Latin and Old Welsh

When indirectly investigating the differences in the teaching and learning strategies adopted for Latin and for Greek in medieval sources such as the manuscript of interest in this paper, a good starting point is to compare the type and complexity of the texts that were transcribed. It is, after all, entirely reasonable to presume that a text of high complexity in a given language implies a high level of competence in that language. It must also be added that a solid grasp of a language would aid the comprehension not only of texts which show a high degree of linguistic intricacy (be it on a morpho-syntactical or a purely lexical level) but also of texts which – regardless of their level of linguistic transparency or obscurity – deal with intricate matters since confidence and good acquaintance with a linguistic system result in a significant reduction of the cognitive load that comes with approaching texts that present difficulties from a content-related point of view. In the specific case of the Liber Commonei, the comparison between its Latin and its Greek contents does in fact align with our preliminary expectations regarding the matter at hand. We can indeed expect to observe a better understanding of Latin than Greek on the part of 9th-century monks in Britain; therefore, it is not surprising to detect a substantially higher level of complexity in the Latin texts which form the first portion of the Liber (ff. 20r–23v): mathematical and astronomical calculations; intricate theological commentaries on biblical passages; and long-winded discussions on capacity, weight, and length measurement systems (along with their schematic references). These are ostensibly more challenging than Bible excerpts when taken at face value, and they surely require not only a passive knowledge but rather a competent use of the Latin language.

Nevertheless, this preliminary comparison offers limited satisfaction to our investigation. While it aligns with our pre-existing empirical assumptions regarding language learning, it ultimately relies on an argument ex silentio: the absence of Greek texts exhibiting comparable intricacy within this specific manuscript (or, in fact, elsewhere in the entire composite codex) does not necessarily imply that the monks who transcribed this manuscript would be unable to engage with them; it is, in fact, merely a suggestion in that direction. To take a more meaningful glance at the language learning strategies involved and at the scribes’ level of competence in Classical languages, I propose instead to consider the testimony offered by the glossing practice as it emerges from this manuscript, both in its Latin and in its vernacular iteration. Through this analysis, it may be possible to confirm (or, eventually, to reject or at the very least re-evaluate) the scenario suggested by the thematic disparity between the Latin and the Greek texts in question.

In this respect, the first observation that must be brought forward is that, while (as mentioned) there is no shortage of bilingual glosses and comments (both interlinear and marginal) on the very first portion of the Liber,Footnote ¹⁰ in its second part (with the biblical extracts in Greek and Latin) glosses are basically absent since the very few interlinear or marginal annotations amount to simple corrections or small additions to the text. This absence is significant in and of itself, as we will see (Section 3), but to properly evaluate the glossing practice observed in the first portion, we need a more substantial comparison.

In light of this, a useful counterpart can be found in the quire that immediately follows the Liber Commonei in the composite codex: the text of Book I of Ovid’s Ars amatoria. Such comparison is warranted: Although the Ovidian copy constitutes a distinct manuscript that only by chance happened to be bound together with the Liber Commonei at a later stage, the two texts were undoubtedly produced in a very similar time, place, and environment – a Welsh monastery in the 9th century. Therefore, the glosses attested in the two adjacent manuscripts can be reasonably expected to represent two separate instances of the same glossing practice, and it is not too far-fetched to assume that different modalities in the glossing of the two texts reflect a difference in the approach to (and in the practical use of) these texts by the British monks, rather than a fundamental difference in the glossing tradition. Naturally, this conclusion cannot be understood as definitive, since there cannot be definitive proof of the fact that, before being bound together at Glastonbury, the two manuscripts were perused in the same context; we will then need to turn to the functional analysis of the two series of glosses to see whether a pattern can be identified.

The Oxonian copy of the Ars amatoria, limited to its first book (with a few sparse textual lacunae, amounting to seven verses in total), has been recently edited by Paul Russell.Footnote ¹¹ Russell’s edition extends not only to the main text of the poem but also to the numerous Latin and Old Welsh glosses present in the manuscript. Among Russell’s comments on the observable glossing practice, his remarks on the ‘pedagogical’ nature of many of the annotations featured in his Introduction are particularly interesting for the purposes of this paper:

Russell’s words are especially significant when paired with the fact that, in the Ars amatoria, many instances of the grammatical glosses are in the vernacular language, with many examples of a formulaic usage of prepositions to signify the correct Latin synthetic case in a language such as Old Welsh that had long lost synthetic case-marking (for example, o/or ‘from/from the’, followed by the Old Welsh gloss, is used to represent a Latin ablative; much in the same way, di ‘to, towards’ is used to translate nouns and adjectives in the dative case, for example, di aperthou ‘to [the] offerings’, which translates muneribus).Footnote ¹³ In cases such as these, the pedagogical value of the glosses and, consequently, the use of the text of Ovid’s poem as a language learning tool is immediately apparent: the presence of Latin and, most importantly, of vernacular elements which are clearly used as metalinguistic categories is particularly telling in that regard.

In turn, the state of the glossing found in the first part of the Liber Commonei presents some major philological difficulties, which prevent a similarly clear-cut analysis. The main issue is that, during the copying process that resulted in the copy of the Liber that survived to the modern day within the Oxoniensis Prior, at least 2 distinct series of glosses and comments were mistakenly conflated into a single one. The end result is a manuscript portion where marginal and interlinear glosses are often misplaced (that is, spatially removed from the main text’s words or phrases which they were originally ascribed to) and miscopied, a situation which not only hinders their comprehension in general but also obscures the interplay between the two batches of glosses that can reasonably be reconstructed and have existed in the lost source manuscript. Nonetheless, eminent Celtic philologists undertook the challenge of editing these glosses.Footnote ¹⁴ Not only did they establish beyond any reasonable doubt that the apograph of the Liber Commonei must have had at least two series of glosses, but they also inferred that, most significantly, the first glossator almost exclusively used Latin, while the second supposed glossator expressed themselves bilingually, often switching between Latin and Old Welsh between sentences or even within the boundaries of a given sentence. Pierre-Yves Lambert went as far as suggestingFootnote ¹⁵ that the second glossator made use of their own vernacular language in those instances where they felt more emotionally invested in responding to the Latin notes left by their predecessor: This dynamic, if proven true, would align with our knowledge of code-switching as a general phaenomenonFootnote ¹⁶ as well as in its iterations in the ambit of Celtic medieval glosses.Footnote ¹⁷

The most significant cluster of vernacular glosses in the Liber Commonei is found in the section that is traditionally known as ‘Weights and Measures’, that is, the epitomised excerpt from Victorius of Aquitaine’s Calculus (ff. 23r–23v, with the addition of a piece of commentary copied in the lower margin of f. 22v and linked to the main text through the use of schematic signes de renvoi). Both the Latin and the Old Welsh glosses found in this portion of the text are not as straightforwardly grammatical (or, even, plainly lexical) as the ones found in the Ars amatoria (see above); instead, these comments engage with the contents of the main text at an ostensibly deeper level, discussing themes as the practical and theological implications of measures of weight mentioned in the GospelsFootnote ¹⁸ or providing equations useful to navigating the geometric system laid out by Victorius of Aquitaine and integrating it with alternative nomenclatures.Footnote ¹⁹ Moreover, it is particularly significant to observe how the Old Welsh glosses on the Calculus very rarely consist of simple lexical clarifications: Most often, the vernacular is used to integrate the aforementioned discussions with additional (and, at times, alternative) relevant information.

As a result, comparing the glosses on the first section of the Liber Commonei with the ones on the first book of the Ars amatoria reveals how the latter was employed to learn Latin at a comparatively basic level (or, at the very least, the poem was read and annotated by monks who exhibited a somewhat limited proficiency in the Latin language), while the perusal of the Liber Commonei arguably required an already established competence in Latin (and, ostensibly, some familiarity with a remarkable variety of scholarly texts) to engage with its intricate contents. An interesting parallel situation has been observed by Paul Russell as part of his research on the bilingual (again, Latin and Old Welsh) glosses to Martianus Capella’s De nuptiis Philologiae et Mercurii (Cambridge, Corpus Christi College, MS 153): When confronting two successive batches of glosses on that manuscript (one derived from a pre-existing continental commentary tradition referred to as the Oldest Glossing Tradition [OGT],Footnote ²⁰ and the other one independently developed at a later stage), Russell ultimately observed:

It is plain to see, then, how a series of lexical and grammatical glosses can be reasonably argued to reflect a more basic knowledge of Latin and, by extension, a less advanced stage in the process of language learning. Reverting back to the contents of the Oxoniensis Prior, this consideration lends further strength to the assumption that, in contrast to Victorius of Aquitaine’s Calculus, Ovid’s Ars amatoria was more or less specifically used as a reference text for Latin language learning. In this regard, it is interesting to follow RussellFootnote ²² in correlating this information with a passage from the scholastic colloquy De raris fabulis, likely originating from Cornwall, where a priest explains the reason behind his limited knowledge of Latin and, accordingly, the poor quality of his Latin production: Ignoro enim regulas grammaticorum nec exempla poetarum ‘For I am ignorant of the rules of the grammarians nor do I know the examples of the poets’ (translation by Russell). This clearly suggests that, as has been argued for the Ars amatoria, grammatical treatises and, most significantly, poetic texts made up the foundations of Latin teaching and learning in early medieval Britain, while, by contrast, in a text such as the Liber Commonei, Latin is better understood as a tool to teach and learn subjects that went beyond the vehicular language itself.

It might be argued that, in the monastic milieu that we are dealing with, the Greek language was one of the subjects that counted among its prerequisites a good knowledge of Latin, and it is now hopefully clear how the distribution of the vernacular glosses in the Liber Commonei serves as an indirect yet solid confirmation for this hypothesis.

Latin and Greek

Let us now turn to considering the evidence offered by the second part of the Liber (ff. 24r–35v, plus ff. 19r–19v; see below), which entirely consists of lectures from the Old Testament in Latin and Greek.

Folios 24r–28v contain Pauca testimonia de prophetarum libris per Graecam linguam (‘brief extracts from the Books of the Prophets in the Greek language’): The Greek text is written down in the left column of each page, while a Latin translation occupies the right column. The first important thing to note when searching for hints of the pedagogic nature of this text is the Greek alphabetic sequence that runs vertically on the left margin of the first page of this portion (24r), to the right of which the Latin transliteration of each symbol is attached. It is interesting to note that, while it is never used in the Greek texts of the manuscript, a qoppa appears in the sequence between pi and rho (along with its Latin counterpart, q). Whoever wrote this alphabet down was clearly deeply concerned with its immediate correspondence with the Latin alphabetic sequence, yet a number of egregious mistakes are found: epsilon is transliterated with ee, just as omega is rendered with oo, suggesting an erroneous interpretation on part of the scribe of epsilon as a long vowel; conversely, heta is transliterated as the Latin h. It is also interesting to note that aspiration is not consistently marked in the Latin rendition, since theta is aptly transliterated with th, but chi is transliterated with c (as opposed to kappa – k) and the correspondent letter for phi is written f. Moreover, psi is incorrectly transliterated as ph. This series of errors would have definitely had some repercussions on a transliteration effort; it must be observed, though, that, in the following portion of the manuscript, where the Greek text is written down in Latin characters, the transliteration is carried out in a way that disregards the prospect in f. 24r completely.

The Latin translation is organised in the right column in a way that reveals the desire to follow the Greek text on the left as closely as possible, not only in a linguistic sense but most strikingly in physical, spatial terms. The most telling example of this phaenomenon is arguably found at f. 28r, where the Latin line in gentibus (translation of the Greek ϵν ταις ϵθνϵσιν [sic], or ‘among the nations’; Malachi 1:11) is repeated twice, in order to maintain the exact correspondence to the Greek text after a ‘slip’:

In the fourth line of the page, occasum (sunset) is omitted from the Latin translation, and as a consequence of this distraction, the rest of the line is filled by the translation corresponding to the subsequent, fifth line of the Greek text on the left column. The scribe, then, reverted this mistake by repeating in gentibus over 2 subsequent lines, the fifth and the sixth one.

This demonstration of effort towards the total spatial adherence of the Latin column to the Greek one shows us how this translation practice can be functionally equated with glossing.Footnote ²⁵ Since the Latin text plays the same functional role as a continuative series of glosses (that is, to annotate the meaning of words or phrases to facilitate linguistic and contextual comprehension of the main text – in this case, the Greek excerpts), it is no wonder that Old Welsh glosses are nowhere to be found in this portion of the Liber Commonei. The exegetic and (here, mainly) pedagogical function of a meta-language is taken on by Latin, revealing how knowledge of Latin was considered a prerequisite to learn Greek.

While these elements alone offer clear indication of the didactic scope underlying the compilation of this manuscript, there are still some aspects to it that need to be investigated. Namely, what can this manuscript tell us about how, in practical terms, Greek language learning was pursued in this particular context? Did the monks rely on a teacher, or was the transmission of written sources their only option for learning Greek? Answers to these questions can be tentatively provided by taking into consideration the texts transcribed in the following pages, from f. 28v to f. 35v, with the addition of f. 19 (recto and verso). These pages contain, once again, excerpts from the Old Testament, but instead of featuring a Greek left column accompanied by a Latin translation on the right, in this case the primary text of the biblical extracts (on the left) is in Latin, and the right column is occupied by a Greek translation (clearly derived from the Septuagint version), phonetically transcribed using the Latin alphabet. The lectures are from the Books of Genesis (the Creation and the Binding of Isaac), Exodus, Isaiah, and Psalm 42 (‘Quemadmodum desiderat cervus’, Psalm 41 in Jerome’s Vulgata). The folio containing the final lecture, from Deuteronomy, was misplaced in its current position (ff. 19r–19v) when the Liber Commonei was bound together with the other components of the Oxoniensis Prior; for all intents and purposes, we shall consider it as the last folio of this quire, as it originally was.

The Greek transcription in the Latin alphabet is quite consistent in its rules throughout the manuscript, and some of its characteristics reveal that a ‘popular’, rather than ‘learned’, pronunciation of Early Medieval GreekFootnote ²⁶ underlies this transcription. The conjunction και, widely present in the text as the translation of Latin et, is consistently written down as ce, revealing the monophthongisation of αι into /e/ and its merging with ϵ;Footnote ²⁷ similarly, η, ι, and ϵι are clearly merged into /i/, as is evident from the very first sentence of the translation of Genesis (f. 28v): En archi epoeisen [sic] o theos ton uranon ce tin gin (ἐn ἀρχῇ ἐποίησϵν ὁ Θϵὸς τὸν οὐρανὸν καὶ τὴν γῆν; in the beginning God created the heavens and the earth). The Greek φ is consistently rendered as f, indicating a completely fricativised pronunciation, as can be seen in that same first paragraph: ce ipen o theos genethito fos (καὶ ϵἶπϵν ὁ Θϵός· γϵνηθήτω φῶς; then God said: ‘Let there be light’);Footnote ²⁸ moreover, the use of nyftan for νύκτα (night [acc.]) (f. 29r) shows not only the expected fricativisation [kt] > [xt], with [x] represented by approximation with the Latin f, but also the hypercorrect addition of a final -n, betraying the generalised loss of final nasal sounds in Medieval Greek.Footnote ²⁹ In turn, the text shows the retention of the distinction between /i/, /y/, and /u/, which aligns with what is known about the chronology of the development of the Medieval Greek vocalic system,Footnote ³⁰ albeit with some graphical inconsistency: Compare yius (ὑιούς; sons; f. 19r) with kyrios, kirios, and quirios (κύριος; Lord), all recorded on the very same page, f. 31v.

The characteristics that we have just listed make it extremely unlikely that this transcription was originally obtained by transliterating a written, Greek apograph. It is far more reasonable to assume that these texts were originally transcribed by someone who actively spoke Greek and therefore was capable of reading out loud the text of the Septuagint. Alternatively, we cannot entirely rule out the possibility that the original transcription was carried out indirectly, that is, by listening to someone capable of reading the Greek Bible (and clearly well acquainted with their contemporary Greek language) and writing down the sounds in the Latin alphabet, with an inevitable degree of approximation. Either way, since these transcriptions (or anything that may closely resemble them) are only found in this manuscript of undisputed Welsh origin, the logical consequence of what has been exposed so far is that, in the 9th century (possibly, earlier), some monks in Wales were being taught (or, at the very least, were being exposed to) Greek by at least one Greek-speaking individual and that biblical excerpts were being used to that end as pedagogical tools – tools that were destined to the further education of those monks who had already acquired a working competence in Classical Latin, as testified by the Liber Commonei itself.