INTRODUCTION

The Etymologiae of Isidore of Seville (d. 636) was the most important medieval Latin encyclopaedic work. In the early Middle Ages, in particular, it was the only widely known encyclopaedic work in the Latin West, attaining notable popularity documented by the survival of more than 500 early medieval manuscripts transmitting this text in full or partially. Approximately fifteen of the surviving pre-1000 codices of the Etymologiae contain at least 100 glosses, while additional codices transmit smaller amounts of marginalia. Altogether, the Etymologiae attracted over 7,000 glosses in the first three hundred years of its circulation (Steinova 2020).

Glosses do not appear uniformly across Isidore’s encyclopaedia covering diverse topics, from the Seven Liberal Arts, medicine and law, to precious stones, insects, games, and footwear. They rather tend to be concentrated in specific sections, indicating that the glossing of the Etymologiae was topic-driven rather than a matter of continuous reading or study. The concentration of glosses in one or more of the twenty books, into which the Etymologiae are divided reveals uncannily what specific interests animated glossators in different places, times, and contexts. In this manner, we can see that the early medieval annotators were most often interested in the book I of the Etymologiae dedicated to the discipline of grammar – it received approximately 4,300 glosses or almost two-thirds of the total attested in the early Middle Ages. The following digital scholarly edition is an attempt to critically present the corpus of these early medieval annotations to the first book of the Etymologiae. It accounts for 54 annotated manuscripts of the first book of the Etymologiae, containing anything from one to over 750 glosses.

While glosses to other books appear almost always in isolation, being restricted to a single manuscript or several textually related codices, many of the glosses to the first book can be found in multiple manuscripts suggesting a degree of circulation. Indeed, when the patterns of gloss parallelism between manuscripts are probed, one can discover clusters – small collections of glosses that seem to have been transmitted together. More importantly, manuscripts that share glosses are often not related philologically. This transmission pattern is consistent with what we know about the use of the first book of the Etymologiae in Carolingian schools. It suggests that many glosses to this book were a product of school engagement. Their circulation seems, at least to some extent, to reflect the relationships between schools in the Carolingian realm and the broader exchange of books, people, and teaching material between various early medieval institutions.

WHAT IS A GLOSS?

While it may seem that glosses do not need to be defined, the material presented in this edition poses many problems well-known to those studying medieval marginalia. Above all, the term 'gloss' is functional, not a descriptive category, as it implies an interpretative relationship to the main text (i.e., a gloss comments on, explains, or renders a passage in a different language). What counts as an interpretation of a text, and therefore what should be considered a gloss (rather than, say, a correction, variant reading or addition) is sometimes not self-evident.

In their appearance, glosses are often indistinguishable from other phenomena inhabiting the margins of medieval manuscripts: corrections, variant readings, marginal tabs/summaries/indices, annotation symbols, secondary additions and interpolations. Medieval annotators rarely thought of their activity in the same fashion as modern scholars do. They were most often motivated by making a certain manuscript more useful or improving its contents rather than by leaving behind a commentary for posterity. Indeed, medieval annotators may not have wished to distinguish between glosses and other categories of marginalia. This certainly seems to be the rule in the manuscripts of the Etymologiae, in which glosses appear indistinguishable from other types of marginalia in palaeographically homogeneous layers. They often lack the telltale attributes, such as being introduced by formulaic abbreviations: i. for id est, s. for scilicet, and sb. or sub. for subaudi. The making of this edition, thus, entailed making certain decisions about which textual elements appearing in the margin to interpret as glosses and which categories of marginalia to include or exclude.

While corrections, variant readings and marginal tabs/summaries/indices were excluded, it was not always possible to distinguish these categories of marginalia from glosses. The distinction sometimes depends on the shape of the glossed text and, therefore, on the decisions made by its editors, as can be gleaned from the comparison of Lindsay's 1911 and Spevak's 2020 editions of book I of the Etymologiae. To give just one example, the text of Etym. I 7.2 in Lindsay's edition reads: Cognomen, quia nomini coniungitur, ut Scipio; while Spevak's edition rather reads: Cognomen, quia nomini coniungitur, ut Scipio Africanus. Indeed, some of the manuscripts examined for this edition add the word Africanus above the word Scipio. If one chooses Lindsay's edition as a base text, we can understand this addition as a gloss. However, from the perspective of Spevak's edition, it is rather a correction. Moreover, even when making this distinction, we should remember that it would probably make little sense to a medieval scribe.

A different problem is posed by the tendency of marginalia to move from the margins into the main text block and blend into the main text. This 'gloss drift' is particularly pronounced in the corpus of the glosses to the Etymologiae, as can be demonstrated on the example of the famous Scholia Vallicelliana (which, however, do not appear in book I and therefore do not feature in this edition). As we know today, this set of annotations to the Etymologiae began its life as personal annotations made by the great Carolingian scholar Paul the Deacon (Villa 1984). They survive as marginalia in a single 12th-century manuscript: Rome, Biblioteca Vallicelliana, A 18. Some of the scholia are also preserved in earlier manuscripts, however, only as interpolations integrated into the main text of the Etymologiae and only extremely fragmented. If it were not for the Vallicelli manuscript, we would lack crucial insights into their nature and origin, and we would probably not think of them as fossilised annotations.

In other cases, we have no way to determine whether particular additions to the text of the Etymologiae began their existence as annotations that made it into the main text block due to the 'gloss drift', or as interpolations inserted in the margin because of the lack of space. Such a distinction may, in fact, be superficial as it was not relevant to medieval readers of the Etymologiae. This is the case with passages from the anonymous treatise De vitiis that appear in some manuscripts as marginalia and in others as integrated interpolations in Etym. I 34 and 36. It is also true for some of the lengthier annotations/interpolations that appear in manuscripts both in the margins and integrated into the main text (e.g., the commentary on Phoenicians in Etym. I 3.5, and the addition of a perfective aspect of a verb in Etym. I 9.3).

In other cases, we find that the text of certain manuscripts contains variants that resemble glosses in their form because they are introduced by the typical formulas (id est, scilicet) or because they provide a word-for-word explanation of a preceding word or phrase. While they do not appear as marginalia in any of the early medieval manuscripts, the fact that they survive today integrated into the main text in a small number of manuscripts provides a good reason to think that these had been glosses that moved from the margin into the main text and survive only in this integrated form. Two identifiable examples include the glosses id est cola et commata to Etym. I 39.2 and hoc est fortium to Etym. I 39.10.

Frequently, fossilisation in the main text indicates the old age of integrated glosses, which originated in the centuries from which no or few witnesses survive. In this regard, it can be pointed out that the edited text of the first book of the Etymologiae contains several passages that resemble integrated glosses in form (they are introduced by the typical formulas) and function (they explain a preceding word or phrase). We should, therefore, be open to the possibility that these passages may not be original to Isidore but rather may be cases of very old fossilised glosses that became embedded so early in the text of Isidore's encyclopaedia that they now survive in most or almost all of the oldest copies of the Etymologiae.

NETWORK AS A MODEL FOR A DIGITAL SCHOLARLY EDITION

This edition is scholarly and critical in the sense that it attempts to present the corpus of early medieval glosses to the first book of the Etymologiae systematically, in its entirety, and provides a critical framework through which it can be read, analysed, and engaged with. The critical framework, in this case, is not based on the genealogical method connected with the name of the German philologist Karl Lachmann. The fluid, non-linear, and polygenetic character of the glosses and the complex nature of their transmission (not only by copying from an exemplar to a copy, but also by collection, orally and perhaps after being long retained in memory) do not allow for it. The users of this edition will, thus, note that the edited text of the glosses is not accompanied by an apparatus criticus. Rather, the editor of this corpus chose an alternative critical framework for the presentation of the material, namely as a network that represents the patterns of relationship and similarity among the witnesses. In this framework, several entities (manuscript witnesses, clusters of glosses transmitted together, and chapters of the first book of the Etymologiae) appear as nodes. Glosses that are common to multiple of these entities appear as edges that connect them. In this manner, the users of the edition can access the corpus from several different angles and simultaneously view it in its entirety and dissected into its many constituent entities (clusters, layers of glosses in particular witnesses, and individual glosses).

The three network visualisation modes around which this edition is built can be activated through the 'Networks' section following this introduction. The users are invited to engage with the edition not only by browsing the chapters and glosses in the left and right panels but also through adjusting the network visualisations and right-clicking on their constituent elements (nodes and edges). The network visualisation modes have three functions:

• to visualise the relationships between the text of the book I of the Etymologiae, layers of annotation as they survive in manuscripts, and clusters of glosses as they may have been transmitted in the Middle Ages;
• to provide access to the complex and non-linear material, since both the nodes and edges displayed can be clicked on to open a corresponding textual entity; and
• to guide the analysis of the edited material with the help of filters and layouts.

THE TEXT DISPLAY

The glosses are presented against a base text provided by the critical edition of the Etymologiae of Wallace Martin Lindsay as reproduced by The Latin Library:

• W. M. Lindsay, Etymologiarum sive Originum libri XX, 2 vols. (Oxford, 1911).

The Latin Library digital text was corrected for small errors. In addition, some corrections, punctuation and layout, and the overview of sources of the main text were adopted from the more recent edition of Olga Spevak:

• O. Spevak, Isidorus Hispalensis. Etymologiae I/Isidore de Séville. Étymologies, livre I: La grammaire (Paris, 2020).

The text of the Etymologiae is displayed in the left panel of the edition’s visual interface. It is divided into 44 chapters, following the division in Lindsay's edition. Words and passages that were glossed (lemmata) are highlighted in the base text and equipped with a superscript g. They can be clicked on to open the corresponding panel with glosses in the right panel of the edition.

As Lindsay's (rather than more recent Spevak's) edition of the Etymologiae was chosen as a basis for this edition, the edition retains square brackets where Lindsay placed them to mark additions proper to certain families. Moreover, it also contains additional passages presented in square brackets not found in Lindsay’s text. These correspond to readings found only in certain witnesses. These readings are represented in the edition only as far as these readings attracted glosses. If a manuscript provides a variant reading relevant for the edition for which a parallel exists in the text of Lindsay's edition or several manuscripts contain variant readings to the same passage, these readings all appear in the same square brackets separated by //. A special case is Etym. I 17.7 (metrical foot anapest), where I chose to include several variant readings even if they do not contain glosses. The edition users can see whether a particular passage appearing in square brackets is presented thus in Lindsay's edition or appears in specific witnesses examined in this digital edition by hovering their mouse above a given passage.

Words and passages presented in italics correspond to citations and examples taken from sources. These sources can be viewed by hovering with a mouse above the words rendered in italics.

THE GLOSS DISPLAY

The glosses, which are the main subject of this digital edition, are displayed in the right panel by clicking on one of the two 'Glosses' tabs at the top of it. They can be viewed either as they appear in all of the witnesses to the text of a specific chapter ('Glosses (chapter)') or as they appear across all chapters in a specific witness ('Glosses (manuscripts)'). Moreover, the users can change which glosses are displayed in the right panel of the edition by adjusting the filter in the 'Glosses (chapter)' tab and choosing to remove the isolated glosses (see below).

Each gloss is associated with a lemma that is provided with a unique identifier. This identifier can be used to refer back to both the lemmata and glosses. The lemmata are displayed next to the identifier in red, while the glosses attached to this lemma appear underneath it, preceded by the siglum of a manuscript in which they appear. Notes on glosses appear in the rightmost area of the gloss tab. The glosses displayed are either isolated (i.e., they appear only in a single manuscript) or shared (i.e., they appear in multiple manuscripts). If shared, glosses are assigned to one of the 34 clusters identified by a letter siglum and a colour and given a weight (displayed as w=1-4). The weight of a gloss reflects its (non-)triviality:

• w=1: a very trivial gloss, which could have been and likely was coined multiple times independently (e.g., completes an ellipsis in a sentence);
• w=2: less trivial gloss, which could have been coined multiple times independently, but less likely than a gloss with the weight 1;
• w=3: non-trivial gloss, which is unlikely to have been coined multiple times, but rather its appearance among shared glosses points to a transmission;
• w=4: very non-trivial gloss, which cannot be considered coined multiple times and therefore clearly reflects a true philological relationship.

Since the orthography of the glosses to the first book of the Etymologiae is inconsistent, it was standardised in this edition using Lewis's and Short's Latin Dictionary (Oxford 1879) as available on the Perseus Digital Library. The standardisation was motivated primarily by the need for automated comparison of glosses in the context of this editorial project. Hopefully, it will benefit those who would like to use the raw data behind this edition for future automated processing. For the same reason of inconsistency, all abbreviations were expanded, including the formulaic id est, scilicet, and subaudi.

The edition further uses the following conventions:

• Glosses in Tironian notes are rendered in the Courier font and followed by a stylus icon: Tironian notes
• Dry-point glosses are rendered in the Times font and followed by a stylus icon: Dry-point glosses
• Glosses in a vernacular language are followed by a language icon: (the language is displayed upon mouseover).
• Glosses with identified sources are followed by a book icon: (the title of the source is displayed upon mouse-over)
• [...] Unreadable, cannot be reconstructed
• [ ] Unreadable, reconstructed
• < > Editorial emendation
• Corrections in a manuscript are indicated using the stylus icon (the reading before the correction is displayed upon mouseover).
• Peculiar readings in a manuscript emended because they represent scribal errors are indicated using the stylus icon (the reading of the manuscript is displayed upon mouseover).
• Peculiar readings in a manuscript retained due to their significance are indicated using the stylus icon (sic is displayed upon mouseover).

CLUSTERS OF GLOSSES

Approximately 1,700 of the more than 4,300 early medieval glosses to the first book of the Etymologiae are shared glosses, that is they appear in multiple witnesses. In most cases, these glosses are shared only by two manuscripts, but sometimes the same gloss can be found in as many as nine or ten manuscripts (e.g., L39.10.7). Usually, glosses are not shared as individual entities but as layers, suggesting that glosses were transmitted from manuscript to manuscript as sets. For this reason, this edition talks about clusters: sets of glosses that appear in multiple manuscripts and may correspond to historical artefacts of transmission. In this edition, 34 such clusters are identified, falling into four groups:

• Clusters of high importance having a weight larger than 20: A-B, D-I, M-Q, S (19 clusters)
• Clusters of medium importance having a weight larger than 10 but smaller than 20: R, T-Z (6 clusters)
• Clusters of special interest representing single glosses or small gloss sets of particular philological value (and therefore given high ranks 3-4): C1-C7 (7 clusters)
• Generic clusters X1 and X2, to which the remaining shared glosses not belonging to any of the other clusters were assigned because their weight is too low to be considered a meaningful indicator of relationships between manuscripts

The users of this edition can note that clusters F, H, I, N, and S have an assigned numeral 1 or 2 (e.g., F1 and F2). This numeral is used to express the mutual closeness of the layers of annotation in the damaged manuscripts Orleans296 (containing only chapters 22-44) and Paris7490 (containing only chapters 5-17). The two manuscripts were probably copied in the same scriptorium. Certain codicological indications (e.g., identical page dimensions, layout and ruling patterns) suggest that the two damaged manuscripts may have once been part of a single codicological entity. However, they were not copied by the same hand and, thus, cannot be readily considered two separate parts of a single medieval codex. Nevertheless, the layers of glossing in both manuscripts are closely related, both textually and palaeographically. There are, thus, reasons to consider them a single layer, as is also indicated from the proximity of the two manuscripts in the network graphs 2 and 3. In the context of this edition, the two (sub-)clusters reflecting glosses shared by Orleans 296 and other manuscripts (represented by F1, H1, I1, N1, and S1) and those shared by Paris Lat. 7490 and other manuscripts (represented by F2, H2, I2, N2, and S2) are assigned the same siglum (to express their closeness), but distinguished by the assigned numerals 1 and 2 (to allow for their separation and filtering).

More can be said about the following clusters:

Cluster F1

138 glosses, weight: 348 (avg. gloss weight 2.36)

Main witnesses: Harley3941 (H3941), Orleans296 (O296)

Associated witnesses: Reims426 (R426)

Associated clusters: F2

Chapter range: mostly 36-44, to a lesser extent in 22-24, 26, 28-29, and 33 (Orleans296 cropped before chapter 22)

Terminus ante quem: before 850 (because of the age of glosses in Orleans296 and Reims426), possibly early ninth century or earlier (as suggested by the system of abbreviations and the corruptions in Orleans296 and Harley3941)

Region of origin: Continent? Traces of insular abbreviations suggest transfer via a centre under insular influence or perhaps an older insular core

Provenance: Fleury?

The only cluster consisting of more than 100 glosses, F1 is the most significant set of annotations to the first book of the Etymologiae surviving from the early Middle Ages. Many of the glosses from this cluster are long, complex, and rich in information (e.g., explaining the narrative context of verses from the Aeneid cited by Isidore), making it clear that the cluster must have been copied and therefore transmitted genealogically.

Layers of glosses in the two main witnesses of F1 could not be copied from one another, as is suggested by the fact that both Harley3941 and Orleans296 feature omissions with regard to each other:

H has an omission/ O full reading	L37.6.12, L37.23.6, L37.29.2 + L37.17.2 (-tur dropped)
O has an omission/ H full reading	L37.18.6, L37.24.9, L38.1.5, L39.13.7 + L38.2.2 (-tur dropped)

In some cases, Harley3941 offers a better reading (L35.7.11, L37.35.4), in other cases, it is Orleans296 (L36.7.4, L37.9.6, L44.5.1), and in two cases, the two manuscripts contain different readings, which cannot be reconciled:

L37.20.4	H voluptas/O voluntas (to cupido)
L39.14.3	H qui flent et lugent/O qui flent vel lugent (to miseris, due to the script?)

The two main witnesses also share several problematic passages that show that they share an ancestor copied in an unfamiliar abbreviation system:

HO	Vergilius dixit de Polyphemo quando persequebatur Aeneam et illefugiens in profundum maris (L37.16.4 O/L37.17.2 H)
HO	ideo dura quia timens H/times O (L37.27.1 H/L37.27.4 O)
L37.11.9	H progressus/O congressus
L37.32.3	Vergilius dixit de illo viro quem misit Iovis (p.c. OH) (rei H) ad Aeneam …
L38.1.5	in illo libro quem scripsit de Plautino H/Plautone O
L43.1.7	H imposuerunt/O composuerunt

The unfamiliarity of the abbreviation system is further indicated by the unusual abbreviation qm with a suspension stroke for quando in Orleans296 (L36.15.2, L37.17.2, L37.29.2, L37.35.17). This abbreviation appears in MilanL99sup, an important eighth-century manuscript of the Etymologiae copied in northern Italy in an insular environment. In the Milanese codex, the pre-Carolingian quando (qo) was reinterpreted as quoniam and abbreviated by the insular suspension qm (Lindsay 1915, p. 220). Moreover, the same abbreviation qm, standing for quando, also appears in glosses to the first book of the Etymologiae in manuscripts with a connection to Mainz (Laon447, Schaffhausen42), an important continental centre with an insular connection.

The palaeographic clues, nevertheless, do not prove that F1 is of insular origin. They perhaps rather indicate that the glosses passed through a centre which was influenced by insular copying practices or that there was an insular core or elements to F1. Two elements speak, on the contrary, in favour of a continental origin of F1. First, Harley3941 contains a notable error in glosses that are partially copied in Tironian notes in Orleans296, and stem perhaps from a misunderstanding of Tironian notes that were present in the common ancestor of the two manuscripts:

L36.10.6	O subaudi mortem/H sub morte (to ire per umbras)
L37.6.8	O scilicet catachresis/ H species catachresis (to haec)

Second, F1 contains one early Romance gloss cinciala/cinicella (L36.7.4). Taking into account only the evidence presented by F1 (and discounting the bearing of F2), we should perhaps imagine that F1 was constituted in a continental centre influenced both by insular and continental habits (however, note what is said about the possible insular origin of F2 below). Since O was present at Fleury by the tenth century and since Fleury maintained relations with Brittany, serving as an important transmission hub of material from northern France to Brittany, there are good reasons to think that Fleury also played a role in the circulation of F1. It is, for example, possible that the common ancestor of Orleans296 and Harley3941 was present at Fleury, an old centre with a strong insular presence in pre-Carolingian times.

The relationship of Reims426 to Harley3941 and Orleans296 is difficult to ascertain, given the small number of glosses it shares with both manuscripts (13 glosses). In one case, it seems to contain a better reading than either of the two main manuscripts:

L37.22.9

HO exceptiones/R ex<c>erptiones (to eglogae)

In three cases, Harley3941 and Reims426 offer a better reading against Orleans296:

L36.13.3	HR nubes/O parva nubis (to nubila)
L36.13.5	HR temepestates/O tempestas (to procellae)
L37.15.13	HR rubicundum/O rubeum (to croceum)

Since Orleans296 is cropped before chapter 22, it is possible and, indeed, likely, that cluster F extended beyond the current attested range, perhaps covering all chapters of book I. Some of the now isolated glosses in Harley3941 were probably echoed in Orleans296 before it had been damaged. For example, there are several longer glosses providing a narrative context to the citations from the Aeneid with the same tenor as those in F1 appearing only in Harley3941 (L21.18.2, L21.18.4). For this reason, too, it is notable that Harley3941 shares many heavy glosses with Paris7490, a loose quire containing chapters 5-17, including some which resemble glosses from F1 in their focus and tenor. As noted above, the relationship between Paris Lat. 7490 and Harley3941 may indicate that glosses in Paris7490 and Orleans296 represent two segments of the same layer.

Cluster F2

19 glosses, weight: 52 (avg. gloss 2.47)

Main witnesses: Harley3941 (H3941), Paris7490 (P7490)

Associated witnesses: Reims426 (R426)

Associated clusters: F1, C3

Chapter range: 16-17 (because Paris7490 contains only 5-17, and H contains no glosses to 5-15)

Terminus ante quem: before 850 (because of the age of glosses in Paris7490), possibly early ninth century or before 800 (given the problems stemming from transcription from an exemplar in abbreviated script)

Region of origin: Continent? Traces of insular abbreviations suggest transfer via a centre under insular influence or perhaps an older insular core

Although cluster F2 contains only 19 glosses, it is easily discernible because of the concentration of these glosses in a few chapters and their heaviness. While Paris7490 is perhaps half a century older than Harley3941, the latter contains overwhelmingly better readings than the former. Paris7490 can be shown to be corrupt in all but one case of disagreement with Harley3941. In two cases, Harley3941 offers a better and longer text of glosses than Paris7490 (L17.15.1, L17.22.5) and in case (L17.2.3), Paris7490 lost the final -tur (Harley3941 lost the final -tur in L17.4.1). In one notable case, Paris7490 contains a corruption that seems to reflect copying from an exemplar written in an unfamiliar script:

L17.2.2

H quasi certator ludicus, vel … nominatus/P [qua]si cantator, vel … nominatur (to Pyrrichius)

The lectio difficilior in this case is certator ludicus. Moreover, the combination of er can be incorrectly interpreted as an if copied from an insular exemplar but not vice versa (and the same applies to the confusion between a final s and a final r, which also indicates an exemplar in insular script).

In addition, Harley3941 and Paris7490 disagree in the case of the old gloss to the Pyrrhic (C3) found also in Aubervilliers342, VLF82 and Wolfenbuttel64):

L17.2.2	H ab igne appellatur, quia ignis Graece pyr dicitur
	P ab igne appellabatur, qui Graece pyr dicitur
	A ab igne appellatur, qui Graece pyr dicitur
	L ab ab igne appellatur, qui[...] pyr dicitur
	W ab igne appellatur, qui ignis Graece pyr dicitur

In one case, Paris7490 and Harley3941 contain different readings, neither of which can be immediately judged better, although in Harley3941 it seems to be more complete:

L17.5.2

H … choris aptissimum. Similiter choriambus/P choris aptissimum canendi

Overall, one can say that Harley3941, even if more recent, preserves consistently better readings of F2, although this may also be an impression due to the limited overlap between Harley3941 and Paris7490 (only in chapters 16-17).

Paris7490 shows a relationship not only with Harley3941 but also with several other manuscripts. The character of the glosses it shares with VLO41, Paris7670, and Reims426 in chapters 5-17 and the general pattern of sharing and layout (e.g. the use of particular signes de renvoi) is surprisingly similar to the one that Orleans296 displays with the same four manuscripts (i.e., Harley3941, Leiden VLO 41, Paris Lat. 7670, and Reims 426). As noted above, given that both Paris7490 and Orleans296 were assigned by Bischoff to the same region, given the paleographic similarities in both glosses and the main script, and the pattern of sharing of glosses, it seems possible that Paris7490 (containing only chapters 5-17) and Orleans296 (containing only chapters 22-44) come from the same scriptorium and may have been glossed by a single group of annotators.

Cluster E

50 glosses, weight: 156 (avg. gloss weight 3.07)

Main witnesses: GothaI147 (G147), Paris7585 (P7585)

Associated witnesses: Aubervilliers342 (A342), Queen320 (Q320), CottonCaligulaAxv (CCAxv)

Chapter range: 21, 27, 32, 34-35, 37-40 (due to the cropping of GothaI147)

Terminus ante quem: certainly before 850 (provided by the date of copying and annotation of GothaI147), possibly early ninth century or before 800 (as is indicated by the abbreviation system and integrated glosses in Paris7585)

Region of origin: British Isles, perhaps Ireland (as is indicated by the presence of Old Irish glosses in GothaI147 and Paris7585, the extensive abbreviation in these two manuscripts, and the abbreviation system in Paris7585)

Provenance: E has a strong connection to tenth-century Canterbury. Two of the witnesses (Paris7585 and CCAxxv) were certainly annotated there, and one witness (Queen320) was likely glossed there

Cluster E is the second-best attested cluster of early medieval glosses of the Etymologiae. It rivals clusters F1/F2 and G in sophistication, richness, and informativeness and contains some notably original annotations (e.g., L37.29.6 and L40.6.13) just as these two other important clusters. While it can be attested today only across several of the chapters of the book I of the Etymologiae due to the fragmentary preservation of GothaI147 (once a manuscript of the entire Etymologiae or the first ten books of this work), it most likely spanned the entire book I. The associated witnesses Aubervilliers342, CottonCaligulaAxv and Queen320 preserve glosses that belong to cluster E in chapters missing from GothaI147 (21, 27, 32, 34-36, and 43). In fact, unlike F1/F2, cluster E may have extended to all books of the Etymologiae, as Aubervilliers342, Paris7585 and the sister-fragments of GothaI147 contain annotations to other books. In their layers of glosses to books II-XX, these three witnesses show a degree of closeness akin to, although not as strong as, GothaI147 and Paris7585 in book I. Cluster E seems to be most fully preserved in Paris7585 today. It is, therefore, likely that some of its roughly 120 isolated glosses to book I and, as long as we accept that E was a set of glosses to the entire Etymologiae, some of the over 500 glosses in other books belong to E as well.

A characteristic element of E is the marking of proper names with proprium (nomen) (abbreviated in both Paris7585 and GothaI147, often to pro, see e.g., L37.34.2, L38.2.6, L39.7.8, and L39.17.4), names of the cities with civitas/nomen civitatis (usually abbreviated in GothaI147, Paris7585 and Aubervilliers342 as civit, civi or cive, see e.g., L37.34.1, L27.34.3, L39.16.2, and L39.17.5), and of various names in general (e.g., L27.29.13, L35.5.3, L37.35.10, and L40.6.8). For this reason, the most likely candidates for glosses belonging to E are some of the isolated glosses in Paris7585 reading proprium (nomen) (L3.6.6, L3.6.10, L36.2.6, L36.2.11, L36.7.5, L36.7.6, and L36.14.10), civitas/nomen civitatis (L3.7.1, L33.2.2, and L39.16.6), or identifying the names of various objects, places, and people (L21.14.5, L32.5.7, L36.17.5, and L37.6.11).

Glosses belonging to E were inserted into both main witnesses by copyists. As they are very heavy (avg. 3.07), we need to think of them as a set that was copied from an exemplar to a copy, just like F1/F2 and G, and, therefore, transmitted for at least some time. In Paris7585, a manuscript copied at Canterbury in the second half of the tenth century, the glossed sections of book I contain many integrated glosses, which may have already been integrated in the exemplar, a sign of their potential antiquity (two of the integrated glosses, L37.29.6 and L40.7.11, are even marked as glossa by a later medieval user). These sections and the glosses, both integrated and marginal, are copied in a script rich in insular Irish abbreviations, oddly out of place in tenth-century England. One of the glosses belonging to E is also integrated in GothaI147 (L39.16.10). Glosses in both Paris7585 and GothaI147 are often extremely abbreviated, which makes their resolution difficult (e.g., L37.28.5, L37.34.1, L39.16.6, and L39.17.4), another sign of potential early insular origin. Both Paris7585 and GothaI147 contain a single stray Old Irish gloss: GothaI147 in L39.18.3 and Paris7585 in book III to poetis in chapter 15 (glossed as filidbus, presumably a Latinization of filid or an incorrect resolution as filiabus, a word that appears later in the same chapter). These clues imply that the exemplar of Paris7585 was most likely an Irish annotated manuscript. For this reason, it is tempting to situate E’s origin in Ireland, perhaps even before 800. However, it also needs to be pointed out that some of the isolated glosses in Paris7585 came from a lost insular commentary on Donatus’s Ars maior that was used in the 840s by Murethach and Sedulius Scottus, but which was presumably not compiled before 800. If it can be assumed that they belong to E, the cluster must have originated in the first half of the ninth century.

How glosses from this cluster ended up in the Breton GothaI147 is difficult to ascertain, especially given that the most recent research indicates that most of the Irish material preserved in Brittany reached this region via northern France and not directly from Ireland or England. If cluster E was available in northern France in the first half of the ninth century, no traces of this sojourn survive.

The three associated witnesses, Aubervilliers342, Queen320 and CottonCaligulaAxv (the latter glossed only chapter 21), contain only a small selection of glosses from E. All three manuscripts were annotated several centuries after their production. Two, Queen320 and CottonCaligulaAxv, have a link to Canterbury just as Paris7585. The eighth-century continental CottonCaligulaAxv can be shown to have been present and glossed in Canterbury in the tenth century. The tenth-century Queen320 may have been produced at Canterbury. While it is not certain that it was still present there in the late eleventh or the early twelfth century when it was annotated, the glosses may be considered evidence to this effect. In both the cases of CottonCaligulaAxv and Queen320, it is possible that the glosses were copied from Paris7585 or its exemplar (as is also indicated by cluster G). The situation of Aubervilliers342 is complicated, as this is a continental Frankish manuscript kept at the Abbey of St. Martin in Tournai in the thirteenth century. However, the glosses as well as K-shaped kaput signs inserted in the margins of this codex (which also appear in Paris7585) suggest that it may have been annotated by an English (or a Breton?) annotator before it reached Tournai, perhaps during the eleventh century.

Both main witnesses of E contain vernacular glosses. Apart from the one Old Irish gloss in each manuscript, Paris7585 contains five Old English glosses, including one to book I (L39.4.8). GothaI147 contains three Breton glosses, while its sister fragments contain additional 12 Breton glosses. The lemmata from book I glossed in Breton in GothaI147 are left unglossed in Paris7585. The single exception is the singular Old English gloss in book I of Paris7585:

L39.4.8

G id est huelim / P taesan (to carminare)

As the stray Old Irish gloss in Paris7585 suggests, the exemplar of Paris7585 (and of GothaI147?) likely contained vernacular glosses, but these were not copied into Paris7585 (the single Old Irish gloss in this manuscript survived due to a misinterpretation). These were not the only glosses omitted in the process of copying. The extensive abbreviation of glosses sometimes seems to have caused problems to the two scribes copying glosses in Paris7585, and GothaI147 and the abbreviated glosses or gloss elements were dropped, as is indicated by this gloss in Paris7585:

L39.17.4

G proprium bri / P proprium nomen

In another case, only GothaI147 contains an extremely abbreviated gloss (L37.28.5). GothaI147 also contains several glosses proprium that do not have a parallel in Paris7585 (L37.34.4, L39.7.4, and L39.26.2), which may also be due to the extreme abbreviation. In two cases, Paris7585 and GothaI147 offer different readings, which perhaps have to do with a distinct resolution of extremely abbreviated glosses:

L39.12.2	G Tales proprium / P Talesius
L39.17.5	G civitas / P civitatis (compare with the abbreviation of various forms of this word as civi)

Overall, there is only one gloss in GothaI147 (L39.10.2), which does not appear in Paris7585 or is not a case of gloss likely to be purged (i.e., because it was vernacular or extremely abbreviated). The agreement between the two main witnesses of E is exceptionally strong.

Clusters I1 and I2

54 glosses, weight: 101 (avg. gloss weight 1.84)

Main witnesses: VLO41 (L41), Orleans296 (O296), Paris7490 (P7490)

Chapter range: 6-9, 12-17 (I2), 22-42 (I1)

Terminus ante quem: 9th century, 4/4 (the date of VLO41)

Region of origin: VLO41 was produced and annotated at Fleury, and since Orleans296 also has a connection to Fleury (and Paris7490 seems to have been produced and annotated in the same centre as Orleans296), an origin in Fleury is probable

Taken on its own, cluster I1 alone is the third heaviest cluster detected in the early medieval manuscripts of the Etymologiae (after F1/F2 and E). When I1 and I2 are considered to reflect a single layer of glosses, they are one of only three clusters (together with F1/F2 and E) to have a weight of over 100. However, clusters I1 and I2 are very different from both these and other notable sets of annotations (e.g., G). Unlike other clusters, I1 and I2 are not visible to the naked eye of a philologue. They are constituted only by glosses of the low weights 1 and 2, not particularly concentrated in any chapter, and don’t have a notable tenor, focus, or regional element. I1 and I2 are examples of what may be considered an ‘invisible cluster’, i.e., a set of annotations that come to one’s attention only with the aid of quantitative methods. Yet, they cannot be dismissed as a phantom conjured by these methods, given the significant weight of I1/I2 and the palaeographic proximity of the two witnesses VLO41 and Orleans296 (and probably also Paris7490). These two (or three) witnesses are connected with Fleury, where both may have been annotated in the span of roughly a century: Orleans296/Paris7490 at the beginning of the ninth and VLO41 at the end of the ninth or the beginning of the tenth centuries. A plausible explanation for the subtleness of this cluster is, therefore, that the layers of glosses in the two manuscripts are not related 'truly philologically', as is the case of glosses in clusters F1/F2, E and G, nor is the younger VLO41 annotated because glosses were transfer to its from the older Orleans296/Paris7490. Rather, the particular pattern characterising I1/I2 (and some of the lighter clusters) may be a vestige of a living oral and mnemonic tradition of exposition of the first book of the Etymologiae at Fleury (in the context of grammatical teaching?). Indeed, it can be shown that this book was used to study and teach grammar at Fleury to a significant degree in the course of the ninth and tenth centuries, as five surviving early medieval grammatical collections transmitting book I of the Etymologiae are connected to this centre. Traces of the same exposition that appears in the form of glosses in Orleans296/Paris7490 can be still detected at Fleury as late as in the late tenth century via the person of Abbo of Flery, who brought some of the Fleury Isidorean lore to Ramsey in the early eleventh century and bequethed it to his English student, Byrhtferth of Ramsey. The scenario that I1/I2 reflects a transmission as facilitated by oral and mnemonic processes is thus plausible.

There are very few disagreements between VLO41 and Orleans296/Paris7490 that could reveal anything about their mutual relationship. In eight cases, Orleans296 or Paris7490 contains longer glosses than VLO41:

L9.3.2	P instruent vel ostendunt / V et instruant
L22.1.9	O iuvenis vel novellus miles / V vocatur novus miles
L22.2.11	O designent / V signent
L36.5.2	O dulcis, matura / V vocat<ur> matura
L37.15.3	O utitur vel trahit / V trahit
L39.13.7	O habitatores vel <qui> colunt terram (shared with Harley3941) / V habitatores
L39.19.1	O fletum, luctum / V id est luctum
L42.2.4	O viguit, aparuit / V viguit

Since the layers of glosses in Orleans296/Paris7490 are several decades older than those in VLO41, one could expect that they would not become shorter or lose constituent elements. However, such shortening can be explained if copying was not the main mode of transmission, as should be suspected here. In four cases, VLO41 has longer and richer glosses than Orleans296/Paris7490:

L12.1.7	P in coniunctione / V id est con[iun]ctione partium
L31.1.8	O separantur / V id est separantur et dividuntur (shared with Reims426)
L37.2.3	O operiuntur / V id est operiuntur et velantur
L37.12.2	O sancta / V id est sancta, candida

Other variance between Orleans296/Paris7490 and VLO41 (e.g., in L33.2.3) seems to be of little importance. Minor variance seems also to be due to the use of Tironian notes in Orleans296/Paris7490 (L7.1.2, L25.2.1). No genuine Tironian notes occur in VLO41 (although this manuscript contains an abbreviation symbol for scilicet that resembles a Tironian note and may have been derived from a Tironian note).

Cluster G

30 glosses, weight: 78 (avg. gloss weight 2.47)

Main witnesses: Harley3941 (H3941), Paris7585 (P7585)

Associated witnesses: Aubervilliers342 (A342), Queen320 (Q320), CottonCaligulaAxv (CCAx)

Chapter range: 2-3, 17, 20-21, 26-27, 29-30, 32, 34-37, 39-40

Terminus ante quem: 10th century, ex. (the date of Paris7585 and Harley3941)

Provenance: Like E, G has a strong connection to tenth-century Canterbury. Two of the witnesses (Paris7585 and CottonCaligulaAxv) were certainly annotated there, and one witness (Queen320) was likely glossed there

Cluster G is the second well-attested cluster connected with the insular world (the other being E). It provides some highly original explanations, such as an etymology of the sicilicus derived from the island of Sicily (L27.29.9), and a clarifying comment explaining that God alone is capable of transforming matter into nothingness (L36.17.4). Like E, this cluster shows interest in explaining the verses from the Aeneid included in the first book of the Etymologiae. The fact that glosses from this cluster belong to a set is confirmed by Harley3941, in which most of them are copied by a single later hand (as opposed to other glosses, including those belonging to cluster E).

Of the two main witnesses of this cluster, Harley3941 transmits abbreviated readings as can be glanced from comparison with other witnesses of G:

L27.29.9	PQ id est nomen ponderis et figura (figurae Q) et sex scrupulos habet (habet sex scrupulos Q) / H nomen ponderis
L27.29.9	PA et sicilicus quia in Sicilia invenitur (inveniebatur A) primo / H sicilicum in Sicilia inventum est
L34.5.5	P pro levatis. Hic pertinet ad genitale / H pro levatis (to arrecti)
L37.24.12	PA id est parvissimi homines qui fiunt (sunt A) in aquis / H parvissimi

Available network displays

	A bimodal network of manuscripts and chapters. The 44 chapters of the first book of the Etymologiae are displayed vertically (green discs). This mode allows the users to see which manuscripts contain glosses to which chapters of book I, the thickness of an edge corresponding to the sum of the weights of the glosses to a specific chapter in a specific manuscript (i.e., a thick edge means that given manuscript contains both many and very non-trivial glosses to that chapter). The manuscripts' horizontal distance to the chapters are proportional to the the number of glossed chapters it contais (i.e., its degree). Its vertical position depends on where most of the glosses are concentrated, so that a uniformly glossed manuscript should appear in the middle, while having more glosses in the first or the last chapters would move it upwards or downwards respectively.
	A bimodal network of manuscripts and clusters. It represents the 53 annotated manuscripts (red discs) and the 34 identified clusters of glosses which appear in more than one witness (colourful squares). This mode allows the users to see which clusters appear in which manuscripts, each edge representing the sum of glosses appearing in a specific manuscript that belong to a certain cluster. The algorithm used to compute the layout of this visualisation reflects the relative closeness of manuscripts and clusters.
	A multigraph representing the mutual relationships between the 53 manuscripts based on the presence of shared glosses. Each edge represents the sum of glosses shared by two manuscripts within a certain cluster. They are coloured in the same manner as these clusters in the previous visualisation mode. The nodes' initial layout is based on geography and corresponds to (roughly) the assumed place of their glossing. The upper left segment of the graph, thus, represents England, while the lower right segment contains manuscripts annotated in northern Italy. The central part of the graph correspond to northern France.

Network interactivity

When activated from this panel the networks are shown in what we refer to as the networks' 'curated display'. This display can be adjusted by the user in a number of ways.

• A mouseover on edges and nodes gives some extra information about the (network) properties of the objects.
• Clicking on a node hides everything but that node's network. Clicking on the node again re-displays the entire network.
• For nodes and edges there is a context menu activated by a right-click. The available actions depend on the object: remove the object from the network, display the corresponding entry in the manuscript database, display or preview the corresponding chapter in the Etymologiae, etc.
• You can move the nodes around for a more informative or pleasing layout.
• Using the slider on the top of the graph, you can filter out nodes (in the first graph) or edges (in the other two graphs) based on their weight.
• Various options can be set using the options panel. These include:
  • The manuscripts and clusters to be visualised in the graph can be filtered.
  • You can filter the non-triviality ranks to be included in the computation.
  • You can select a different algorithm for computing the layout of the network.
  • The button 'Curated display' ignores these preferences and computes the editor-defined layout. The button 'Compute graph' takes into account the preferences set above.
  • Here, you can save the graph as a picture for use in publications.

BIBLIOGRAPHY

The following authors are referenced in the introduction to the edition and notes to the glosses:

• D. O'Croinin, 'A 'Lost' Old Irish Gloss Rediscovered', Peritia 32 (2021), forthcoming.
• Französisches Etymologisches Wörterbuch, edited by Walther von Wartburg et al., 25 vols. (Basel and Nancy, 1973-2002).
• L. Holtz, 'Sur trois commentaires irlandais de l'Art majeur de Donat au IXe siècle',Revue d'histoire des textes 2 (1972), 45-72.
• W. Meyer-Lübke, Romanisches Etymologisches Wörterbuch, 3 vols. (Heidelberg, 1935).
• W. M. Lindsay, Notae Latinae: an account of abbreviation in Latin mss. of the early minuscule period (c. 700-850) (Cambridge, 1915).
• E. Steinmeyer and E. Sievers, Die althochdeutsche Glossen II: Glossen zu nichtbiblischen Schriften (Berlin, 1882).
• E. Steinova, 'Annotation of the Etymologiae of Isidore of Seville in Its Early Medieval Context', Archivum Latinitatis Medii Aevi 78 (2020), 5-81.
• C. Villa, 'Uno schedario di Paolo Diacono. Festo e Grauso di Ceneda', Italia medioevale e umanistica 27 (1984), 56-80.
• I. Williams, 'Irische und Britannische Glossen', Zeitschrift für celtische Philologie 21:1 (1940), 290-306.

UPDATE LOG (status of April 2025)

Nov 2021

• added 2 glosses to Bern101 (L9.7.5_1 and L28.1.5_1)

Sep-Oct 2022

• L7.2.3_7 moved among shared glosses
• percurrit corrected to praecurrit in L17.7.5_1
• a note added to three glosses in L27.29.9
• corrected an error making two glosses belonging to L34.11.4 attached to L34.11.3
• corrected a mistake in L39.10.7 (L39.10.7_3 was added to cluster C7 instead of L39.10.7_2)
• updated bibliography and corrected small errors in introductory texts

March-April 2025

• added text about cluster G to Clusters
• added a gloss from Schaffhausen42 to L39.23.3 and a gloss from Wolfenbuttel64 to L39.3.8
• added publications related to the edition to the homepage