation itself, on account of which both the figures and the names are made. Some even add ʻorder’, but this is part of the value of the letters.”

This, disregarding the relationship between littera and elementum, and, more importantly between potestas and elementum, might be thus visualized:68

nomen ʻname’

littera figura ʻsymbol’

potestas ʻvalue’

These attributes were playing their part when the letters of the alphabet were arranged in groups. The order of the letters within the alphabet was traditional and neglected their individual features (see 2.1.). Priscian does not give a complete list of the 23 letters of the Latin alphabet, but they are invariably arranged according to their traditional order when listed group-wise.

The three accidents of the letter are not dealt with as such in the Auraicept. Still, traces of them are found in its theoretical fundament. The following is a statement on the analysis (tomus) of the letters:

6,1 Catte tomus fri fid? 6,2 Ni ansae: co·feisser a llin 7 a n-úati, a mét 7 a lagat, a cumang 7 a

69 n-éccumang, a nert 7 a n-aimnert (Ahlqvist 1982: 50).

At first glance one might think that cumang in this extract translates potestas, but the very last sentence of the canonical text reads: lánchumang intib uilib eter feda 7 táebomna co n-aurbu úath (Ahlqvist 1982: 51). “There is full force in them all, both vowels and consonants, with the exception of ‘h’.” (cf. Ahlqvist’s translation). The term is therefore not obviously used to denote the different values of the letters, but rather their ʻforce’ in a more narrow sense, viz. their ability to express sound at all. The canonical core of the Auraicept, owing to its brevity, has little to offer when it comes to the actual use of these terms. Yet it should be noted that the

68 Contrast potestas autem ipsa pronuntiatio (cited above) with elementa proprie dicuntur ipsae pronuntiationes [...] (GL vol. II: 6-7). A certain confusion is apparent. The figure reproduces the one provided in the edition of FGT (Hreinn Benediktsson 1972: 44). In its original context it is used to illustrate the more general medieval conception of the letter. 69 “6,1 What is the nature of analysis in respect of letters? 6,2 Not difficult: that you may know their maximum and their minimum number, size, force and function.” (Ahlqvist 1982: 50).

51 letter (fid) is said to be one of the ‘seven items by which Irish is to be measured/analysed’ (Ahlqvist 1982: 50). The commentary to the above quotation, on the other hand, provides a few important remarks on the actual application of the technical vocabulary:

Caite deochair eter a cumang 7 a neart? A cumang cetamus: Intan gabaid guth a n-ænur .i.

a no o no u. A nert immorro intan dos-bere primshuidhiugud i sillaib, amal ata bais 7 lais (Calder 1917: ll. 965-68).70

In this extract cumang does show affinity with potestas as defined by Priscian (being the ʻpronunciation itself’), while nert is associated with internal rhyme in stressed syllables. Another way of expressing the value of the letters is found in an expression as x gabus greim (+ gen.) ‘x which takes the effect of ... (gen.)’ is used several times in the commentary (e.g. Calder 1917: ll. 106-08). This is reminiscent of the Latin expression x transit in y, and might be seen as another reflex of the Latinate tripartite division. The following question introduces a section where this principle is investigated, first with respect to the vowels, then with respect to the consonants:

Conadar didiu isin bethi-luis-nin: Caidi in fid gabus greim taebomna 7 in fid gabus greim

da tæbomna 7 in fid gabus greim focail 7 in fid na geibh greim taebomna na feda na focail? (Calder 1917: ll. 1375-78).71

It is striking that the three attributes of the letter did not make a greater impression on the Irish; that the presence of this part of traditional grammatical teaching has to be inferred. The Irish fondness for enumeration in triads as an epistemological tool is otherwise well attested. The letter-name was of great concern to the medieval grammarian. Being one of the three accidents of the letters, the name had to fit the pronunciation. Diachronic sound-changes distorted the harmony, which was reintroduced by the assignment of new ‘cosmetic values’ (McManus 1988: 127) to some of the letters. This might therefore have been part of an effort

70 “What is the difference between their force (cumang) and their strength (nert)? Their force first: When they provide voice on their own, i.e. ʻa’ or ʻoʻ or ʻuʻ. Their strenght, however, when a prime position brings them into a syllable, as bais and lais,” (ibid.). 71 “It is demanded, too, in the Beithe Luis Nin: What is the vowel that takes the force of a consonant, and the vowel that takes the effect of two consonants, and the vowel that takes the effect of a word, and the vowel that does not take the effect of a consonant, vowel, or word,” (ibid.).

52 by the scholastic Ogamists to make their vernacular alphabet compatible with the theory of the letter as expounded in current grammatical texts, in addition to ‘accounting for an embarrassment of riches from an earlier period’ (loc. cit.).

The FG represented the three attributes (tilfelli) of the letter with the terms nafn, líkneski and jartein (Hreinn Benediktsson 1972: 44). The letter of the FG is named stafr – a term which was probably taken from runic terminology (hence already a technical word and analogous to fid) and which took on a meaning comparable to littera. The same term is used in the ThGT. Of the three attributes, however, only nafn remains in use. This obviously translates nomen. The two other terms72 used in the FGT are replaced by fígúra and veldi respectively.

Annat tilfelli stafs er fígúra. Þat er mynd eða vǫxtr stafanna, svá gerr sem nú er ritat. Þriðja tilfelli stafa er máttr, ok er þat sjalfframflutning stafa ok merking þeirra. Fyrir þann sama mátt stafanna eru bæði fundin nǫfn þeirra ok fígúrur. Sumir meistarar kalla skipan hitt fjǫrða tilfelli stafs, enn þat kallar Priscianus einn part þann er mætti stafsins heyrir (normalized from B. M. Ólsen 1884: 45). 73

The second attribute, fígúra, is further defined as the mynd or the vǫxtr of the letters. The last of these terms are also used in the FGT. Whereas the FG, according to Hreinn Benediktsson, is able to make a distinction between his líkneski (i.e. fígúra) and vǫxtr, such a fine-tuned distinction is not part of the nomenclature of the ThGT (or of the Latin grammarians for that matter).74 According to Priscian each vowel-letter may be pronounced in ten or more ways. The term he uses is ipsae pronuntiationes (GL vol. II: 7). This is reflected in the ThGT in a

72 The reader is referred to the relevant section of the introduction to FGT (Hreinn Benediktsson 1972: 53 ff.) for a discussion of the use of these. 73 “The second characteristic of the letter is shape (figura), that is the form or shaping of letters, made as is now written. The third characteristic of the letter is value (potestas) and that is the actual pronunciation of letters and their signification. The characteristics of name and figura are both found before the value of letters. Some scholars call order the fourth characteristic of the letter, but Priscian calls it only a part, which belongs to the characteristic of value of the letter.” (Wills 2001: 89 [§4.4]).

74 Hreinn Benediktsson might overstate the terminological propriety exhibited by the FG. When the FG does in fact use his terms interchangeably, that is, according to Hreinn, because no distinction is demanded.

53 passage which has been taken to imply the existence of tonemic accent in Icelandic at the time (compare modern Norwegian) (as argued in Myrvoll & Skomedal 2010). If this is the case, Óláfr has detected an authentic distinctive feature in his language by applying (or forcing) on it the Latinate accents - which were in reality unsuited for Icelandic, just as they arguably were for Latin, which had in turn borrowed them from the Greek system. The following passage, then, is both a learned construction and the product of an effort to accommodate real linguistic features, not present in the source language, within the set framework. Thus it becomes a source to actual linguistic development.

[...] enn þó eru merkingar þeirra miklu fleiri, þvíat Priscianus segir, at hverr raddarstafr hafi tíu hljóð eða fleiri. Svá sem ‘a’, ef þat er skamt, hefir fjórar hljóðsgreinir. Hvassa hljóðsgrein fyrir útan áblásning ‘h’, sem hér: ʻári’. Þunga hljóðsgrein enn fyrir útan ‘h’, sem hér: ... . Hvassa hljóðsgrein með áblásning ‘h’, sem hér: ʻháfi’, ok þunga hljóðsgrein með ‘h’, sem hér: ʻhàfandi’. Langt ‘a’ hefir sex hljóð: ef þat hefir áblásning ‘h’, þá berr þat annat hvárt hvassa hljóðsgrein eða þunga eða umbeygiliga sem hér: ... . Slíkt hið sama, ef þat hefir eigi áblásning hljóðar þat þrjár leiðir, sem þessi nǫfn: ‘ári’, ‘àranna’, ‘âra’. Slíkt hið sama má ok aðra raddarstafi greina, en ‘i’ ok ‘v’ hafa því fleiri hljóðsgreinir at þeir eru stundum samhljóðendr, sem í þessum nǫfnum, ‘iarl’, ʻvitr’, ok er þá ‘v’ venð kallat í norrœnu máli (normalized from B. M. Ólsen 1884: 40-42).75

The passage of course recognizes the phoneme /h/, but categorizes it as ‘aspiration’ (áblásning), just as the Latin grammarians, instead of granting it status as a ‘letter’ (see 2.7.). In the context of the attributes of the letters, ‘h’ is an interesting case, as both the Latin and the runic alphabets recognize it as a letter, but Latinate grammatical theory does not unequivocally accept it as anything but a mark of aspiration comparable to the spiritus asper of the Greeks.76

75 “[...], but still there are many more phonetic values for them, because Priscian says that each vowel has ten or more sounds, just as a if it is short has four distinctions of sound: an acute accent without an aspirated h, as here, ‘ari’ (eagle); a grave accent without aspirated h, as here: ‘á’ (river); an acute accent with aspirated h, as here: ‘hafi’ (have); and a grave accent with h as here: ‘hafandi’ (having). Long a has six sounds: if it has an aspirated h, then it has either an acute accent or grave or circumflex, as here [...]. Likewise, if it does not have aspiration, it is sounded in three ways, as in these words: ‘ari’, ‘aranna’, ‘ara’. Likewise, one can also make distinctions in other vowels, but i and v have further distinctions of sound, in that they are sometimes consonants, as in these words: ‘iarl vitr’ (wise earl), and then v is called ‘venð’ in the Norse language.” (Wills 2001: 85 [§3.5]).

76 The grapheme ‹h› became redundant in written Latin due to the loss of /h/ in Vulgar Latin.

2.5. Graphemic Inventory The total supply of letters in the vernacular alphabets could be a source of pride to the grammarians. In the Irish case due emphasis is put on the existence of ‘additional characters’. Their presence in the Ogam gives it a total of 25 letters, which is more than the Latin alphabet is able to muster (i.e. 23). This is a quantitative victory over the other alphabets, and, by extension, over the other languages: for these letters represent sounds of speech (see also 2.10.). Óláfr could hardly opt for such a quantitative solution to strengthen the prestige of the 16-runes’ fuþark. Instead he noticed that Priscian attributes the same number of letters to the ‘most ancient of Greeks’ (Gl vol. II: 11). Thus a triumph of temporal distance was within reach: Stafanǫfn eru sextán í danskri tungu í þá líking sem Girkir hǫfðu forðum daga (normalized after B. M. Ólsen 1884: 40). “There are sixteen letter-names in the Norse language, just as the Greeks had in former days” (Wills 2001: 85 [§3.2]). The practice is also vested with authority independent of Latin, and this is a crucial point, Latin being the lingua franca of the time and the prime competitor to the vernacular. As is the case in the Latinate grammatical tradition, ‹h› is not regarded as a proper letter (2.4.). Therefore, in order to make the 16-letter fuþark actually reach a total of 16 letters, without contradiction, Óláfr includes the rune f twice in his discourse; once among the semivowels, albeit reluctantly, and then as one of the mutes (B. M. Ólsen 1884: 44). He does not mention ‹h› at all in Mg, although it is part of a pangram which he quotes (see 2.9.) (B. M. Ólsen 1884: 46). The status of ‹h› has been briefly touched upon in the preceding sub-chapter (2.4.) and the effect this has on the presentation in the ThGT is therefore in no way incidental. The text-internal rationale for the application of Latinate doctrine on the Icelandic language and alphabet (see 2.10.), which is articulated in the introduction to M (B. M. Ólsen 1884: 59- 60), is given precedence over minutiae. Yet this is conceivably not as much a case of misrepresentation or irresponsible handling of facts as it is a change of perspective and of bestowing theory upon the evidence.

The following sections (2.6.-2.8.) will deal with discourses relating to the different sub-groups of the alphabets, presenting the individual letters where they are of relevance to the argument. Attention is targeted at the development of theory based on those fundaments which are provided by Priscian. To some extent these sections pick up the threads of 2.3. above.

2.6. Vowels Ex his vocales dicuntur, quae per se voces perficiunt vel sine quibus vox literalis proferri non potest, unde et nomen hoc praecipue sibi defendunt; ceterae enim, quae cum his proferuntur, consonantes appellantur. sunt igitur vocales numero quinque: a e i o u (GL vol. II: 9).77

The fusion of the inherited Ogamic alphabet and Latin grammatical theory is nowhere as straightforward as with the basic vowels. These have a one-to-one correspondence, which is recognized in transliterations in the manuscripts, although their order within the respective alphabets is not the same. This is of course a fact which owes to the historical origin of the Ogam (1.3.) and the composition of the phonemic system of Primitive Irish (McCone 1996: 24), but these circumstances were unknown to the medieval literati. The acrophonic letter- names, however, had transmitted the correct values of the letters as far as the vowels are concerned, and so these were easily deducible. The five basic vowel-letters were at some point accompanied by another set of five letters. On the authority of the maxim lectio difficilior potior, some of these had consonantal origin (McManus 1991: 143). McManus has invested much effort in retracing the value of the Ogamic letters and the reader is referred to his publications (consult 4.3.) for a comprehensive account. Instead I will concentrate attention at one of the arguments which are developed in the Auraicept regarding the additional characters. The uncertain value of the additional characters made them especially suitable to heuristics, and the Early Irish grammarian was not above eulogy (in more than one sense of the word). The letter combinations which are given below (bottom right) are used in manuscript orthography and were assigned to the additional Ogamic characters. The forms of the letters are not of much concern to the following argument (see 1.3.):

77 “Of these, those are called ‘vowels’ which carry out voice through themselves, or without which scriptible voice may not be pronounced, whence they also defend this name excellently; for the others, which are pronounced together with these, are called ‘consonants’. The vowels, then, are five in number: ‘a’, ‘e’, ‘i’, ‘o’, ‘u’.”

cárechtair (‘character’) ainm (‘name’) Equivalent ᚐ ailm a ᚑ onn o coic feda airegda ᚒ úr u (‘five primary vowels’) ᚓ edad e ᚔ idad i ᚕ ébad ea ᚖ ór oi forḟeda ᚗ uilen ui (‘additional characters’) ᚘ iphín io ᚙ emancholl æ/ae

The theoretical relationship of the additional characters to the original five vowels is not consistent throughout the commentary to the Auraicept. In the oldest layer of the text they are presented as truly unique signs – invented for the purpose of denoting specifically Irish sounds. This is in fact diametrically opposed to their probable historical origin as characters which were framed in order to accommodate foreign letters (McManus 1991: 143ff.). It is articulated as follows:

[...] 7 cach són dona·airnecht cárechtair isna aipgitrib olchena, ar·íchta cárechtairi leo-som isin bethe-luis-nin ind oguim, ut est: ᚕᚖᚗᚘᚙ (Ahlqvist 1982: 48).78

Irish gains the upper hand on both the graphic and the phonic levels. The forḟeda are only arguably vowel letters – this is discussed under 2.8. below. They are nevertheless included among the vowels in a passage in the commentary, which is concerned with their formal ennumeration (compare 2.5. on the ThGT). This passage is the cause for the twofold division of the additional characters in the table above; where ᚕ EA and ᚖ OI are separated from the three others. It takes as its starting point the pseudo-historical origin of Irish at the site of the Tower of Babel (see 2.10.) and the scholars which were present there. Almost everything relating to the Irish language or alphabet was given eponymous origin at this juncture of Irish history. So also the letters were associated with historical characters (this wordplay would not have been

78 “[...] and every sound for which a sign had not been found in the other alphabets besides, signs were by them invented in the B-L-N of the Ogam, that is ...,” (ibid.).

57 recognized by the medieval Irish and is therefore not the cause of the association), according to the commentary – or to one of its many voices; the arboreal variety has already been mentioned (2.3.) and is at variance with this version, which runs: The school of Fénius Farsaid consisted of 25 scholars of particular eminence. The five principal vowels were named after the five that were noblest among these. ‘Others say’ (as·berat araile) that there are seven principal vowels and that there were consequently seven particularly outstanding scholars. The two additional characters are added promptly to the basic five (Calder 1917: ll. 255ff., 1135ff.). Their inclusion is facilitated by naming them feda (as opposed to forḟeda) in this particular passage. The resulting total is the same as in the Greek alphabet and hence the outcome is a rhetorical achievement based on alternative organization of evidence.79 The total inventory of a particular category of letters could be five or seven simultaneously. A total of ten principal vowels is then reached by the inclusion of the three remaining forḟeda ᚗ UI, ᚘ IO, and ᚙ Æ (Calder 1917: l. 427). The etymological mode of inquiry which is so prominent in the Auraicept favoured multiple explanations to single phenomena (see 2.10.). Occasionally the results are less surprising than the steps leading to them. For instance, through lemma-based exegesis of the canonical Irish text, the commentary might well arrive at the teachings of Priscian.

Caide ruidhles 7 dileas 7 coitchind 7 indles in focail is guta? Ni ansa. Ruidhles di guth shet, uair etaidh guth a hænur. Diles di guth fuiit, uair nos-foidend fein. Coitchind di .i. guth fotha, uair is fotha hi isna foclaib. Indles di immorro guth fotha, uair ni fotha hi inti fein (Calder 1917: ll. 380-84). 80

This extract introduces the conception of the vowel as a fundamental unit of the voice through etymological analysis of the Irish word together with a discursive strategy which was developed by the scotti peregrini, notably Sedulius Scottus (for which see Law 1994: 105). An expression such as guth-ṡét ‘voice-path’ compares well with Priscian’s litera uel quasi legitera,

79 The two letters are named ébad and ór respectively, and according to Auraicept Muman (íar nAuraicept Muman), these were added to the signary before the other three additional characters (McManus 1991: 143; Calder 1917: ll. 1359ff.). 80 “What are peculiar, proper, common and improper of the word ‘vowel’? Not hard. Peculiar to it: voice-path, since it finds voice by itself. Proper to it: they express a voice, for it expresses itself. Common to it: i.e. voice-foundation, for it is a foundation in the words. Improper to it, however, is voice-foundation, when it is not a foundation in itself,” (Calder 1917: 29).

58 quod legendi iter praebeat (GL vol. II: 6). The commentary engages both the earlier strata of the text and the complementary Latin tradition in its exposition of grammatical doctrine. Latinate doctrine surfaces as vernacular (‘native’) terminology.

The ThGT introduces the vowel-letters in a passage which in several ways is a departure from Priscian. The basic 16-letter fuþark has five vowels:

Í norrœnu stafrófi eru fimm hljóðstafir svá kallaðir: úr u, óss o, íss i, ár a, ýr =, ok er íss stundum settr fyrir ‘e’, þá er hann er stunginn, svá sem aleph eða ioth setjaz fyrir tveim raddarstǫfum í ebresku máli. u er því fyrst sett, at þat hljóðar í framanverðum vǫrrum. o er þar næst; hann hljóðar í munni. i stendr þar næst, þvíat hann hljóðar í ofanverðum barka, en í neðanverðum barka, ef hann er punktaðr, ok hljóðar þá sem ‘e’. Þar næst er a skipat, þvíat þat hljóðar í brjósti. = er tekit af ebreskum stǫfum. En látínumenn skipuðu stǫfum gagnstaðliga þessu, sem hér er greint. Þeir settu ‘a’ fyrst, þvíat þat hljóðar næst hinu nezta tóli raddarinnar, er vér kǫllum lungu, ok þat má fyrst skilja í bernsligri raust. Enn u er því fremst skipat, at þat er fremst ok næst sjálfu efni raddarinnar er vér hyggjum at loptit megi kalla, ok hafa því hvárirtveggju meistarar vel ok náttúrulega skipat stǫfunum í sínu máli. Raddarstafir þessir hefjaz allir af sínu hljóði ok leggjaz í samhljóðendr ᚱ reið ok S sól (Wills 2001: 84 [§3.6-15], slightly adapted).81

81 “In the Norse alphabet there are five vowels called: úr ᚢ, óss ᚮ, íss ᛁ, ár ᛆ, ýr ᛨ, and íss is sometimes used for ‘e’, when it is ‘dotted’, just as aleph or ioth are used for two vowels in the Hebrew language. ᚢ is placed first, because it is sounded on the front of the lips. ᚮ is next; it is sounded in the mouth. ᛁ stands next because it is sounded in the upper windpipe, but in the lower windpipe if it is dotted, and it then sounds like ‘e’. ᛆ is placed next, because it is sounded in the chest. ᛨ is taken from Hebrew letters. But Latin speakers arranged the letters in the opposite order to that expounded here. They placed ‘a’ first because it is sounded closest to the lowest organ of speech, which we call the lungs, and it can be distinguished first in a child’s voice. But ᚢ is placed in front because it is the most forward and the closest to the substance of voice itself, which we think may be called air; and thus both (Latin and Norse) scholars have ordered the letters in their language well and in a natural way. These [names of] vowels all begin with their own sound and end in the consonants ᚱ (reið) and S (sól).” (ibid., slightly adapted). The quotation reads differently in W – where ‘h’ is said to be taken from the Hebrew letters (B. M. Ólsen 1884: 5). This does not fit with the discussion, which deals with vowels, and is therefore secondary.

They could be arranged as follows (see also Raschellà 1994):

fígura nafn Equivalent

úr u (also v) u o óss o i / e íss i /e/æ (also j) a ár a = ýr y

Passages such as this were not commonplace in medieval grammatical literature.82 Nevertheless it has parallels of varying correspondence in several anonymous Latin tracts on grammar. One of these, which opens with the words Magnus quae vox, is contained in manuscript Clm 14737 in the Bayerische Staatsbibliothek in Munich (Law 2003: 119, 292) and was probably composed in the 10th century (op. cit.: 148). The relevant passage is found at ff. 159v-160r. This is concerned with the proper arrangement of the vowels in the alphabet according to their place of articulation, and provides a list corresponding to that which Óláfr implies above (gagnstaðliga þessu, sem hér er greint) regarding the arrangement of the Latin letters. A closer parallel is provided by Thurot (1868: 135), as noted by B. M. Ólsen, where the addition regarding ‘childish speech’ is found. ‘A’ is the first letter of the alphabet because nature (and thus, ultimately God) made it so; it is the first discernible sound of speech. The argument is internally coherent until Óláfr reaches the letter = ýr. This is said to be taken from among the Hebrew letters. He avoids self-contradiction by placing it outside his system - though unable to properly fit it into his articulatory scheme, Óláfr is able to legitimize the letter’s inclusion by reference to its origin. Wills (2001) lists the following as a parallel or source in Priscian: utimur etiam y Graecorum causa nominum (GL vol. II: 9), “we also use ‘y’ because of Greek nouns”. This could certainly have functioned as a ‘source’ by analogy: Priscian places the equivalent letter at the very periphery of his scheme and attributes it to a foreign tradition. Óláfr further mentions that the runic letter i (íss) might denote ‹æ› when it is punctuated. This practice is validated by reference to a corresponding Hebrew practice of

82 « La physiologie des sons articulès, le mécanisme de leur production, n’était pas du domaine propre de la grammaire, » (Thurot 1868: 135).

60 disambiguation by means of vowel symbols. 83 He does not say anything about the origin of the practice. Óláfr’s vocalic inventory corresponds closely to contemporary Danish (Finnur Jónsson 1927: 26; Raschellá 2016: 187) and Norwegian short-twig runes (Spurkland 2001: 166; Raschellà 2016: 187), with the exception of the last vowel-letter listed above (5.1.). The set of five vowel letters was the basic one in the graphemic system, which could be augmented with a set of additional characters derivated from these for purposes of disambiguation. These letters were nevertheless subordinated and their use was not quite consistent. In the theoretical framework of the ThGT they are mainly treated as diphthongs. This is not fully in consonance with actual practice (see 2.9. on the pangram), in which they denoted the umlauted vowels (1.3.).

Diphthongs The Auraicept puts strong emphasis on the diphthongs (or digraphs; the distinction is not clear-cut as the same word is used for both). These have been briefly introduced as a tentative part of the basic system of vowel letters. In the commentary to the Auraicept, there are three reasons why the Ogam has dipththongs or forḟeda:

Air it trega ar a tugad forfeda eter isin aibgitir in ogaim .i. do fregra do defoghur amal

adberar isna breathaibh nemedh .i. genmota forfeda a fail defoghur na nguta 7 dono do sainigedh foghur forsna fedhaibh, air is buga fogair bis isna forfedhaibh, ut est, neam ᚕ and: næmh ᚙ and: nem ᚓ and (Calder 1917: ll. 1296-1301). 84

Priscian is not the source of the above excerpts, which to some extent abandon Latinate doctrine in favor of a peculiarly native phenomenon, namely the vowel-based denotation of

83 These symbols were deviced by the Masoretes, active in Palestine during the 8th-10th centuries. The vowels in Hebrew were not previously expressed in writing. 84 “For there are three reasons why forḟeda were introduced at all into the alphabet of the Ogam: to correspond to a diphthong (deḟogur) as it is said in the Bretha Neimed, i.e. except that forḟeda have the two sounds (deḟogur) of the vowels, and then to differentiate sounds upon the forḟeda, for it is a softness of sound that exists in the forḟeda, as neam (‘heaven’): EA is there, næm (‘saint’) Æ is there, nem (‘poison’): E is there.” (cf. Calder 1917: 101). The examples show how the additional characters are used to denote consonantal quality (palatal vs. neutral /μ/), though this is not stated explicitly in the surrounding text (see Poppe 1996a: 67n18). The difference between nem ‘heaven’ and nem/neim ‘poison’ would be the palatal quality of -m.

61 distinction of phonemes through consonantal quality. This was often, though by no means consistently, employed in manuscript orthography (see McCone 1996: 26, 32f. for the early beginnings of this usage). The phenomenon is reflected in the ‘cosmetic values’ of the forḟeda, which are therefore dependent on non-ogamic practice, but not on the theory of diphthongs as laid out by Priscian and which is subtly accommodated in the ThGT.

According to the ThGT, there are three reasons why the Latin language has diphthongs:

Á látínu er diphthongus fyrir þrjár sakir fundinn: fyrir hljóðfegrð ok sundrgrein ok samansetning. Enn í norrœnu fyrir tvennar sakir: fyrir grein ok hljóðfegrð (normalized from B.M. Ólsen 1884: 49). 85

Although there are fewer causes to account for the existence of diphthongs in Icelandic, there are more diphthongs in it than in Latin. Óláfr does not miss his chance to point this out (B. M. Ólsen 1884: 47). His argument is primarily based on phonology, but his depiction of the phenomenon crosses the border between the auditory and the written several times. The word used to denote ‘diphthong’ is either the Icelandic tvíhljóðr m. (hljóð n. renders sonus at §3.3 (B. M. Ólsen 1884: 40)) or Latin diphthongus. As mentioned in the previous sub-chapter, there is some discrepancy between the diphthongs that Óláfr lists and the actual use of the composite runic letters or ligatures. The definition of diphthongs in the ThGT is remarkably unlike the one in Instt. gramm. (GL vol. II: 37) and takes runic terminology as its basis. Thus a diphthong is a ‘conglutination’ (samanlíming) of two vowels, which is represented by a set of five límingarstafir in the runic alphabet. This might be contrasted with Priscian’s diphthongi autem dicuntur, quod binos phthongos, hoc est voces, comprehendunt (ibid.) “They are called diphthongs because they comprise two sounds.” The emphasis on the graphic domain is characteristic of the general tenor of Mg. So also is the fact that Priscian is abandoned at this exact point. The context of the argument is within an analysis of the pangram which is quoted in the fourth chapter of Mg (see 2.9.). 86 The attribution of the pangram to King Valdemar would likely have conferred to it significant authority.

85 “A diphthong is found in Latin for three reasons: because of euphony, distinction (sundrgrein) and composition. But in Norse for two reasons: because of distinction (grein) and euphony.”

86 See Kristján Árnason 2016 (p. 208) for a discussion of the compounded letters in relation to poetics.

2.7. Consonants In traditional Latinate doctrine, the consonants may not be pronounced by themselves, but are dependent on the vowels quae per se voces perficiunt (GL vol. II: 9, ll. 5-9). It is important to note that these two categories are imposed as a result of analysis of the litterae in the context of individual pronunciation (2.1., 2.3.). Their differing functions in words are recognized, as for instance consonantal ‹i› and ‹u›, but has no bearing on the system as it is initially represented and which is authorized by tradition. The Latin consonants are further divided into mutes and semivowels (2.3.).

The Irish word for ‘consonant’ is taebomna (taeb ‘side’ + omna ‘bole’, ‘tree’) and might refer to the visual make-up of each consonantal character, which “in the Ogham alphabet, is a branch only on the side of the stem, or placed obliquely across it,” (eDIL: taebomna). There is considerable semantic overlap between omna (eDIL: ‘the bole of a tree or sometimes a tree’) and fid. Thus the following gloss might be as close to the origin of the term as any other: .i. do thaebaib na fidh n-aireghedha biit (Calder 1917: l. 413), “i.e. they are at the sides of the vowels”, harmonizing the compound with Latinate thought (represented by con-sonans). Also taebúaim (taeb + fúaimm ‘sound’, cf. translation in Calder 1917) appearing in one attempt at etymology (l. 412) supports this explanation.87 Both explanations are concurrent in the commentary and the conceptualization of the consonants along these lines provides instructive example of how native learning or native imagery is used alongside Latinate theory and forthright borrowing from it to create a synthesis.

The consonants of the ThGT are named samhljóðendr (sg. samhljóðandi).88 The same term, a calque on the Latin word, is used in the FGT. Óláfr recognizes that the Icelandic language allows for a greater number of consonants within a single syllable than does Latin and provides two examples consisting of eight or nine letters (not phonemes) (B. M. Ólsen 1884: 50-51). He then goes over to discuss consonantal clusters and the system of hendingar (syllables participating in internal rhyme) in skaldic poetry.

87 Read as taeb-úaim ‘side-seam’ (úaimm is also used for ‘alliteration’), following Calder, it would support the interpretation based on the visual appearance of the Ogam consonants. 88 The correspondence between the noun sonus and the related participle con-sonans is thus paralleled by Óláfr: hljóð n.: samhljóðandi.

Semivowels The semivowels give particularly revealing insight into the nature of the early Irish linguist’s engagement with Priscian. They provide us with a case where comparison between the Icelandic and the Irish tradition is particularly able to highlight the somewhat contingent development of discursive strategies. The main definition of the letters has already been given. The Instt. gramm. supplies this with the following, integrating the relative phonetic quality of the semivowels as a defining principle:

Semivocales sunt, ut plerisque Latinorum placuit, septem: f l m n r s x; [...] hae ergo [hoc est semivocales] quantum vincuntur a vocalibus, tantum superant mutas. [...]. ʻSemivocalesʼ autem sunt appelatae, quae plenam vocem non habent, ut ʻsemideos’ et ʻsemiviros’ appellamus, non qui dimidiam partem habent deorum vel virorum, sed qui pleni dii vel viri non sunt (GL vol. II: 9). 89

Priscian’s main definition of semivowels, which clearly differs from the one attempted by Dionysius Thrax,90 and which is based on the name of the relevant letters (VC),91 is rendered futile when applied on the Ogam due to the fact that all its characters are named after the acrostic principle (hence by necessity CV). The definition with most relevance here is that which is based on the name of the letters. The name of the semivowels have the structure VC (e.g. ef, el, em.), while the mutes have the structure CV (be, ce, de etc.). No letter in the Ogam classifies as a semivowel according to this definition:

89 “There are seven semivowels, according to most latinists: f, l, m, n, r, s, x; [...] These, then, (i.e. semivowels) outdo (superō) the mutes to the extent they are defeated (pass. of vincō) by the vowels. They are called ‘semivowels’ because they do not have full sound, as we call ‘half-gods’ and ‘half-men’ not those who have halves of gods or men, but those who are not fully gods or men.”.

90 Τούτων ἡμίφωνα μήν ἐστιν ὀκτώ· ζ ξ ψ λ μ ν ρ ς. ἡμίφωνα δὲ λέγεται, ὅτι παρ’ ὅσον ἧττον τῶν φωνηέντων εὔφωνα καϑέστηκεν ἔν τε τοῖς μυγμοῖς καὶ σιγμοῖς (Uhlig 1883: 11-12). “There are eight semivowels: ‘ζ’, ‘ξ’, ‘ψ’, ‘λ’, ‘μ’, ‘ν’, ‘ρ’, ‘ς’. These are called ‘semivowels’ because they are less well- sounding than the vowels and consist of moaning and hissing.” 91 Thus the letters which “begin in ʻe-ʼ and end in themselves” (ab e incipientes et in se terminantes) are classified as semivowels (see GL vol. II: 8).

2,1 Attaat di ernail forsind apgitir Laitindai .i. guttai 7 consona. 2,2 Attat dano di ernail

forsin beithi-luis-nin ind oguim .i. feda 7 táebomnae; 2,3 di ernail dano forsna consonaib

lasin Laitneoir .i. lethguttai 7 múttai. 2,4 Inna lethguttai cétamus, a tuistidi remib, 2,5 in mútti immurgu, a tuistidi ina ndegaid do ṡuidib. 2,6 Nirbu immaircide són lasin nGoídel

2,7 armbad aicned dóib díb línaib: a nguth remib 7 ina ndegaid, 2,8 air is ed robo

immaircide la suide 2,9 combad a toísech do·airesedar lais 7 a ndéidenach do chur úad, 2,10 conad mútti uili in beithe-luis-nin ind oguim acht feda nammá (Ahlqvist 1982: 48).92

The Ogam contains only mutes and vowels. The semivowels (lethguttaí) are rejected from the system. The auditory base which is an important part of Priscian’s outline is lost from sight, and so it presents no obstacle to the exploitation of this fact for rhetorical profit. There are two plausible explanations: incompetence or conscious omission. Incompetence with regard to Latin grammatical doctrine might safely be attributed to some of the later scribes of the Auraicept-tradition, as Latin quotations in the commentary are more often than not impaired and are accompanied by translations which must derive from earlier copies (see further Poppe 2002: 308), but hardly to the author(s) of the earlier strata, who must have been part of a monastic community (Rekdal 2012: 169) and unquestionably well acquainted with the language through grammatical texts. The second option, that of conscious omission, is naturally not the default choice, and poses the question ‘why?’. As pointed out in the introductory chapter, the Ogam’s four (original) groups of characters might have been inspired precisely by the Latin theoretical framework, and so a break in the Ogamic tradition and the resulting rejection of any such similarity must be seen as related events. The rejection of the semivowels might simply have been caused by a wish to provide the alphabet with another easily articulated difference from the default (i.e. Latin) system, for sake of theoretical independence.

92 “There are two categories in the Latin alphabet, namely vowels and consonants. 2,2 There are two categories in the Ogam B-L-N, namely vowels and consonants; 2,3 also two categories of the consonant according to the Latinist, namely semivowels and mutes. 2,4 The semivowels first, their supporting vowels come after. 2,5 It was not appropriate, according to the Irishman, 2,7 that to have a preceding and a following vowel should be a property of both of them, 2,8 for the appropriate thing according to this one was 2,9 that it might be the initial part that might remain with him, casting the final part away from him, 2,10 so that only mutes make up the Ogam B-L-N, except for the vowels only,” (Ahlqvist 1982: 48).

Of course the rejection of semivowels had the advantage of according with Priscian’s main definition (GL vol. II: 8). Furthermore, it eloquently borrowed its principle from the Instt. gramm., from the treatment of the name of ξῖ by anastrophe (ksi › iks) (see 2.3.). The acrophonic principle behind the names of the letters of the Ogam was a traditional feature of the alphabet. The commentary to the above quotation (Calder 1917: 497ff.) disregards this fact and implies that the names of the semivowels were subjected to anastrophy and subsequently represented mutes. According to the commentary, this was done in order to follow the example set by the Greek alphabet, which apparently had no semivowels. It must be remembered that Fénius Farsaid was a ‘Scythian Greek’ (2.10.), and that support by sacred language in general could enhance authority.93

Cid ar mad fearr leis-sium a mbith comtis muiti huili quam (.i. inas) a mbeith leathguta 7

muti amal robatar icon Laitneoir? Ni ansa. Ar sechem Grec, ar ni filet lethgutai la suidibh 7 roba do Grecaib do Feinius; no dno is ar uaisli uird na nGrec ut dicitur: Omne uile

priusponitur, omne bonum postponitur .i. samaigthir gach ndereoil ar tus 7 gach sainemail co forbu (Calder 1917: 512-19).94

Despite the fact that the Irish language is praised for its wide-ranging vocabulary and coverage in terms of letters, even the lack of one of Priscian’s main groups might be used as an argument for the superiority of Irish over Latin. It is all about perspective. The Ogamist must have taken considerable pride in being able to accommodate a principle from Priscian to such a display of autonomy.

Having noted that the lack of semivowels was made into a core topos in the Irish tradition, it is thought-provoking that nothing of the sort is found in the Icelandic material, given that the runic names are also acrophonic. Instead Óláfr has noticed that Priscian holds that the ‘sense’ (merking) of the letter may not be (fully) determined by the make-up of its name. The

93 Now of course the Greeks had semivowels (ἡμίφωνα): ζ z, ξ ks ψ ps, λ l, μ m, ν n, ρ r, and ς s, but, as has been mentioned, these were not defined on the basis of their names, which are acrophonic.

94 “Why would he prefer them all to be mutes [in the Ogam] rather than being semivowels and mutes, as they were with the Latinist? Not difficult. In order to follow the Greeks, for they had no semivowels, and Fénius was a Greek - or indeed it is because of the nobleness of the order of the Greeks, as it is said: [...], i.e. every bad thing is placed first, every good thing is placed to conclude”.

66 acrophonic principle is therefore dismissed in his scheme, which closely corresponds to that laid out in the Instt. gramm. (supra):

Samhljóðendr eru .xi. í rúnamáli, .v. þeir er nálægir eru raddarstǫfum, eru kallaðir half- raddarstafir af úfróðum mǫnnum, þvíat þeir hafa meiri líking raddarstafa ok merkiligri hljóð enn aðrir samhljóðendr. Þat er r, n, S, M, l. f er af sumum mǫnnum með þessum stǫfum taliðr í látínustafrófi, þvíat hann hefr sitt hljóð af raddarstaf sem aðrir þesskyns samhljóðendr. Enn Priscianus segir eigi mega ráða stafsins merking, hvárt hans nafn hefsk af raddarstaf eða eiginligu hljóði, sem marka má í þessu stafrófi ok mǫrgum ǫðrum, er náliga hefjask allir stafir af sínu hljóði, bæði raddarstafir ok samhljóðendr ok halda þó fullkomliga sínum merkingum (normalized from B. M. Ólsen 1884: 43-44). 95

Óláfr notes that the name of the letter does not influence its denotation. In support of Priscian’s statement he mentions the fact that ‘in this alphabet’ (í þessu stafrófi), i.e. the runic alphabet, allmost all letters begin with the sound they denote. It is interesting to note that Óláfr here adheres closely to the Latin text. So closely, in fact, that the nominis prolationes which form the basis of his analysis (consider e.g. þvíat hann hefr sitt hljóð af raddarstaf sem aðrir þesskyns samhljóðendr) are the Latin ones rather than the acrostic runic names.96 This in marked contrast to the Auraicept na nÉces. Óláfr applies observations made on Latin and the definitions obtained from them on Icelandic.97 When he makes no note of this, as is typically the case, it is important to remember the programmatical stance announced in the beginning of the second part of his treatise (B. M. Ólsen 1884: 60), on the close affinity or identity of the

95 “There are eleven consonants in runic speech (rúnamál). Those five which are closest to the vowels are called semivowels (half-raddarstafir) by unlearned men, because they resemble the vowels more, and have more marked sound (hljóð) than other consonants. These are r, n, s, m, l. f is by some men counted among these letters in the Latin alphabet, because it begins its pronunciation (hljóð ʻnominis prolatio’) from a vowel, as other consonants of that kind. But Priscian says that the sense (merking) of a letter may not be determined by whether its name begins with a vowel or its proper sound, as may be observed in this alphabet and many others, that almost all letters begin with their own sound, both vowels and consonants, and still they fully keep their sense (merking ʻpotestasʼ)”.

96 I.e. Reið, nauð, sól, maðr, lǫgr and fé (for an overview of the runic names consult Spurkland 2001: 21). In the treatment of the runic vowels the opposite is the case. 97 This is not always carried out with success – consider the treatment of ‘h’ in M, the second part of his treatise (B. M. Ólsen 1884: 68).

67 traditions (for which see 2.10.). The runic letters were eminently unsuited to illustrate Priscian’s point. This is downplayed by Óláfr with a twist of logic. Priscian presents seven semivowels in the above quotation: ʻf’, ʻl’, ʻm’, ʻn’, ʻr’, ʻs’ and ʻx’. The ThGT, on the other hand, does not deal with ʻx’. His treatment of ʻf’ builds on Priscian. He says that ʻsome men’ place it together with the semivowels, as it is pronounced with a preceding vowel, i.e. ef (VC). Then he observes, as does Priscian, that the latter’s systematization of the letters of the Latin alphabet has no bearing on phonology: The letters keep their value regardless of the pronunciation of their names.

Mutes As an introduction to the treatment of mute letters in our texts, a few extracts concerning terminology will be examined, all ultimately derived from Dionysius Thrax. The Greek terms for ‘semivowel’ and ‘mute’ are both derivatives of the term for ‘vowel’, arrived at by the use of the prefixed adjective ἡμι- and the pejorative prefix ἀ- respectively: [...] ἄφωνα δὲ λέγεται, ὅτι μᾶλλον τῶν ἄλλων ἐστὶν κακόφωνα, ὥσπερ ἄφωνον λέγομεν τὸν τραγῳδὸν τὸν κακόφωνον (Uhlig 1883: 12). 98 The relationship between the terms was only partly carried over into the Latin terminology (i.e. semi-vocales); muta obviously translates ἄφωνα, but is not derived from vocalis. The illustration offered by Dionysius Thrax is therefore not quite as clarifying when used with the Latin terms and was abandoned. It was, however, replaced with an example which is somewhat off-the-point, but which in somewhat altered form found its way into the vernacular grammars. Consider the following extract from Instt. gramm.:

Reliquae sunt mutae, ut quibusdam videtur, numero novem: b c d g h k p q t. et sunt qui non bene hoc nomen putant eas accepisse, cum hae quoque pars sint vocis. qui nesciunt, quod ad comparationem bene sonantium ita sint nominatae, velut ‘informis’ dicitur mulier, non quae caret forma, sed quae male est formata, et ‘frigidum’ dicimus eum, qui non penitus expers est caloris, sed qui minimo hoc utitur: sic igitur etiam ‘mutas’, non quae omnino voce carent, sed quae exiguam partem vocis habent (GL vol. II: 9).99

98 “They are called ‘mutes’ because they are more bad-sounding than the others, just like we say that a tragedian with a bad voice is mute.”

99 “The rest are mutes, according to some they are nine in number: ‘b’, ‘c’, ‘d’, ‘g’, ‘h’, ‘k’, ‘p’, ‘q’, ‘t’, and there are some who do not think that this name fits them well, as also these are a part of speech. They do not understand that they are thus named in comparison with the well-sounding [letters], as a women is called ‘unshapely’, not because she lacks a shape, but because she is ill-shaped, and we call him ‘cold’,

The gist of this explanation is essentially the same as in that which is offered by Dionysius Thrax, although Priscian’s examples are slightly less suitable – in-formis, for instance, resembles the Greek composition quite clearly, but is also irrelevant as an illustration of the Latin term muta.100 The analogy is furthermore flawed, as the rationale underlying the division into three groups of letters is surely the degree to which they have ‘sound’ or ‘voice’. Priscian’s example was adopted in the Auraicept, where the pejorative prefix do- is used in order to translate the Latin in-.

[...] no onni as mutus .i. amlabar 7 ni airsinni batis amlabar di raith, air itat a fhoghuir intib cidh diat becca, ut Priscianus dixit: Informis mulier dicitur non quia caret forma sed male

formata est .i. atberar in bannscal dodheilb 7 ni iarsinni seachmallas o dheilbh acht midhealb fuirri nama. Is amlaidh sin iarum na muiti nidat nemfograigh acht is terc fogur intibh tantum. Unde mute .i. mifhotha dicuntur ut Donatus dixit: Mute sunt que per se

nec proferuntur et per se sillabam non faciunt .i. atat na muiti 7 it e na denat in sillaib

101 treothu fen 7 noco turcbaiter treothu 7rl. (Calder 1917: 469-78).

The Irish composition mí-ḟotha is ‘excavated’ from the Latin term muta. The construction guth-ḟotha is found elsewhere in the commentary and is derived from the Irish term gutta ‘vowel’. These two explanatory formations are of interest not only because they both avail

not because he wholly lacks warmth, but because he uses very little of it. So also [we call them] ‘mutes’, not because they lack all voice, but because they have a small share of voice.”

100 Cf. Isidore of Seville: Vocales autem et semivocales et mutas a veteribus sonas et semisonas et insonas dictas (Canale 2014, vol. I: 70). “Vocali, semivocali e mute furono chiamate dagli antichi sonore, semisonore ed insonore,” (op. cit.: 71.). The term used by ‘the ancients’ closely resemble the Greek terms. 101 “[...] or from the word mutus, i.e. ‘speechless’, and not because they would be speechless altogether, for their sounds are in them even when they are small, ut Priscianus dixit: Informis dicitur mulier non quae caret forma sed quae male est formata, i.e. a woman is called unshapely not because she is devoid of shape, but only because she has an ill shape. Thus, therefore, the mutes are not soundless but a scanty sound is in them tantum. Whence they are called mutae, i.e., foundationless, ut Donatus dixit: Mutae sunt quae nec per se proferuntur nec per se syllabam faciunt, i.e., the mutes are these letters which do not make a syllable by themselves, and are not pronounced by themselves, etc.” (Calder 1917: 37). Note that Calder in his translation has replaced the corrupted Latin of the Irish mss. with Priscian’s original text.

69 themselves of the word fotha ‘fundament’, which, from the point of view of the medieval etymologist, has the convenient anlaut f- (in composition /f/- › Ø- ), but importantly also because the relationship between the terminological units, both with a basis in sound, is provided afresh, and, it would seem, with a clear basis in a part of the teaching of Priscian which is patently ignored in the older material. It is certainly curious that the reception of the older text in this way harmonizes it with doctrine from which it originally had distanced itself. Against this background we would be justified in asking whether the display of autonomy in the Auraicept might be classified according to chronology, but the commentary does also offer significant departures from traditional grammatical doctrine.

Óláfr replaces the example with an example of his own.

Í þessu stafrófi eru ok fimm stafir, er vér kǫllum dumba stafi. Þat eru: F Q K t B. Ok eru stafir þessir eigi því kallaðir, at þeir hafi ekki hljóð, heldr því at þeir hafa lítit hljóð hjá raddarstǫfum, í þá líking sem sá maðr er lítils kallaðr verðr eða enskis af góðri ætt, er lítt er mannaðr hjá sínum gǫfgum frændum (normalized after B. M. Ólsen 1884: 44-45). 102

Óláfr does not include ‘h’ among his mutes. This is the only one of Priscian’s nine letters which is not represented in his translation of this passage, allowing for the fact that the runes might be used to denote voiced/unvoiced pairs,103 and that the distinction in usage between the Latin graphemes ‹c›, ‹k› and ‹q› was irrelevant for his system. Included among the Icelandic mutes, however, we find F f, which is not included by Priscian. Óláfr includes this letter among his semivowels too – hence the inconsistency between his statement about the number of consonants in the runic alphabet (samhljóðendr eru .xi. í rúnamáli) and the cited forms is rendered less obvious. The confusion about the status of ‘f’ made it a suitable replacement for ‘h’ – it could be listed both as a mute and as a semivowel without violating the explanation laid out in prose.

102 “In this alphabet there are also five letters which we call mute letters; these are: f, þ, k, t, b, and these letters are not called mute because they do not have sound; rather because they have little sound compared with vowels, in the same way that a man is said to be of little value, even if he is from a good family, if he is accomplished in little compared with his noble kinsmen.” (Wills 2001: 87 [§3.19]; runic letters in his trans. are here replaced with transliteration).

103 For instance t could render /t/ or /d/.

In the Irish tradition, the mutes have their ‘parent vowels’ (tuistidi) after them (i.e. when pronounced as part of the alphabet). The tuistidi are defined as in lucht ota a teasargain no a tinnscital .i. a nguthaidhe, “those whence is their deliverance or their origin, i.e., their vowels,” (Calder 1917: ll. 483-84):

Cid ara n-ebairt a tustidhe ’na ndiaidh maso thustidhi tinnscital, uair ni gnath in tindscedul fo dheoidh. Ni hedh eimh as ail dosum sund ar mad tustidi in tinnscedul etir, acht mad æ roscichestar inna menmain .i. dliged gotha fil i tosuch na leathguta do airisim leis fo

dheoidh 7 in dliged consonata fil intib fo deoidh do chur uad prius (Calder 1917: ll. 484- 491).104

2.8. De ordine litterarum De ordine literarum is the last section of the first book of the Instt. gramm.. This section deals with the relative position of the letters in the larger unit of the syllable. This, termed ordo, was arguably (according to Priscian) to be conceived of as related to the value (potestas) of the letters: quidam addunt etiam ordinem, sed pars est potestatis literarum (GL vol. II: 9). “Some also add ‘order’, but that is a part of the value of the letters”. As a logical consequence of this, Priscian deals with ‘order’ before he reaches the syllable (GL vol. II: 37ff.). It is related to the ‘force (potestas) of the elements’. The phenomena discussed are primarily phonological and related to shifting phonological environments. It should be pointed out here that Priscian does not deal with the order of the letters within the alphabet; indeed no word corresponding to ʻalphabet’ is found in his work. Reference to the alphabet is rather made with phrases as apud antiquissimos Latinorum (where we would perhaps have used ʻin the oldest Latin alphabet’), and similarly a Graecis. Priscian does not give a complete list of the 23 letters of the Latin alphabet, but when he lists them group-wise, this is invariably according to their traditional order. Otherwise he does not deal with the sequence of the letters within the alphabet.

104 “Why did he say the parent vowels are after them, if beginning be parents, since it is not usual that the beginning is last? That certainly is not his intention here, that parent vowels should be the beginning at all, but that science will be perceived in his mind, i.e., the law of voice which is at the beginning of the semivowels should remain with it to the last, and the consonantal law that is in them to the last should be uttered forth first,” (Calder 1917: 37).

Neither does the canonical portion of the Auraicept provide a complete list of the letters in their proper order, but the whole inventory of the Ogam is illustrated in the commentary, after the alphabets of the three sacred languages have been listed (Calder 1917: l. 1138). This is a conventional (for the MS record) depiction of the four families of letters together with the supplementary fifth family and a transliteration in sub- and superscript. This key to the Ogam alphabet (14th century) is reproduced in the appendices (5.1.). In the following discussion it should be kept in mind that it is possible that the actual shapes of the letters as they are presented in the manuscript witnesses may have undergone changes during the transmission of the texts. A distinction between the epigraphic record and the MS record is obviously paramount, but it is equally important to keep in mind that our manuscripts are copies of lost sources. This should be borne in mind, although the effect on the present line of argument will be minimal.105 In the following they will be reproduced as they appear in the editions. It is true that both treatises put some emphasis on the forms of the letters, but reconstruction based on their description would hardly be possible. The discussions on the letter-shapes do, however, seem to fit quite well with those that have been most faithfully transmitted.106 As an aid in illustrating the order of letters, Óláfr quotes a mnemonic piece or pangram (Raschellà: ʻfrase-modelloʼ (1994: 684)), which he attributes to king Valdimarr (i.e. Valdemar Sejr, King of Denmark 1202-1241). This also illustrates the shapes of the runes: Þessa stafi ok þeirra merkingar compileraði minn herra Valdimarr konungr með skjótu orðtæki á þessa lund [...] (B. M. Ólsen: 45-46; Wills 2001: 88 [§4.5]).107 This piece contains all the 16 letters ‘in Danish tongue’ (í danskri tungu) together with a few variant letters, and reads as follows (A):

105 The case would have been different with a text like the FGT, which was written as an attempt to reform orthography, and the influence of which would have been easier to assess had the original document survived.

106 There is nevertheless no way to be sure about whether Óláfr used the symbol Y or = - W leaves the illustration out when it should first occur. When it is cited next its shape might suggest an original with Y. A has =, which is the usual shape in late Icelandic runic inscription (see B. M. Ólsen 1883: 82). 107 “These letters and their significations my lord King Valdemar compiled with a short word-formula in this manner: ...” (Wills 2001: 89 [§4.5]).

Transcription: SPrængD mannc hø

Transliteration: Sprængd mannz høk flẏði toui boll (A: 2v13-14; cf. B. M. Ólsen 1884 45- 46; cf. Raschellà 1994: 684;).

(Tentative) translation: “Bold Tove chased away the exhausted man’s hawk,” (Raschellà 2016: 187; cf. Wills 2016: 120).

The phrase is eminently hard to translate. 109 The function of the phrase is, however, much clearer than its precise meaning (see Raschellà 1984: 684; Wills 2001: 128 [§3.2.2]). The introduction of the mnemonic phrase marks a shift in Óláfr’s approach, but even though it employs a mnemonic piece as the basis for discussion, it is accommodated within Priscian’s framework. The methodological shift between the third and fourth chapters of Mg is therefore not as profound as suggested by Wills (2001: 144; 2016: 119). In fact, Ólafr introduces the relevant section by quoting Priscian on ordo: Sumir meistarar kalla skipan hitt fjórða tilfelli stafs, en þat kallar Priscianus einn part þann er mætti stafsins heyrir (normalized after B. M. Ólsen 1884: 45). “Some scholars call order the fourth characteristic of the letter, but Priscian calls it only a part, which belongs to the characteristic of value of the letter,” (Wills 2001: 89 [§4.4]). In addition to providing the forms of all the letters together with their value, the pangram gives Óláfr the raw-material he needs in order to discuss the concept of ‘conglutination of letters’ (samanlíming), which was frequent also in epigraphy. The phenomenon of connecting adjacent letters into ligatures is of course also paralleled in manuscript calligraphy. It therefore fits quite nicely into Óláfr’s program: Priscian’s discussion is accommodated to the native alphabet, which is capable of effective representation of sounds - just as the conventional Latin script.110

108 The third letter, r, is not represented in this ms. A facsimile is to be found in Stafrænt handritasafn at http://www.am.hi.is:8087 (visited April 28th 2016). 109 See comprehensive discussion in Raschellà 2016 (171ff.). A new attempt at interpretation was made in a paper delivered by Katrín Axelsdóttir at the 22nd Germanic Linguistics Annual Conference (2016), at Háskóli Íslands, with the title “All the King’s Runes”. An abstract is available at http://glac2016.hi.is/wp-content/uploads/2015/09/GLAC-22-2016-Abstracts-2016-05-13.pdf (last visited October 18th 2016). I was not present at the conference and have not heard the paper. 110 At Óláfr’s time the script is generally classified as pre-Gothic, developing towards Textualis (see Haugen 2004).

Hence the fourth chapter is an original attempt at an application of a calligraphic feature (the adjoining of runic letters had originally no basis in phonology) to the predominantly aural discussion in Priscian (but see 2.1.). The additional runic letters (see 1.3.; 5.1.) enabled the fusion of these two spheres – the diphthongs, for instance, were represented by ligatures in the increased runic alphabet (5.1.). The ThGT gives a short treatment of the order of letters in diphthongs (B. M. Ólsen 1884: (47)-48). In dealing with the diphthongs, Óláfr invokes a hierarchical principle: The vowels are arranged so that the one which is ʻcloser to the innermost organs of speech’ (i.e. lungs and windpipe, B. M. Ólsen 1884: 35) is placed first. When providing examples of this, he is faithful to his earlier explanation of the place of articulation of the vowels (located at op. cit.: 42): [...] er ok hœgra þat hljóð fyrri at setja í samanlímingu er fyrri skapaz enn hitt eptir er síðar formeraz (op. cit.: 49). “[...] it is also easier to place the sound which is created first at the start of the conglutination and that one following, which is formed later,” (Wills 2001: 89 [§4.15]). The place of articulation is interconnected with precedence in time.

2.9. The Opposition between the Native Grammarian and the Latinist The close reading of passages revealing the theoretical concerns of the grammarians has undoubtedly made the reader aware that the texts systematically adopt a distinction between their subject tradition and the object tradition with which this is compared. Direct comparison between features of the vernacular language and those of Latin is carried out both in the Auraicept and in the ThGT. Both contrast the native grammarians with the Roman grammarians (the Auraicept: ind laitneoir (sg.); ThGT: látínumenn (pl.)). In the Auraicept this is a discursive trait which is found throughout the commentary, with secure basis in the core text, where the pair laitneoir (‘Latinist’)/fili (‘poet, man of learning’) is set up (e.g. Calder 1917: l. 668) and the formula “the one’s x is the other’s y” (‘x’ lasin laitneoir is ‘y’ lasin filid) is frequently followed. This polemic, based on the painstaking comparison of grammatical phenomena, does only to a very limited extent deal with letters, but e.g. with vocabulary (op. cit.: p. 82; rhetorical climax by citing in Laitneóir féin ‘the Latinist himself’, i.e. Isidore, arguing for vernacular supremacy in relation to Latin, at ll. 1098- 1100) and comparison/inflection (p. 50). What is said about letters, though, is said in the same vein and is a statement where one of the foilchesta (ᚍ NG) and three of the forḟeda (ᚗ UI, ᚘ IO and ᚙ Æ) are exhibited in order to demonstrate the comprehensive inventory of letters in the vernacular alphabet as compared to that of Latin (ll. 1078ff.). McManus is probably right in

74 attributing the interest in these letters to uncertainty with respect to actual usage (McManus 1991: 137) and the consequent need of illumination. Perhaps the fact that they did not correspond to any single letters in MS usage (ibid.) made them expressly captivating to the Old/Middle Irish Ogamist who would have been first and foremost familiar with conventional script and therefore liable to view the Ogam letters against this background. The dichotomy is applied as part of a rhetorical programme. I would like to suggest that the Auraicept itself, operating on the interface between Irish and Latin, somewhat ironically challenges the relevance of such an opposition in Irish textual culture by that very act.111

Óláfr does not develop the comparison with the Latin language and its letters into a corresponding polemic. He is more diplomatic:

Enn látínumenn skipuðu stǫfum gagnstaðliga þessu sem hér er greint [...] ok hafa því hvárirtveggju meistarar vel ok náttúrulega skipat stǫfunum í sínu máli (normalized from B. M. Ólsen 1884: 43).112

Björn M. Ólsen emphasised the apparent opposition between Latin learned men and those versed in vernacular literacy (1884: 54). This resembles the opposition between in Goídel and in Laitneóir, which comes to the fore in the Auraicept. He further contends that the above passage is to be read in conjunction with a passage found in the prologue to the four grammatical treatises in W. This reads:

Skal ýðr sýna hinn fyrsta letrshátt svá ritinn eptir sextán stafa stafrófi í danskri tungu, eptir því sem Þóroddr rúnameistari ok Ari prestr hinn fróði hafa sett í móti látínumanna stafrófi, er meistari Priscianus hefir sett (B. M. Ólsen 1884: 154). 113

111 Further evidence against such a dichotomy is provided by the Continental manuscripts (Thes. Pal.) where Irish is found side by side with Latin; often written in the same hand. The majority of the 9412 glosses in Cod. Sang 904 (of which 3478 are in Old Irish) are, for instance, written by glossator A (consult http://www.e-codices.unifr.ch/en/csg/0904/3, last visited October 12th 2016). 112 “But the Latinists arranged the letters contrary to that which is explained here [...] and thus both masters/scholars have arranged the letters in their language well and naturally.”

113 “I will show you the first way of writing (letrsháttr), written thus after the sixteen letters’ alphabet in Danish tongue, after that which Þóroddr Rune-master and the learned Ari Priest have put against the alphabet of the Latins, which Priscian set up.”

The prologue to the four grammatical treatises is attributed to the redactor of W and was evidently composed to fit the compilation in this manuscript (see Males 2013: 42n1). While it remains possible that W could be a copy of a redaction more or less identical to it, this would date the prologue to approximately 1350, making it roughly one hundred years younger than the original ThGT. It is therefore not strictly relevant to our reading of the ThGT, 114 which pertains to the cultural context of its composition.

2.10. Excursus: Mythography and the Origin of the Vernacular Alphabet Unlike their enigmatic historical origin, sketched out in the introduction, the mythological or pseudo-historical accounts of the alphabets are of great concern to the present argument, and it is therefore worthwhile to direct attention towards them before reaching the conclusion. The following section will be the least balanced one, to the advantage of the Auraicept, which is so steeped in pseudo-historical information that overlooking this aspect would be tendentious. Its inner logic depends on historical fabrication. This is not the case in the ThGT. Though it too partakes in pseudo-history, this does only implicitly affect its conceptualization of the vernacular alphabet. Isidore of Seville assumed that racial, political and social identity was linguistic in origin (see Irvine 1994: 276). The same is assumed in the Auraicept na nÉces (on which Charles-Edwards 1998: 76). The origin of language was a very central theme to the medieval historian. It is, however, reasonable to assume that the origin of the vernacular alphabet was of far less concern, as the alphabets themselves, as literary tools, were of less importance. This indeed appears to be the case and is reflected in the following.

Myth and Etymology Medieval thought was preoccupied with the search for correlations between extralinguistic reality and linguistic facts. The mindset thus put on display was at a vast remove from that of the modern scientist: Each linguistic object was worthy of study not because it was part of a system, but because it could point towards higher things. Knowledge of the world could be advanced through the study of the niceties of language. Etymology, in modern linguistics concerned with form rather than meaning, by retracing sound-laws, was in the Middle Ages

114 Except for, of course, that one might have expected a certain amount of editorial alterations in the copy of the ThGT in W. As mentioned in 1.7., such alterations do not significantly alter the sense of the text.

76 almost heedless of form,115 but very much concerned with meaning, and was therefore regarded as an important epistemic tool. Its insouciant handling of form made etymology, ‘a game with no rules’ (Bergin 1938: 206), a ready tool in the hands of the interpreter or the stylist and useful both in the composition of Sachprosa and saga. For the typical analysis of a compound word, carried out by isolating and explaining its constituent elements, the relationship between these elements and the things denoted by them had to be one-to-one - a proposition that may be difficult to grasp, but meaning that the form of the word stood in some kind of relation to the concept (the logical problems here may account for the lack of explicit theoretical justification of the proposition throughout the Middle Ages). The origin of the Irish language, as presented by the Auraicept na nÉces, provided just such a correspondence. This made etymology a legitimate tool for the analysis of the core text itself, and it is indeed one of the principal modes by which this text is explained and elaborated upon in the commentary. It is remarkable that the Auraicept na nÉces, dependent as it is on etymological practice similar and presumably indepted to Isidore’s

115 Consider for instance a procedure often followed in the commentary to the Auraicept: 1) The word is split into separate components. These components need not correspond to our morphemes. Example: attaat .i. ata æ i n-ait (Calder 1917: l. 328), but also atai æ uait is given as interpretation. 2) As witnessed in the example under 1), the etymologist may freely add syllables to his analysis. His approach in this way resembles the present-day student who has to undo the effects of syncope in order to identify an Old Irish verbal form. Indeed this method may have been encouraged by the fact that the commentators were often working with texts from an earlier stage of the language (i.e. Old Irish) and thus had to deal with linguistic change (see Kelly 2002: 234-35). Already the Middle Irish scholar was confronted which archaic and/or specialized technical vocabulary and obsolete forms (for instance verbal forms with infixed pronouns). But Isidore would also apply such an analysis: salsum ʻspicedʼ from sale asparsum ʻsprinkled with salt’ (Canale 2014: 634; see also Barney, Lewis, Beach and Berghoff 2006: 24). Derivations such as these might have served a mnemotechnic purpose (Barney, Lewis, Beach and Berghoff 2006: 24). The fourth book of the rhetorical treatise Rhetorica ad Herennium gives a description of paronomasia which is worth to quote in this context: Adnominatio est cum ad idem verbum et nomen acceditur commutatione vocum aut litterarum, ut ad res dissimiles similia verba adcommodentur. Ea multis et variis rationibus conficitur. “Paronomasia (āgnōminātiō) is the figure in which, by means of a modification of sound, or change of letters, a close resemblance to a given verb or noun is produced, so that similar words express dissimilar things. This is accomplished by many different methods,” (Caplan 1954: 300ff.). It is intriguing (and hardly coincidental) that the different methods here mentioned are quite similar to the ones applied in Irish etymological procedure.

Etymologiae, does not grant Latin this privilege. Irish is given epistemological priority.116 It is difficult to accept that the commentator(s) were satisfied by a mainly Latin analysis of their language when it is stated that even the relative sense of an Irish uttering (compare esp. Calder 1917: ll. 1097-101) might be unaccessible through Latin. The practice of vernacular etymological inquiry is a feature also found more generally in the Irish corpus of vernacular texts and probably most explicitly in the 9th century etymological glossary of Cormac Úa Cuilennáin – possibly the earliest in any Western vernacular (ed. Meyer 1913). It did not make such an impression in Old Icelandic literature. Perhaps therefore the authorization of such a practice is not prioritized in vernacular grammar. Other sources of authority are given main priority, the most notable being skaldic poetry, to which reference is constantly made. Etymological inquiry was clearly one of the principal interpretive tools available to the medieval Irish literary scholar.117 The medieval scholar recognized five different varieties of Irish; one of these was Isidorian-type etymology and was called bérla n-etarscartha. Exegesis carried out by etymology would, however, be rendered futile if the relationship between the language and the real world could not be guaranteed. To the medieval mind, the obvious place to seek such authorization would be in the origins of the language. The Bible provides only few remarks on language, but the importance of these to medieval linguistic thought was in reciprocal proportion to their number. Genesis places the origin of language in Eden and establishes a one-to-one relationship between each thing and the word which denotes it (see Law 2003: 107 on vox/verbum), here quoted in the Latin Vulgate (Weber-Gryson 1994):

116 Both Latin and Greek play their part in the etymological derivations found in Sanas Cormaic. 117 Consider, for instance, the beginning of the Auraicept in the BB (314 α 16): Incipit Auraicept na nEges .i. eraicept, ar er gach toiseach (Calder 1917: ll. 1-2). “Incipit Auraicept na nÉces, i.e. ‘eraicept’ for every beginning is ‘er’.” A close parallel to this etymological inquiry is found in the commentary to Donatus’ Ars maior by Murethach (Burnyeat 2007: 189n20). Intertextual relationship between the Auraicept and the etymological dictionary Sanas Cormaic is capable of proof; consider for instance: Tucait bindiusa, ut est, is i in ghabair 7 d’eoch ban is ainm .i. goar .i. solus isin Fenic[h]us (no isin Breatnais) co tuc in file be fris ar tucait mbinniusa (Calder 1917: ll. 633-35). This may be compared to the explanation found in Sanas Cormaic: 675. [...] .i. is goor gach solus i suidi, unde dicitur goar don eoch gil 7 rl. [...] Rotuill in fili gāidhelach .b. fris ar tucait mbindessa, ar rop āille leo gobar quam goor, unde gobar nominatur (Meyer 1913: 55).

19 formatis igitur Dominus Deus de humo cunctis animantibus terrae et universis volatilibus caeli adduxit ea ad Adam ut videret quid vocaret ea omne enim quod vocavit Adam animae viventis ipsum est nomen eius 20 appellavitque Adam nominibus suis cuncta animantia et universa volatilia caeli et omnes bestias terrae Adam vero non inveniebatur adiutor similis eius (Genesis 2: 19-20).118

Adam was the principal name-giver. Humanity was, however, deprived of linguistic, and thus racial, unity at Babel. The original Adamic language was irretrievably lost and replaced by a multitude of languages.119

7 venite igitur descendamus et confundamus ibi linguam eorum ut non audiat unusquisque vocem proximi sui 8 atque ita divisit eos Dominus ex illo loco in universas terras et cessaverunt aedificare civitatem 9 et idcirco vocatum est nomen eius Babel quia ibi confusum est labium universae terrae et inde dispersit eos Dominus super faciem cunctarum regionum (Genesis 11: 7-9).120

To Augustine this passage represented the tragedy which befell mankind as a result of arrogance. He contrasted it with the story of Pentecost in the Acts of the Apostles (2: 1-11) (Law 2003: 104-105) and the humility, spurred by faith, which at that occasion caused mutual understanding across language boundaries. The Irish solution to the Babelian problem benefited from ‘blind spots’ in the Bible and was anything but humble: A school was established at the Tower, by the eponymous Fénius Farsaid, who sent his disciples abroad to learn the multitude of languages. Fénius was himself

118 “19 And the Lord God having formed out of the ground all the beasts of the earth, and all the fowls of the air, brought them to Adam to see what he would call them: for whatsoever Adam called any living creature the same is its name. 20 And Adam called all the beasts by their names, and all the fowls of the air, and all the cattle of the field: but for Adam there was not found a helper like himself,” (Douay-Rheims Bible).

119 Commentators after Augustine identified the Adamic language with Hebrew (Law 2003: 104). This view seems to be predominating in Icelandic and Irish grammatical discourse. So also in Dante’s De vulgari eloquentia (Tavoni 2011: 1180). 120 “7 Come ye, therefore, let us go down, and there confound their tongue, that they may not understand one another’s speech. 8 And so the Lord scattered them from that place into all lands, and they ceased to build the city. 9 And therefore the name thereof was called Babel, because there the language of the whole earth was confounded: and from thence the Lord scattered them abroad upon the face of all countries,” (Douay-Rheims Bible).

79 a polyglot, learned in the principal or sacred languages, and out of the many languages resulting from the dispersion he selected the best parts and created a synthetic language, in bérla teibide ‘the Select language’: Irish.121 As a consequence Irish became far more ‘comprehensive’ than any other language. The best, ‘the widest’ and the most splendid was extracted from every language and put into Irish. This goes for vocabulary, but also for the phonemic inventory of the language (Calder 1917: ll. 1068-78). As a consequence the native writing system is equally comprehensive (op. cit.: ll. 1056-57). Instead of being ‘an unfortunate reminder of human wrongdoing at Babel’ (Law 2003: 190), Irish came as close to the pre- confusion stage and as close to the solution as possible. The confusion and dispersion of tongues which arose at Babel presented the Irish mythographer with an opportunity to provide his vernacular with an origin in the Biblical sphere. Names were devised for those things which were not previously represented in any language (Calder 1917: ll. 1061-62). In the commentary, the variety of building-blocks of the Tower of Babel are likened with the building-blocks of language. In the first of the Classical Modern Irish122 grammatical treatises it is said, alluding to Genesis 2: 19-20, that Fénius Farsaid gave names to ʻevery concrete thing which eye can see and hand can touchʼ:

[...] 7 as é Fē[i]nius Farsaidh féin tug ainm ar gac[h] ní subsdainnteac[h] do-chí súil 7 ghlacus lámh (Mac Cárthaigh 2014: ll. 131-132).123

121 The story is told in its whole, more or less, at several places in the commentary to the Auraicept (e.g. Calder 1917: pp. 2-4, pp. 8-22), though the details are not always consistent; parts of the story resonates throughout the entire work. It is nevertheless accompanied by other accounts: Ma do reir na sgrephtra naibe 7 ughdair in leighind is in aipgitir Ephraide is bunad don aipgitir Gregda 7 Laitianda 7 don beithe- luis-nion a n-ogaim cen co n-indisit ugdair in leighind, ut est aleb isinn Eabhra, alpa uadha-side isin Greic, a isin Laitin, ailm isin Gaodilc ogaim (Calder 1917: ll. 3421-3425). “But according to Holy Scripture and the scholastici [see eDIL: léigenn] it is the Hebrew alphabet that is the source of the Greek alphabet and the Latin [alphabet] and of the BLN of the Ogam, although the scholastici do not mention it, that is: ‘Aleph’ in Hebrew, that is ‘alpha’ in Greek, ‘a’ in Latin, ‘ailm’ in Irish of Ogam.” Neither the Irish nor the Icelandic tradition presents a uniform narrative of the origin of language and alphabet. 122 Classical Modern Irish was a highly regularized language, used (c. 1200-1600) in poetry and codified in several grammatical treatises and syntactical tracts.

This extract is picked from a passage which draws heavily on material derived from the Auraicept and its commentary (Mac Cárthaigh 2014: 188), although the remark itself must either be original or stem from other sources. What stands out here is a conscious wish to liken the situation after the fall of the tower at Babel to the original state of affairs, where the correspondence between each thing and the word which denoted it was absolute. The parallel between the origin of Irish and the origin of the Adamic language (arguably Hebrew) is obvious given the general Old Testament setting of the story. This passage explicitly states what is only implicitly said in the commentary to the Auraicept. Fénius Farsaid is given the same role as Adam once had. His role as a polyglot is repeatedly underscored also in the Lebor Gabála (LG; ed. Macalister 1939; see e.g. §149). The discussion on language above is of great concern to the vernacular alphabet. The language was its raison d’être and the prestige of the writing system would therefore be interconnected with it.

Retracing the Tetralogy: The Four Books of the Auraicept na nÉces Erich Poppe has described the Auraicept as being organized by two principles: ‘a horizontal sequence of books’ and ‘a vertical sequence of base - or canonical - text and gloss-commentary’ (2002: 296). He further remarks that the latter principle is found also in the Hiberno-Latin exegetical grammars. The vertical sequence has been dealt with in the introductory chapter (1.5.). At this point, having introduced the dispersion of languages at Babel and the Irish solution, we shall turn to the horizontal sequence of books. The Auraicept is traditionally (by text-internal evidence) divided into four books. The relation of these to the actual manuscript state of the treatise and its probable historical development has been briefly touched upon (1.5.). The first of the four books of the Auraicept to be presented in both recensions is the one which is attributed to Cenn Fáelad mac Ailella (d. 679) of the Cenél nEógain. According to tradition, however, this was the last of the tetralogy to be composed. This fact is expounded on in the following passage:

Caidi log 7 aimser 7 perso 7 tugait scribind in Uraicepta? Ni oenlog tra lasna cethri libro, amal atbert in fili: a n-as tuiseach, is ed is deghenach, a n-as dedhenach, is ed as toisseach .i.

123 “And it was Fénius Farsaid himself who gave name to every concrete thing which eye can see and hand can touch.”

a n-as toisseach iar n-urd lebhurda, is ed as dedhenacho arricht .i. lebor Cindfaeladh mic Oilella (Calder 1917: ll. 63-67).124

This passage receives some support in the following statement from the YBL-version: Cendfaolad mac Ailella doathnuaighiuster i nDoire Lurain maille re hurmor na sgreaphtra (Calder 1917: ll. 2638-40).125 Thus internal evidence suggests that his death would provide the terminus ante quem for the whole Auraicept na nÉces. This is, however, at odds with Middle Irish language features, which occur passim, both in commentary and in the language which is subject to analysis (e.g. analytic forms of the verb) (Calder 1917: 22). It is also in conflict with the format of the extant versions of the work, as noted in the introduction. Cenn Fáelad was noted for his breadth of learning and strong memory. His literary activity was, for instance, noted by Ruaidhrí Ó Flaithbheartaigh in his Ogygia seu rerum hibernicarum chronologia (1685) more than a thousand years after the death of the author.126 Abigail Burnyeat discusses at some length (2007) the role of Cenn Fáelad in staging the fusion of Latinate and vernacular learning.127 At this point medieval tradition is very much in line with the argument of the present thesis: Bretha Étgid presents an origin myth for the three schools of léigend, fénechas and filidecht. This prologue resembles the introduction to the ʻBook of Cenn Fáelad’ rather closely, and, as pointed out by Burnyeat (2007: 214-15), this might well indicate that the passages are interrelated in one way or the other. The prologue to Bretha Étgid is the more comprehensive in that it gives a fuller account of the nature of the learning which Cenn Fáelad assimilated while in his sick-bed. In his prologue, Cenn Fáelad refers to his predecessors as augtair na nGoídel (see below). The second book is attributed to the poet Ferchertne, who flourished at the royal seat

124 “What are the locus, tempus, persona and causa scribendi of the Auraicept? The four books do not share a common locus, as the poet put it: ʻWhat is first, that is what is last, and what is last, that is what is first.’ That is, that which is first according to book order, that is what was invented last, i.e. the book of Cenn Fáelad mac Ailella.”

125 “Cenn Fáelad son of A. revised it in Doire Lurain together with most of Scripture.” Note that this passage attributes the composition of the Auraicept to Ferchertne and its revision the Cenn Fáelad, thus deviating from the four-book scheme. 126 Quem librum uraicęaċt na nęigios, .i., ʻPræcepta Poetarum inscriptumʼ, et Centena Carminum genera complexum Kenfoela filius Olilli, Donaldo Rege Hiberniæ multis abinde Sœculis apud Doire- lurain interpolavit (Ogygia, part. 3. c. 30: 217).

127 A more elaborate version of the story is found in the law-text Bretha Étgid (see Burnyeat 2007).

82 of Emain Macha during the reign of Conchobar mac Nessa (Calder 1917: ll. 735ff.). He was, on the authority of the Auraicept, coeval with Christ. The third book is attributed to Amairgin Glúngel and the period of the initial settlement of Ireland: The time of the sons of Mil (Calder 1917: ll. 1028ff.).128 The fourth and last (remember that ‘the last is the first’) book was written by Fénius Farsaid, Iar mac Nema and Goídel mac Etheoir during Exodus (Calder 1917: ll. 1102ff.). It has been noted (1.5.) that the œuvres of the presumptive original authors, in the event that the tetralogy has any text-critical basis whatsoever, must have been rather insignificant. The fourfold division of the Auraicept is therefore hardly meant to aid the reader in sifting through the material, not least due to the vagueness of any ordering principle (see Calder 1917: xxvi; Acken 2008: 73). It might be argued that the subdivision was first and foremost intended to allow greater authority to the text and to the endeavour it represented by projecting it into the past.129 In the later tradition we have evidence for a strong tendency to attribute seminal works to significant historical or pseudo-historical figures - in some contrast to the fact that we seldom have evidence for such ascriptions in older material or in the old texts themselves.130

What is presumably or at least possibly the oldest distinguishable layer of the text (since it is treated as such in the manuscripts) likewise begins with reference to authority and to the realm of Scripture:

1,1 As·berat trá augtair na nGoídel combad sí tucait airic in bérla Féne gním n-ingnad n- indligthech for·chaemnacair isin domun .i. cumtach in tuir Nemruaid (Ahlqvist 1982: 47).131

128 Amairgin Glúngel is represented as the fosterson of Caí Caínbrethach in one manuscript of the Senchas Már (Hancock 1865: viii). The first sentence which was ever passed on Irish soil was thus derivative of Mosaic law. 129 The implication this has for our interpretation of the Auraicept as a whole has been only been hinted at (Burnyeat 2007: 195-199). 130 This situation is paralleled in the Icelandic case. Consider for instance Snorri’s works (and indeed the ThGT). 131 “Now the authors of the Irish say that the cause for the invention of the language of the Féni [Irish] was a strange wonderful deed that took place in the world, i.e. the construction of Nimrod’s tower” (ibid.: 47).

This is presented as the beginning of the book of Cenn Fáelad. The tradition thus implies that the composition of the Auraicept na nÉces was an ongoing project already a short time after the dispersal of tongues at Babel, through the initial period of Irish settlements and then during the age of the Ulster cycle. The presentation of the material is in reverse order, taking the reader on a journey back to the time when the language first made its appearance in script:

Is e seo a thossach in libhair-sea iar Fenius 7 iar nIair mac Nema 7 iar nGædel mac Ethiuir.

Is iat sin a persainn; 7 is i a aimser .i. aimser tancatar mic Israel uili a hEigipht. I nDacia arricht ci asberatar [asberat E] alii co mbad i mmaigh Sennair. Tucait a scribind .i. a

thotlugud do scoil mhoir co Fenius 7 co Iar 7 go Goedel mac nEitheoir a thepe doibh inna

nUraicept iarna idnacul do Maissi 7 iar foglam do Chae Cainbreathach occa; conad iarsin

arriachta na aipgitri a n-æntabaill, amal asbeir: Cat iat aibgitri 7rl (Calder 1917: ll. 1102- 1111).132

This is the requital offered by the Book of Fénius Farsaid: Irish did not chance to appear on the Cross, but, in spite of that, it was arguably impressed on the stone tablet of Moses side by side with the three sacred languages.133 The above paragraphs states that the Auraicept was bestowed unto Moses. In the

YBL-redaction a genitive attribute follows the verbal noun: iar n-idnacul rechto do Mhaisi 7 íar foglaim do Chaí Chaínbrethach oca (slightly normalized from Calder 1917: ll. 4143-44) “after the bestowal of the (Mosaic) law and after he had learnt it to Caí Caínbrethach.” 134 This reading, more in keeping with the Old Testament, situates the origin of Irish vernacular law at Mount Sinai, and makes it derivative of the Mosaic law which is represented by the Decalogue (cf. Hancock 1865: viii; Meroney 1945: 20). The BB reading, though inferior from a text- critical point of view, provides the Auraicept na nÉces with this privilege. This is what was

132 “This is the beginning of this book according to Fénius and Íar mac Nema and Goídel mac Ethéoir. They are its persons, and this is its time, i.e. the time in which all the Sons of Israel came out of Egypt. It was invented in Dacia, although others say it was on the plain of Shinar. Its causa scribendi: that the great school requested of Fénius, Íar and Goídel mac Ether that they should fashion the Auraicept after its bestowal unto Moses, and after he had learnt it to Caí Caínbrethach. Hence it is after this the alphabets were invented on a single table, as he says: What are the alphabets ? etc.”

133 For the close interaction between the early Irishmen and the Israelites, the reader should consult the major pseudo-historical compilation Lebor Gabála Érenn (ed. R. A. S. Macalister). 134 In the earliest literature recht without the article is used to denote the Mosaic law especially (DIL: recht).

84 bestowed unto Moses and what he further taught Caí Caínbrethach. The history of origin as presented in the LG follows the Early Irish from the dispersal at the tower of Babel and to their settling in Ireland. By its reverse chronology the Auraicept takes the reader back there again; to the origins of language.

Ilium and Iceland The Asiatic origins of the Norse gods and, with them, important parts of Norse culture (handcrafts, poetry etc.), is well-known from Snorri’s depictions in Ynglinga saga, grounding the history of the Norwegian kings in the Classical world, and in the prologue to Snorra Edda (Faulkes 2005: 3-6), his manual of poetics, where especially the Norse language is identified as the ‘tongue of the Asians’ (tunga Ásíamanna) (op. cit.: 6). The emphasis on Odin’s different capabilities is stronger in Ynglinga saga than in the SnE-prologue. Among these skills (íþróttir) we find poetry, battle-tactics, shapeshifting and his role as law-giver. He was originally a chieftain from the capital of Asia, namely Ásgarðr (Troy); and thus Norse poetry and Norse laws, among with other features, were inherited from the Trojans. In the introduction to M, the second part of the ThGT, which deals with barbarisms and solecisms together with other rhetorical figures, Óláfr says the following about the third book of Donatus’ Ars maior:

Í þessi bók má gerla skilja, at ǫll er ein listin: skáldskapr sá, er rómverskir spekingar námu í Athenis borg á Grikklandi ok sneru síðan í látínumál, ok sá ljóðaháttr eða skáldskapr, er Óðinn ok aðrir Ásíamenn fluttu norðr hingat í norðrhálfu heimsins, ok kenndu mǫnnum á sína tungu þesskonar list, svá sem þeir hǫfðu skipat ok numit í sjálfu Ásíalandi, þar sem mest var frægð og rikdómr ok fróðleikr veralldarinnar (normalized from B. M. Ólsen 1884: 60).135

135 “In this book [Donatus, Ars Major 1] it may be clearly understood that everything is the one art: the poetry which Roman orators learnt in Athens in Greece and then turned into the Latin language; and the song-metre or poetry which Óðinn and other men of Asia brought north into the northern half of the world, and taught men this kind of art in their own language, just as they had arranged and learnt it is Asia itself, where beauty and power and knowledge were the greatest in all the world,” (Wills 2016: 120).

This passage has only implicit consequences for the perception of the runic alphabet. The focus is on language and on its relation to poetry (as opposed to etymology as in the Auraicept).136 The cultural background which is invoked in the ThGT is Classical and pre- Christian, rather than Biblical, as is the case in Irish tradition. In the context of the present argument the most important aspect of this passage is that it provides a general justification for the theoretical engagement with the vernacular (authority by antiquity), together with a specific justification for the application of Latinate doctrine on it: The vernacular tradition exists in close affinity with the Latinate tradition. This is a key to understanding Óláfr’s urge to downplay the differences between these also on the most minute level when he deals with the letters of the alphabet.

136 A paper on the rhetorical consequences resulting from the choice of origin myth in the Icelandic and Irish traditions, entitled “Foundation Myths and their Rhetorical Consequences in the Auraicept and the Icelandic Third Grammatical Treatise”, was given by Males at Tionól 2014, held at the Dublin Institute for Advanced Studies.

3. CONCLUSION The final part of the thesis is devoted to the first question asked in the preamble (1.1), the other two questions having been the subject of the preceding analysis. This concluding chapter will be broken down into three parts: the first is dedicated to ‘vernacular grammar’ as a genre, the next to the texts as educational tools, while the third is devoted to their role in the history of linguistic study in the medieval period.

3.1. Practical and Theoretical Utility: Grammar as Genre Although of synoptic character, Lambert’s 1987 article, already quoted on the Ogam, contains a few points that are in line with the argument of the present thesis.

Mais l’Auraicept est une grammaire profondément artificielle ; s'il est exagéré de dire, comme Bergin 1938, qu'elle ne contient que des spéculations futiles (26), il est sûr en tout cas que l’Auraicept ne pouvait être d'aucune utilité pratique, ni pour les poètes, ni pour les juristes. Son incompréhension pour les abstractions linguistiques est sans doute la première cause de ce demi-échec ; l’Auraicept était surtout une composition littéraire née du besoin d'avoir une Grammaire pour l'irlandais (Lambert 1987: 20).

Two things should be said concerning the limited practical value of the Auraicept. The first is that practical utility does not guarantee the survival of any given text in a manuscript culture. If it did, we should have great difficulties in explaining the paucity of actual school-texts relating to grammar both from medieval Ireland and Iceland (cf. Guðrún Nordal 2001). The plentitude of manuscript witnesses to the Auraicept does not attest to widespread use of the text in education in the Old/Middle Irish periods, though of course the complexity of the text from a text-critical point of view does testify to significant circulation in the latter of these periods. It should be noted that this view offers a slight depart from consensus, which I believe has exaggerated the weight of the mention of the Auraicept in the MVII (Thurneysen 1891: 32, 115; Ahlqvist 1982: 17). I would like to suggest the following development: The Auraicept in its early stage represents two main ambitions, viz. strengthening the status of the vernacular language and alphabet by means of polemic at a time when Latin was by far the dominating language in a literary context, and the introduction of the ‘grammatical genre’, which was the cornerstone of literary culture, but not represented in Irish at the time. This basis then became a locus for linguistic inquiry, through gloss and extended commentary, and was transformed into an increasingly prestigious work to be consulted on matters of language and poetics. Hence it

87 was used as a ‘propedeutic to the study of metrics’ (McManus 1991: 148) in the Middle Irish schools, but its origin is to be sought in learned discourse on language rather than in an introductory context.137 The abscence of syntactic archaisms in the Auraicept (Ahlqvist 1982: 36-37) might, with a view towards this development, be explained as a result of the text being a ‘very simple and rather basic piece of “Sachprosa” that would not differ materially from the spoken language of its day.’ (ibid.), but might equally well be attributed to the likely event that it, unlike the law- texts, had no basis in oral tradition. This argument might be extended to another text, which likewise shows archaic morphological features, but no syntactic archaisms, namely the Cambray homily (late 7th / early 8th century; MS Cambrai, Bibliothèque Municipale 679 (c. 763-780), fos. 37rb-38rb) – in point of fact the oldest extended Old Irish prose passage that has survived. Texts such as these were the products of a literary culture - and this should help explain why they lack the syntactic archaisms of the law-texts, while at the same time the Auraicept has been subjected to Middle Irish exegetical commentary in a manner strongly reminiscent of that which is associated with the law-texts. Two genres of disparate origins were treated the same way in the later literary culture; just as the oral tradition represented by the law-texts had inherited archaic features, the early written language itself aged over time.138 The practical value of the ThGT is perhaps equally difficult to assess. Óláfr’s role as a teacher at Stafaholt has been mentioned (1.7.) and there is nothing improbable in the opinion that the text was meant to be used in relation to this school. His learning, if not his position as subdeacon, would perhaps have entitled him to a great deal of influence on the methods of instruction and the topics to be covered. The singular character of the ThGT would demand such independence if it was to be used in education. The manuscript record testifies, as has been mentioned, to some circulation (see 1.7.); though less than is the case for the somewhat earlier poetics of Snorri, which has a wholly different approach. The influence of the latter is evidenced in later medieval material (see further Males 2014), while abundant reference to the ThGT was made in the post-medieval period (see Wills 2001: 18), not least due to its information on the runes. I believe that the present thesis supports the contention that the ThGT belongs to a period of ‘vernacular assertiveness and the establishment of a literary past’ (Males 2016: 264), as does the Auraicept.

137 See further Males 2016 for a somewhat corresponding development in Old Icelandic grammatical literature at large. 138 It is of course also possible that the canonical Auraicept is younger than Ahlqvist’s dating.

To this I would like to add that these two treatises establish not only a literary past, but ‘a past of letters’. The FoGT is imitative of the ThGT, but concerns itself with the rhetorical figures in verse, thus adding to M (the second part of the ThGT), which is outside the scope of the present thesis. Mg has no medieval offspring in the Icelandic tradition. It is conceivable that the runic chapters in Mg stand in the same relation to the rhetorical ones in M as does the mythological background in Gylfaginning to Skáldskaparmál. The fundaments of theory could aid in the larger synthesis of poetic composition. This line of thought might be extended towards the Auraicept, which is also concerned with poetry, though the structure of the work is less well-defined and not as gradual. To round up the discussion on practical and theoretical utility it is requisite to add that, as a literary expression, the vernacular grammar would ‘add to’ vernacular learning more generally. In his De schematibus et tropis, Bede provides us with biblical examples on all the tropes and figures which are descriped in the third book of Ars maior (see Clunies Ross & Wellendorf 2014: xxx-xxxi). This is, mutatis mutandis, not very unlike the project of Óláfr Þórðarson. The same attitude was applied to the letters. In the introduction to the vernacular alphabets (1.3.) it was anticipated that the grammatical genre staged the entrance of the vernacular alphabet into the domain of the Latin alphabet; not as a full-blown book-script, but as a fundamental tool of linguistic thought and literary ideology. This is reiterated here as the preceding analysis will hopefully have shed light on the implications of this remark. The originality of the treatises does not lie in a reassessment of scribal practice or of the correlation between speech and writing, but rather in their attitude and willingness to challenge the authority of the lingua sacra and its alphabet – in theory if not in actual deed. In theatrical terms, the vernacular grammars present the reader with a persuasive mise-en-scène. The foreign cast (Latinate features) is replaced with a native one, whose imperfections are played down and whose advantages are accentuated by skilfull direction of the spotlight.139

139 Compare Vivien Law on the Occitan grammarian Guilhem Molinier: “This attitude is characteristic of medieval grammarians of the vernacular. They take over only those categories which seem to them to suit their language and jettison the rest, rejoicing in the similarities but perfectly happy to admit that their language has no declensions or whatever it might be. Almost paradoxically, their assumptions about the status of Latin and the vernacular helped enormously. Knowing as they did that Latin, being one of the three sacred languages, was qualitatively different from any vernacular, they did not expect to be able to discover a vernacular analogue for every single phenomenon of Latin grammar. Interested

3.2. Texts as Part of a Curriculum The monastic schools of Europe did not regularly present a course of study with well-defined structure, with clear progression and a graded curriculum, features which are rather innovations associated with the universities (see Law 2003: 158). It is therefore conspicuous that just such a graded curriculum is mentioned in MVII; where the course of study is divided into twelve years, each of which is devoted to certain topics and prescribed texts (Thurneysen 1891: 32 ff.; for a brief discussion and overview see pp. 115-120).140 The Auraicept is, as have been mentioned, one of these prescribed texts. A rather striking difference between the Auraicept and the ThGT is the degree to which the first text has been vastly expanded by several layers of commentary. The question- and-answer form might give the impression of schoolrom usage, but at the same time this was a well-established form of literary expression (which, in the Auraicept, does not follow the dialectical method – the ‘questioner’ is more or less an empty vessel eager to be filled with knowledge). It is therefore of limited value as evidence of use in practical education. Rijcklof Hofman has even argued (2013) that the original form of the canonical Auraicept may have been made up of core text plus commentary. For all that, it is highly valuable as evidence of reception, and the result is a markedly dynamic discourse which rules out any possibility of assessment through a parameter as authorial intention.141 In the case of the ThGT, on the other hand, authorial intention must be, and has been, taken into consideration. Óláfr’s connection to the Danish court and the implications of this has, however, been somewhat neglected in the preceding argument. A fuller discussion on this topic would have been a desideratum, though it is somewhat extrinsic to the methodological approach that has been followed.

only in arriving at a serviceable framework for describing the vernacular, they had no compunction about dropping what didn’t suit them. They were completely free from any patriotic urge to force their language into the Latin mould,” (Law 2003: 203). I would like to attribute to the Auraicept and the ThGT a considerable ‘patriotic urge’, and the texts sometimes make themselves guilty of stretching their evidence, but otherwise the attitude described by Law is certainly relevant to both.

140 MVI and MVII belongs to the same school, while MVIII is independent of these (Thurneysen 1891: 124-25). 141 The attribution of the authorship of the Auraicept to Cenn Fáelad might be dismissed as irrelevant to our purposes. The tradition is inconsistent and the one possibility rules out the other. Ahlqvist (1982: 18) is more cautious on this point than the present writer.

3.3. History of Linguistics and the Later Tradition The Auraicept na nÉces is classified as a ʻparticular grammar’ (as opposed to ʻuniversal grammar’) by Vivien Law (2003: 191-92), aiming to ‘describe the features of a specific language’. This is arguably true as far as the oldest layer of the text is concerned. For, as the multiple quotations from the commentary will undoubtedly have shown, this ventures at times into discourse on details of considerable abstraction, and resorts to extra-linguistic realities to explain linguistic phenomena. Consider for instance the discourse on tomus fri fid (ʻmeasure with regard to fid ’) (Calder 1917: 79ff.). These are features that Law assigns to the universal grammars, among whom she counts the ThGT (Law 2003: 191). Priscian combined traits from the Schulgrammatik and regulae genres (Law 2003: 88). His influence thus makes this divide less suitable in describing the vernacular grammars. I think Law’s classification of the Auraicept is problematic at the very least, and have difficulties in seeing the value of universal versus particular in the context of vernacular grammar.142 It is of course possible to force it onto the texts, but then considerable overlap has to be acknowledged. The questions asked in the universal grammars, such as ‘what is a proper noun?’, ‘how many word classes are there?’ (Law 2003: 190), are of the same kind as we find in the Instt. gramm., and there is no reason to suppose that the division universal/particular would be of concern to the vernacular grammarians. Furthermore it provides only minuscule aid in categorizing the material from a modern point of view - a separation of vernacular- medium grammars of Latin from vernacular-medium grammars of the vernacular (ibid.) seems far more relevant. Compared with the Bardic grammatical tracts the Auraicept is, predictably, reactionary in spirit (McManus 1991: 148), but in a late seventh- or early eight-century context it was anything but conservative. The ThGT is quite different at this point; in important ways it was an innovation, but it certainly shows reactionary, if not revivalist, tendencies. At the time of composition a comprehensive treatment of skaldic poetry had been carried out by Óláfr’s uncle Snorri Sturluson, and, although their object of study was the same, their approaches could hardly have diverged more (see further Kristján Árnason 2016: 211). The questions which were asked about language more or less throughout the textual history of the Auraicept belong to the same paradigm. A shift in the methodological approach to language is clearly seen in the Irish Grammatical Tracts (IGT; ed. Bergin 1916-55; IGT-I ed. Mac Cárthaigh 2014). The first editor of these texts saw this as fortunate:

142 The context in which it is launched.

At a much earlier period we find tracts like the Uraicept, full of absurd speculations on the origin of the Gaelic language. These are repeated and re-echoed for centuries in the worthless etymological glosses which disfigure the law commentaries. But the tracts which I now propose to publish belong to a different class. In the present instalment, it is true, the Uraicept is quoted at the beginning; but when once he has left behind the mythical invention of the alphabet, the author is on solid ground, and confines himself to facts (Bergin 1916: i).

The Latin grammarians’ discourse was heavily dependent on terminology with a basis in writing. It is therefore only natural that Ogam terminology became important in Irish grammatical discourse, but the IGT-I shows that, at this stage, the vernacular alphabet has itself ceased to be of interest in the analysis of language, although the term ogham itself is used to denote ‘spelling’, as opposed to gáoidhealg or ‘pronunciation’. The use of the Ogamic names of the letters, e.g. coll go n-úathadh (Bergin 1916: §53) ‘coll with aspiration’, suggests that the heritage of the Auraicept sticks somewhat deeper than Bergin would concede. That heritage might be justly termed ‘scholastic’, being disentangled from the older tradition (but see Ahlqvist 1982: 10), although, as mentioned in the introduction (1.3.), these two traditions have most likely sprung out from the same source. The continuity factor shows that the ideas conveyed by the texts were not all leading to dead ends. They also represented new beginnings, some of which are still visible today. Consider for instance the continued (though antiquarian) use of grammatical terminology from the ThGT (Gade 2007). The methodological integrity of the Icelandic grammatical tradition shows a significant decrease after the composition of FGT (i.e. mid-12th century). This is seen in the lack of independence from Latin sources and in faithful presentation of evidence. The integrity or independence of ideology, on the other hand, seems to stand in reciprocal proportion to this. The SGT (ed. Raschellà 1982) is admittedly reluctant to fit this scheme, though, as it “seems to represent an attempt at definitive disengagement from the Latin grammatical tradition: on the one hand, through the partial recovery and positive re-evaluation of a native tradition and, on the other hand, through the addition of new elements, which were partly drawn from ‘new’ philosophical experiences (such as the rediscovery of Aristotelian thought) and partly elaborated by the author himself,” (Raschellà 1982: 121-22). This makes the overall character of SGT very different from that of the ThGT. Rather than disengaging himself from the Latin grammatical tradition, Óláfr reinforces its importance in the study of the vernacular. The

92 reader is reminded that the chronological relationship between these two texts is the opposite of their order in W, as argued by Raschellà (1982). On the methodological integrity of the Irish tradition Ahlqvist states the following, which will introduce my final remark on the place of the vernacular treatises within the history of linguistics:

A fairly long acquaintance with Latin grammar may help to explain not only the critical attitude adopted in the Auraicept na nÉces towards Latin, but also the very fact that the native language had come to be felt as something worthy of study. This stands in sharp contrast with the circumstances of the reception of Latin grammar in Anglo-Saxon England. [...] It seems to me not impossible that the contrast that we find here between Ireland and Anglo-Saxon England is connected with the length of time for which native scholars had been engaged in studying Latin and the vernacular language (Ahlqvist 1989: 44-45).

To the present writer, it seems not implausible that Ahlqvist’s argument might be extended to cover also the Icelandic case, though ‘length of time’ is of course a fairly opaque criterion.143 At any rate, the ThGT, postdating as it does a considerable amount of vernacular literature, is a comparably late production, and so the native scholars (following Ahlqvist) would have had a fairly long acquaintance with Latin grammar before its composition. The sharp contrast with Anglo-Saxon England would have to be abandoned, though, as there is strong reason to believe, with Gade (2007), that this tradition provided at least the Third Grammarian with indispensable impetus.144 With an eye towards the practical utility of the vernacular grammars, it is perhaps just as reasonable to assign the lack of linguistic introspection in Anglo-Saxon England to its not being a prerequisite for the composition of vernacular literature: thus, instead of invoking ‘time’ as a parameter, the argument is harmonized with that which is outlined in the present thesis.

143 Why would long-term engagement with Latin literature shift attention towards the vernacular? 144 “In most parts of medieval Europe such a phenomenon [use of the vernacular rather than Latin] would likely indicate a relatively poor knowledge of Latin in the local community and this has been a common modern assumption about medieval Iceland,” (Clunies Ross 2005: 145n5). The expressed purpose of Ælfric’s grammar supports Clunies Ross’ contention.

4. BIBLIOGRAPHICAL REFERENCES AND ABBREVIATIONS

4.1. Manuscripts AM 28 8vo: Codex Runicus. Consult Handrit.is: https://handrit.is/en/manuscript/view/da/AM08-028.

Cod. Sang. 904: Prisciani grammatica. Codices Electronici Sangallenses. (CESG), E-CODICES, Virtual Manuscript Library of Switzerland: http://www.e-codices.unifr.ch/

RIA 536 (23 P 12) fol. ʻBook of Ballymote’ (BB). Irish Script on Screen (ISOS). http://www.isos.dias.ie

GKS 1812 4to. Handrit.is: http://handrit.is/is/manuscript/view/is/GKS04-1812 (visited November 11, 2015).

4.2. Editions and Translations Ahlqvist, Anders. (1982): The Early Irish Linguist: An Edition of the Canonical part of Auraicept na n-Éces, with Introduction, Commentary and Indices, Helsinki: Societas Scientiarum Fenica.

Auraicept na nÉces – see Ahlqvist (1982); Calder (1917).

Barney, Stephen A., Lewis, W. J., Beach, J. A. and Berghoff, Oliver. (2006): The Etymologies of Isidore of Seville, Cambridge University Press, Cambridge.

Bergin, Osborn. (1916-55): Irish Grammatical Tracts, supplements to Ériu, vol. 8 (1916), vol. 9 (1921-23), vol. 10 (1926-28), vol. 14 (1946-47) and vol. 17 (1955).

Björn M. Ólsen (1884): Den tredje og fjærde grammatiske avhandling i Snorres Edda tilligemed de grammatiske avhandlingers prolog og to andre tillæg, Samfundet til udgivelse af gammel nordisk literatur, København.

Calder, George. (1917): Auraicept na n-Éces. The Scholar’s Primer, Edinburgh: Grant.

Canale, Angelo Valastro. (2014): Isidoro di Siviglia. Etimologie o origini, 2. vols., UTET libreria, Torino.

Caplan, Harry. (1954): Rhetorica ad Herennium, Loeb Classical Library, Harvard University Press, Cambridge, Massachusetts.

Clunies Ross, Margaret and Wellendorf, Jonas. (2014): The Fourth Grammatical Treatise, Viking Society for Northern Research, University College London.

Faulkes, A. (1991): Snorri Sturluson. Edda. Háttatal, Viking Society for Northern Research, London.

- (1998): Snorri Sturluson. Edda. Skáldskaparmál, Viking Society for Northern Research, University College London.

- (2005): Snorri Sturluson. Edda. Prologue and Gylfaginning, Viking Society for Northern Research, London.

Finnur Jónsson. (1886): Den første og anden grammatiske afhandling i Snorres Edda, Samfund til udgivelse af gammel nordisk litteratur, København.

- (1927): Óláfr Þórðarson: Málhljóða- og málskrúðsrit. Grammatisk-retorisk afhandling, Copenhagen: Bianco Lunos bogtrykkeri.

FGT: First Grammatical Treatise – see Hreinn Benediktsson (1972).

FoGT: Fourth Grammatical Treatise – see B. M. Ólsen (1884); Clunies Ross & Wellendorf (2014).

GL: Grammatici Latini – see Keil (1855-80).

Hancock, W. Neilson, et. al. (1865): Ancient laws of Ireland, 6 vols, vol. 1: Senchus Mor, Stationery Office: Dublin.

Hreinn Benediktsson. (1972): The First Grammatical Treatise, Reykjavík, Institute of Nordic Linguistics.

IGT: Irish Grammatical Tracts – see Bergin (1916-55).

Instt. gramm.: Institutiones grammaticae - see Keil (1855-80).

Keil, H. (1855-80): Grammatici Latini. In aedibus B. G. Teubneri, Lipsiae.

LG: Lebor Gabála Érenn – see Macalister (1938-56).

Macalister, R. A. Stewart. (1938-56): Lebor Gabála Érenn. The Book of the Taking of Ireland, Irish Texts Society, 5. vols., London.

Meyer, Kuno. (1913): Sanas Cormaic. An Old Irish Glossary Compiled by Cormac Úa Cuilennáin, King-Bishop of Cashel in the Tenth Century. Edited from the Copy in the Yellow Book of Lecan, in: Anecdota from Irish Manuscripts, Bergin, Best, Meyer and O’Keefe, vol. V., Hodges Figgis & Co, Dublin [repr. with editor’s corrections, Llanerch Publishers 1994].

MV(I-III): Mittelirische Verslehren I-III - see Thurneysen (1891).

Raschellà, Fabrizio D. (1982): The So-Called Second Grammatical Treatise. An Orthographic Pattern of Late Thirteenth-Century Icelandic, Filologia Germanica, Testi e Studi II, Felice Le Monier, Firenze.

Rask, Rasmus K. (1818): Snorra-Edda ásamt Skáldu og þarmeð fylgjandi ritgjörðum eptir gömlum skinnbókum, Stockholm.

SGT: Second Grammatical Treatise – see Raschellà (1982).

SnE: Snorra Edda – see Faulkes (1991-2005).

Stokes, Whitley and Strachan, John. (1903): Thesaurus Palaeohibernicus. A Collection of Old-Irish Glosses, Scholia and Verse, Vol. II., Dublin Institute for Advanced Studies, Dublin.

Stokes, Whitley. (1905): Félire Óengusso Céli Dé. The martyrology of Oengus the Culdee, (HBS, 29), London: HBS.

Tavoni, Mirko. (2011): De vulgari eloquentia, in Dante Alighieri. Opere, ed. Marco Santagata, vol. 1 (2nd edition 2015), Mondadori.

The Holy Bible Douay-Rheims Version, John Murphy Company, 1899.

Thes. Pal.: Thesaurus Palaeohibernicus – see Stokes & Strachan (1903).

ThGT: Third Grammatical Treatise – see Rask (1818); B. M. Ólsen (1884); Finnur Jónsson (1927) and Wills (2001).

Thurneysen, Rudolf. (1891): Mittelirische Verslehren, in Irische Texte, 3. serie, Whitley Stokes & Ernst Windisch, 4. vol. Leipzig: Hirzel, I: pp. 1–182. (Use has been made of Carl Marstrander’s copy in Mag 315, UHS Library, UIO, equipped with hand-written notes on the texts and a partial translation into Norwegian of MVI and parts of MVII).

- (1917): “Morands Fürstenspiegel”, in Zeitschrift für celtische Philologie, vol. 11, pp. 56-107.

Uhlig, Gustav. (1883): Dionysii Thracis Ars grammatica. Grammatici graeci, vol. 1, Leipzig: Teubner 1883.

Weber, Robert and Gryson, Roger. (1994): Biblia Sacra Vulgata. Editio quinta, Vierte verbesserte Auflage, Deutsche Bibelgesellschaft, Stuttgart.

Wills, Tarrin. (2001): The Foundation of Grammar: An edition of the first section of Óláfr Þórðarson’s grammatical treatise, PhD thesis, University of Sydney, published at http://homepages.abdn.ac.uk/cgi- bin/cgiwrap/wag017/mg-new.cgi.

Zupitza, Julius. (1880): Ælfrics Grammatik und Glossar, Weidmannsche Buchhandlung, Berlin.

4.3. Monographs and Articles Acken, James. (2008): Structure and Interpretation in the Auraicept na nÉces, VDM Verlag 2008.

Adams, George Brendan. (1970): “Grammatical analysis and terminology in the Irish Bardic schools”, in Folia Linguistica 4, pp. 157-166.

Ahlqvist, Anders. (1980): “Det medeltida Irlands språkvetenskap,” in: Nordic languages and modern linguistics 4: Proceedings of the fourth International Conference of Nordic and General Linguistics in Oslo 1980, ed. Hovhaugen, Even, Universitetsforlaget, pp. 202-211.

- (1986): “Gramadóirí Gaeilge agus Laidine”, in Léann na cléire, LCC XVI, in eagar ag Pádraig Ó Fiannachta, An Sagart, Maigh Nuad, pp. 54-71.

- (1989): “Latin Grammar and Native Learning,” in: Sages, Saints and Storytellers. Celtic Studies in Honour of Professor James Carney, Maynooth Monographs 2, ed. Ó Corráin, Breatnach, McCone, An Sagart, Maynooth, pp. 1-6.

Barns, Michael P. (2012): Runes: A Handbook, Boydell Press.

Bauer, Alessia. (2013): “Reflexionen über die eigene Sprache: irische und norröne Tradition im Vergleich”, in Between the Islands and the Continent: Papers on Hiberno-Scandinavian-Continental relations in the Early Middle Ages. Studia medievalia septentrionalia, vol. 21, ed. Rudolf Simek and Asya Ivanova, Wien 2013, pp. 9-29.

Bergin, Osborn. (1938): “The Native Irish Grammarian”, in Proceedings of the British Academy (24), pp. 205-35.

Björn M. Ólsen (1883): Runerne i den oldislandske literatur, København 1883.

Burnyeat, Abigail. (2007): “The Early Irish Grammaticus?”, Aiste I (2007), pp. 181-217.

Charles-Edwards, T. M. (1998): “The context and uses of literacy in early Christian Ireland”, in Literacy in Medieval Celtic Societies, Huw Pryce, Cambridge University Press, Cambridge 1998, pp. 62-83.

Clunies-Ross, Margaret. (2005): A History of Old Norse Poetry and Poetics, D. S. Brewer, Cambridge.

DIL: Dictionary of the Irish Language, - see Quin (1983). eDIL: Electronic Dictionary of the Irish Language at http://www.dil.ie.

Gade, Kari Ellen. (2007): “Ælfric in Iceland,” in: Learning and Understanding in the Old Norse World: Essays in Honour of Margaret Clunies Ross, Brepols Publishers, pp. 321-339.

Guðrún Nordal. (2001): Tools of Literacy. The Role of Skaldic Verse in Icelandic Textual Culture of the Twelfth and Thirteenth Centuries, University of Toronto Press.

Haugen, Odd Einar and Thomassen, Einar. (1990): Den filologiske vitenskap, Solum forlag, Oslo.

- (2004): “Paleografi”, in: Handbok i norrøn filologi (ed. Odd Einar Haugen), Fagbokforlaget, pp. 175-215.

Hayden, Deborah. (2011): “Poetic Law and the Early Irish Linguist: Contextualizing the Vices and Virtues of Verse Composition in Auraicept na nÉces”, in Language and History, vol. 54, pp. 1-34.

- (2012): “Some Notes on the Transmission of Auraicept na nÉces”, PHCC 32 (2012), pp. 134-79.

Hofman, Rijcklof. (2013): “Latin Grammars and the Structure of the Vernacular Old Irish Auraicept na nÉces”, in Spoken and Written Language. Relations Between Latin and the Vernacular Languages in the Earlier Middle Ages, Mary Garrison, Marco Mostert and Arpad P. Orbán with the assistance of Wolfert S. van Egmond, Utrecht Studies in Medieval Literacy Brepols, Turnhout, pp. 185-98.

Irvine, Martin. (1994): The Making of Textual Culture. ʻGrammatica’ and Literary Theory 350-1100, Cambridge University Press, Cambridge.

Johansson, Karl Gunnar. (1997): Studier i Codex Wormianus. Skrifttradition og avksriftsverksamhet vid ett islandsk skriptorium under 1300-talet, Acta Universitatis Gothoburgensis, Göteborg.

Jón Axel Harðarson (2003): “Rezension von Krömmelbein (Hrsg.) Dritte Grammatische Abhandlung”, Philologia Fenno-Ugrica (9), pp. 59-71.

Kelly, Fergus. (1988): A Guide to Early Irish Law, Early Irish Law Series, vol. 3., Dublin Institute for Advanced Studies, Dublin.

- (2002): “Texts and transmission: the law-texts”, in Ireland and Europe in the Early Middle Ages: Texts and Transmission, by Próinséas Ní Chatháin & Michael Richter, Four Courts Press, Dublin, pp. 230-242.

Kristján Árnason. (2016): “Vernacular and classical strands in Icelandic poetics and grammar in the Middle Ages”, in NOWELE. North-Western European Language Evolution, vol. 69:2 (2016), pp. 191-235.

Lambert, Pierre-Yves. (1987): « Les premières grammaires celtiques, » in: Histoire Épistémologie Langage, tome 9, fascicule 1, 1987. Les premières grammaires des vernaculaires européens, pp. 13-45.

Law, Vivien. (1982): The Insular Latin Grammarians, The Boydell Press, Woodbridge, Suffolk.

- (2003): The History of Linguistics in Europe. From Plato to 1600, Cambridge University Press, Cambridge.

- (2004): “The Study of Grammar”, in Carolingian culture: emulation and innovation, ed. Rosamond McKitterick, University of Cambridge Press, pp. 88-110.

Luhtala, Anneli. (2005): Grammar and Philosophy in Late Antiquity: A Study of Priscian’s sources, Studies in the History of the Language Sciences, vol. 107, John Benjamins Publishing.

Males, Mikael. (2013): “Wormianusredaktören: Språk, tro och sanning vid 1300-talets mitt”, Arkiv för nordisk filologi (128), pp. 41-77.

- (2014): “Skáldskaparmál as a Tool for Composition of Pseudonymous Skaldic Poetry”, in Chase Eddic, Skaldic and Beyond. Poetic Variety in Medieval Iceland and Norway, Fordham University Press, New York 2014, pp. 62-75. - (2015): “Er Ólafur Þórðarson höfundur Eglu?”, Són. Tímarit um óðfræði, 13. hefti, ed. Anna Þorbjörg Ingólfsdóttir and Haukur Þorgeirsson, Óðfræðifélagið Boðn, Reykjavík, pp. 173-181. - (2016): “Applied Grammatica: Conjuring up the Native Poetae”, in Intellectual Culture in Medieval Scandinavia c. 1100-1350, ed. Stefka Georgieva Eriksen, Brepols, Turnhout 2016, pp. 263-308.

McCone, Kim. (1996): Towards a Relative Chronology of Ancient and Medieval Celtic Sound Change, Maynooth Studies in Celtic Linguistics I, Department of Old and Middle Irish, St. Patrick’s College, Maynooth.

McLaughlin, Roisin. (2009): “Fénius Farsaid and the Alphabets”, Ériu, vol. 59 (2009), pp. 1-24.

McManus, Damian. (1986): “Ogam: archaizing, orthography and the authenticity of the manuscript key to the alphabet”, “, Ériu, vol. 37, pp. 1-31.

- (1988): “Irish letter-names and their kennings”, Ériu, vol. 39, pp. 127–168. - (1991): A Guide to Ogam, Maynooth Monographs 4, An Sagart, Maynooth.

Meroney, Howard. (1945): “Fénius and Gáedel in the ʻLebar Cindfáelad’”, in Modern Philology, Vol. 43., No. 1., University of Chicago Press, pp. 18-24.

Micillo, Valeria. (1999): “Die grammatische Tradition des insularen Mittelalters in Island. Spuren insularer Einflüsse im Dritten Grammatischen Traktat”, in Adaptation und Akkulturation im insularen Mittelalter, ed. Erich Poppe & L.C.H. Tristram, Münster 1999, pp. 215-229.

Myrvoll, Klaus Johan & Skomedal, Trygve (2010): “Tonelagsskilnad i islendsk i Tridje grammatiske avhandling”, Maal og Minne, pp. 68- 97.

Ó Cróinín, Dáibhí. (1995): Early Medieval Ireland. 400-1200, Longman.

Ó Cuív, Brian. (1983): “The Linguistic Training of the Mediaeval Irish Poet”, Statutory Lecture 1969, School of Celtic Studies, Dublin Institute for Advanced Studies.

ONP – Ordbog over det norrøne prosasprog. Den arnamagneanske kommisjon, København 1989.

Pedersen, Holger. (1909-13): Vergleichende Grammatik der keltischen Sprachen, vols. 1-2., Vandenhoeck und Ruprecht, Göttingen.

Poppe, Erich. (1996a): “Die mittelalterliche irische Abhandlung Auraicept na nÉces und ihr geistesgeschichtlicher Standort”, in Theorie und Rekonstruktion. Trierer Studien zur Geschichte der Linguistik, Klaus D. Dutz & Hans J. Niederehe, Münster: Nodus, pp. 55-74.

- (1996b): “Latinate Terminology in Auraicept na nÉces”, in History of Linguistics 1996, Vol. 1: Traditions in Linguistics Worldwide, Studies in the History of the Language Sciences 94, David Cram, Andrew R. Linn and Elke Nowak (Amsterdam 1999), pp. 191-201.

- (2002): “Latin Quotations in Auraicept na nÉces: Microtexts and their Transmission”, in Ireland and Europe in the Early Middle Ages: Texts and Transmission, by Próinséas Ní Chatháin & Michael Richter, Four Courts Press, Dublin, pp. 296-312.

Quin, E. G. (1983): Dictionary of the Irish Language (DIL), Royal Irish Academy, Dublin 1983.

Raschellà, Fabrizio D. (1994): “Rune e alfabeto latino nel trattato grammaticale di Óláfr Þórðarsson”, in Sagnaþing helgað Jónasi Kristjánssyni sjötugum 10. apríl 1994, ed. Gísli Sigurðsson, Guðrún Kvaran and Sigurgeir Steingrímsson, 2 vols., Reykjavík: Hið íslenska bókmenntafélag, pp. 679-90.

- (2005): “Old Icelandic Grammatical Literature: The Last Two Decades of Research (1983-2005,” in Learning and Understanding in the Old Norse World: Essays in Honour of Margaret Clunies Ross, ed. Judy Quinn, Kate Heslop, and Tarrin Wills, Brepols Publishers, pp. 341-372. - (2016): “Óláfr Þórðarson and the ‘Norse alphabet’: A thirteenth-century Icelandic grammarian’s account of runic writing”, in NOWELE. North-Western European Language Evolution, vol. 69:2 (2016), pp. 155-191.

Rekdal, Jan Erik. (2012): “Macaronic texts in the early Irish tradition”, in High vs. Low and Mixed Varieties. Status, Norms and Functions across Time and Languages, ed. Gunvor Mejdell and Lutz Edzard, Harassowitz Verlag, Wiesbaden, pp. 169-178.

Russell, Paul. (1996): “Moth, Toth, Traeth: Sex, Gender and the Early Irish Grammarian”, in History of Linguistics 1996, Vol. 1: Traditions in Linguistics Worldwide, Studies in the History of the Language Sciences 94, David Cram, Andrew R. Linn and Elke Nowak (Amsterdam 1999), pp. 203-13.

Seppänen, Minna. (2014): Defining the art of grammar: Ancient perceptions of γραμματική and grammatica, (Dissertation), Turun Yliopiston Julkaisuja/Annales Universitatis Turkuensis, University of Turku, Turku 2014.

Spurkland, Terje. (2001): I begynnelsen var fuþark. Norske runer og runeinnskrifter, Cappelen Akademisk Forlag/Landslaget for norskundervisning, Oslo.

100

Svanhildur Óskarsdóttir. (ed.) (2013): Handrit úr fórum Árna Magnússonar, Den arnamagnæanske samling, Nordisk forskningsinstitutt, Bókaútgáfan Opna.

Thurn. Gramm. : A Grammar of Old Irish – see Thurneysen 1946.

Thurneysen, Rudolf. (1928): “Auraicept na n-éces”, in Zeitschrift für celtische Philologie, vol. 17, pp. 277-303.

- (1946): A Grammar of Old Irish, tr. D. A. Binchy and Osborn Bergin, Dublin: School of Celtic Studies, Dublin Institute for Advanced Studies.

Thurot, Charles. (1868): Extraits de divers manuscrits latins pour servir à l’histoire des doctrines grammaticales au moyen âge, Paris.

Tranter, Stephen. (1997): Clavis Metrica. Háttatal, Háttalykill and the Irish Metrical Tracts, Beiträge zur nordischen Philologie, 25. Band, Helbing & Lichtenhahn Verlag AG, Basel und Frankfurt am Main.

- (2000): “Medieval Icelandic artes poeticae”, in Clunies Ross Old Icelandic Literature and Society, Cambridge Studies in Medieval Literature 42, Cambridge University Press.

Wills, Tarrin. (2016): “The thirteenth-century runic revival in Denmark and Iceland”, in NOWELE. North- Western European Language Evolution, vol. 69:2 (2016), pp. 114-129.

ZCP: Zeitschrift für celtische Philologie, est. 1897 by Kuno Meyer and Ludwig Christian Stern.

101

102

5. APPENDICES

5.1. Keys to Alphabets (Epigraphic) Ogam is usually transliterated by means of capital letters. Runic inscriptions by minuscule letters with bold weight. These practices are followed here.

BASIC INVENTORY:

B L F S N M G NG Z R ᚁ ᚂ ᚃ ᚄ ᚅ ᚆ H ᚇ D ᚈ T ᚉ C ᚊ Q ᚋ ᚌ ᚍ ᚎ ᚏ ᚐ A ᚑ O ᚒ U ᚓ E ᚔ I

ADDITIONAL LETTERS WITH MS VALUES (AURAICEPT NA nÉCES):

EA OI UI IO Æ ᚕ ᚖ ᚗ ᚘ ᚙ

The MS key (as in BB: 176ra1, 2) reproduced in Gaelic type:

b l f Š n h d t c q m g n g s r r a o u e i ea oi ui io æ

The following key to the runic alphabet is reproduced from Spurkland 2001 (166) and is a standard inventory of Norwegian runic inscriptions during the High Middle Ages:

16-RUNES’ FUÞARK:

f U q o R k h n i a sSc t B m l Y f u þ o r k h n i a s t b m l y

LATER ADDITIONS:

e æ ö/\ g d p/P c e æ ø ǫ g d p c/z

With the following changes/additions the above equals the inventory of the ThGT:

1) = y replaces Y. 2) The following are added: z ey, ø eo (ǫ).

103

See also Raschellà 2016 (p. 160). Consult also AM 28 8vo (1r, additional leaf), which contains the runic alphabet together with transliteration in Latin letters.145 The order of the letters follows the Latin alphabet.

In quotations from the texts I have generally, though not invariably, used the original signs in the original text and transliteration in the translations.

5.2. On Quotations from Editions and Translations The canonical Auraicept is quoted from Ahlqvist 1982. Commentary from Calder 1917. The ThGT is quoted either from B. M. Ólsen 1884 or Wills 2001. Quotations from the first are normalized throughout. The latter edition contains a normalized version itself. Translations are quoted from those editions where they occur, unless I have had reason to disagree with the choices of the translators. Priscian is mostly translated by the present writer, but some passages have been translated elsewhere and are quoted here. The typesetting of the editions is generally respected, but minor alterations (e.g. in punctuation, which does not derive from manuscript usage at any rate) occur for sake of legibility. Issues caused by different levels of textual representation (normalized, diplomatic, facsimile) is generally not relevant to the argument, and though I have refrained from standardizing Calder’s text (with the exception of one passage from the YBL-version located at Calder 1917: ll. 4143-44).

145 A high-resolution photography of this leaf is obtainable at: https://image.landsbokasafn.is/source/AM_28_8vo/AM_28_8vo,_accMat01r_-_200-hq.pdf (visited October 17th 2016).

104