Changes in progress in late Northumbrian: the extension of -s as genitive and plural marker

Some of the linguistic changes which are crucial in the history of English and have traditionally been ascribed to Middle English can already be observed in late Northumbrian. One of these changes is the extension of genitive singular -es from the a-stems to other noun classes. Another is the spread of nominative/accusative plural -as from the masculine a-stems to the neuters and to other declensions. The aim of this article is to establish the actual scale on which these two interparadigmatic analogical changes are found in the glosses to the Lindisfarne Gospels and to the Durham Collectar, both dating from the tenth century. The present study intends to shed more light on the process of morphological restructuring affecting the declensional system of late Northumbrian, and has three main aims: (a) to determine the pattern of dissemination of the innovative inflectional endings; (b) to establish which variables are significant in these analogical processes: declension, type and token frequency, grammatical context, Latin lemma, demarcation; and (c) to account for the different developmental tendencies of these two linguistic changes.


Aims and methodology
Although for the Old English period the focus has normally been on West Saxon, Anglian dialects (Mercian and Northumbrian) play a very important role in the history of English, as recognized by, among others, Hogg (1997: 98). Their importance can be seen in the fact that they already show some linguistic changes that have generally been ascribed to Middle English, such as accusative/dative syncretism, the extension of -s to the present indicative, or the weakening of unstressed vowels. The present article studies the extension of genitive singular -es and nominative/accusative plural -as from the a-stems to other noun classes in two of the most important extant witnesses of late Northumbrian (tenth century): the gloss to the Lindisfarne Gospels (London, British Library, Nero D.iv) and the gloss to the Durham Collectar (Durham,Cathedral Library,A.iv.19), composed about two decades later by Aldred the Provost, whom scholarly consensus identifies with the glossator of Lindisfarne (cf. Jolly 2016 and Roberts 2016, among others). Ross (1937: 99) remarked that -es was extended to practically all classes in Lindisfarne (lufes, saules, sibbes, brydes, oxes, widues), and that this must be considered as the normal form of the genitive singular in this gloss, although older forms are often preserved as well. Analogical extension of -es is also widespread in Durham, as shown by Rodríguez Ledesma (2018). With regard to the plural, Ross (1937: 100) observed that -as was extended to other classes in Lindisfarne (suordas, ebolsungas, witgas), although this inflection is not found in the u-stems and in the neuter n-stems. There is, however, a clear distinction between these two analogical processes: 'whereas the gen. sg. in -es is normal in practically all classes, the nom. acc. pl. in -as only occurs sporadically outside its own class ' (1937: 100). 2 It is the endings -o, -a that are normally found for the nominative/accusative plural in almost every class, their origin being partly the weak declension and partly the neuter a-stems (1937: 101).
The aim of this article is to establish the actual scale on which these two analogical changes are found in Lindisfarne and Durham. To this end a quantitative analysis has been carried out of all nouns attested in the glosses whose etymological inflection for both the genitive singular and nominative/accusative plural is other than an s-form (117). They consist of feminine nouns (nouns ending in -UNG, nouns ending in -ness, remaining ō-stems, i-stems, root-stems), kinship r-stems and n-stems. For the study of nominative/accusative plural, all neuter a-stems attested in the glosses in these cases have additionally been considered (94).
Although some general observations about these analogical processes are found in previous literature, no systematic investigation has been carried out. The present study intends to shed more light on the morphological restructuring process affecting the declensional system of late Northumbrian, and has three main aims: (a) to determine the pattern of dissemination of the innovative inflectional endings; (b) to establish which variables are significant in these analogical processes: declension, type and token frequency, grammatical context, Latin lemma, demarcation; and (c) to account for the different developmental tendencies of these two linguistic changes (the extension of genitive singular -es and nominative/accusative plural -as from the a-stems to other noun classes). The results of this study also contribute to topics of scholarly interest, such as the question of authorship (is Aldred the glossator of all four gospels?), or the supposed conservatism of Durham compared with Lindisfarne.
The data have been retrieved using the Dictionary of Old English Corpus (DOEC), which is based on Skeat's edition of Lindisfarne (1871-7) and on Lindelöf's edition of Durham (1927). 3 Given the inaccuracies detected in the DOEC with regard to Lindisfarne, all tokens have been checked against the facsimile edition (Kendrick et al. 1956) and the digitalized manuscript in order to detect possible errors.
For this study, the declension and gender of nouns has been assessed on the basis of Old English grammars (Wright & Wright 1914;Campbell 1959;Hogg & Fulk 2011), glossaries (Cook 1894;Lindelöf 1901), and dictionaries, such as the Dictionary of Old English (DOE) when available (A-I) and Bosworth & Toller. In some cases, however, there is no agreement among the sources, and different genders are assigned to a noun. 4 Because of these difficulties and disagreements, the classification presented in this article merely indicates that the nouns studied were mostly or usually a member of some particular class, in line with Lass (1997: 108).
The presentation and analysis of the data are organized as follows: feminine nouns are studied first (section 2), and they are subdivided into five categories: nouns ending in -UNG (section 2.1), nouns ending in -ness (section 2.2), remaining ō-stems (section 2.3), feminine i-stems (section 2.4) and feminine root-stems (section 2.5). Kinship r-stems are analysed in section 3, weak nouns or n-stems in section 4 and the plural of neuter a-stems in section 5. Examples of masculine a-stems with vocalic-plural forms are given in section 6 and conclusions are provided in section 7. In each section the results of the study of the spread of -s in Lindisfarne have been compared with those found in Durham in order to offer a more comprehensive account of the process of morphological restructuring affecting the declensional system of late Northumbrian.

Productivity of the a-stems
Previous studies have criticized the classification of Old English nouns found in traditional grammars, such as Campbell (1959) and Brunner (1965), on the grounds that they are 'presented in terms of their historical origins and etymologically predicted forms, rather than as synchronic structures' and are based on a set of Early West Saxon texts that 'provide a standard against which the deviations of other texts can be measured' (Hogg 1997: 96). By making a distinction between declensions which are synchronically active (major declensions) and those which are fossilized (minor declensions), Hogg claims that there are only three major declensions: the masculine/neuter as-declension (stan), the feminine a-declension (lufu) and the n-(or weak) declension (nama) (1997: 97). Krygier (2002) similarly uses the criterion of paradigm unity to reclassify Old English nouns into three declensional types: declension I (the es-type), declension II (the e-type), declension III (the an-type). According to Krygier, one of the advantages of this new model is that it is 'truly synchronic ' (2002: 316). However, it can be criticized on the grounds that it views Old English as a homogeneous entity, whereas in fact it comprises texts written in different dialects and at different times, for many of which (those written in late Northumbrian, for example) this model does not apply. Another advantage, according to Krygier, is that 'the model has clear retrodictive power', i.e. the 'ability to correctly predict and explain the subsequent developments as they really happened ' (2002: 317), but this argument can be criticized as being post hoc, since the model was established once it was already known the direction in which English had changed. 5 Lass (1997: 108) sums up these difficulties: Given the general behaviour of Germanic gender-assignment and noun-declension, it's never entirely safe to say that some particular noun N 'was an X-stem of gender G'. The best we can say, given the rather exiguous remains of Old Germanic in general, is that, in the vast and murky continuum of Germanic dialects, any given noun was mostly or usually a member of some particular class.
Although I agree with Hogg (1997) and Krygier (2002) that etymology is not an adequate criterion for a synchronic analysis of Old English, I have followed the traditional classification (with Lass's caveat) since it is useful when it comes to identifying and describing the different analogical processes taking place in late Northumbrian. 6 These analogical changes can be described as interparadigmatic, since they involve the spread of inflections of a morphological pattern, that of the a-stems, to nouns belonging to other declensions that followed a different pattern, the result being the reduction of allomorphy across inflectional classes. Two important facts about analogy stressed by Hogg are that it 'can take place in improbable directions ' (1979: 66) and that 'failure of analogical change is often quite unpredictable ' (1979: 62).
The a-stems included the highest percentage of nouns in Old English. Quirk & Wrenn (1957: 20) state that 45 per cent of the nouns in the Old English lexicon are masculine, of which nearly four-fifths have genitive singular in -es and nominative/accusative plural in -as, and 25 per cent of the nouns are neuter, almost all of which have genitive singular in -es and nominative/accusative plural in -u or without ending. Hogg & Fulk stress the frequency factor and remark that, out of the 100 most frequent nouns in Old English, 54 belong to this class (including subclasses): 47 a-stems (22 masc., 25 neut.); 6 ja-stems (3 masc., 3 neut.); 1 wa-stem (masc.) (2011: 14). 7 According to these figures, the quantitative preponderance of the a-stems can be seen not only in terms of 'type frequency' (it contained the highest percentage of nouns in Old English), but also of 'token frequency' (nouns belonging to it occurred very frequently in texts). Bertacca (2001: 5-6) gives several reasons for the productivity of the a-stems in Old English: (i) it included over 50 per cent of the nouns in the Old English lexicon; 5 Hogg & Fulk (2011) offer both the traditional diachronic and a synchronic classification of Old English nouns, but for the latter they resort to etymology in order to explain the allomorphic variation found in the different paradigms. 6 For the same view, see Adamczyk (2018: 13) and Fernández Cuesta & Rodríguez Ledesma (2020: 41). Keeping the traditional model is useful for comparison with other texts written in other dialects or at different stages in the Old English period, many of which preserve the etymological inflections to a much greater extent than Lindisfarne or Durham do. 7 Of the 200 most frequent nouns in Old English, over 30 per cent are strong masculines with nominative/accusative plural -as (Newman 1999: 75).

700
NIEVES RODRÍGUEZ LEDESMA (ii) except for some nouns (ende), this declension had word-based inflection, 'which is morphosemantically and morphotactically more transparent than stem-inflection' (2001: 5); (iii) many borrowings were integrated into this declension; (iv) except for a comparatively small number of neuter nouns (see table 7), the a-stems were the only ones in which the nominative/accusative plural had a distinctive inflection, which was not used for any other case; and (v) the a-stems attracted nouns from other inflectional classes. However, some of his arguments lack explanatory power (iii and v, for example) and, therefore, cannot be considered 'reasons' for the productivity of this class; they merely prove the fact that this was a productive declension, since it attracted other nouns.
Others are circular or ad hoc: with regard to reason (ii), for instance, there were declensions, such as the n-stems, which had stem-based inflection rather than word-inflection but were, nevertheless, productive in both Old and Middle English, since they attracted nouns from other inflectional classes and borrowings from other languages. Adamczyk (2008: 108) agrees with some of the reasons given by Bertacca for the productivity of the a-stems, namely the quantitative preponderance of this declension (i), and the fact that the inflection for the nominative/accusative plural was not used for any other case (iv). The 'distinctiveness' of its case markers is another factor: Both the GEN.SG. ending -es and the NOM/ACC.PL. marker -as were more distinct than their feminine o-stem counterparts (-e, -a) and were, consequently, more expansive markers, whose productivity was steadily growing throughout the Old English period. (Adamczyk 2018: 193) 2 Feminine nouns 2.1 Nouns ending in -UNG 8 Table 1 gives the list of nouns belonging to this class which are attested in the genitive singular in Lindisfarne and the number of occurrences of the different inflections. 9 As can be seen in the table, -es is the dominant ending in the genitive singular of these nouns. There are some tokens of zero inflection, a feature of Northern dialects normally associated with Middle English but already attested in Old Northumbrian (see Rodríguez Ledesma 2015Ledesma , 2016. By contrast, there is only one occurrence of the etymological inflection. In this class, therefore, the innovative -es inflection is well established in the paradigm, showing very little synchronic variation. Nouns such as ðrowung or claensung gloss Latin (L.) genitives ending in -is, so there is the possibility of influence of the original, the <s> in Latin triggering an <s> in Old English: (1) ðrounges deda asaegd biðon Passionis gesta narrantur 'the deeds of the passion are narrated' (LkHeadGl (Li) 90) However, -es is also found when rendering Latin forms ending in a vowel, as in (2), and in occurrences attested independently of the original, as in (3): (2) ic salde iuh maeht henisaes ł hniðrunges ł onufa nedrum dedi uobis potestatem calcandi supra serpentes 'I gave you power to tread on serpents' (LkGl (Li) 10.19) (3) mamon, þaet is gidsunges hlaferd 'Mammon, that is the lord of avarice' (MtMarg (Li) 6.24) Some of these nouns are attested in the nominative/accusative plural in the gloss. Most of them take the etymological inflection ending in a vowel: faestnunga (1x); groetencgo, groetengo, groetingo (3x); leasunga (1x); mercunga, merconga (2x); ðrowungo (1x). However, there are also occurrences of forms in -as, such as ebolsungas, ebolsongas, efalsongas, efolsungas (6x) for the noun efolsung, or gitsungas (1x) for gitsung. The spread of -as as marker of nominative/accusative plural in these nouns is, therefore, not as advanced as that of -es in the genitive singular.
a In these nouns there is variation between -a and -e (Hogg & Fulk 2011: 110). In Lindisfarne, -o is also found (costungo), probably illustrating the weakening of vowels in final position. b Cf. the form of'suiðung' in example (8).

NIEVES RODRÍGUEZ LEDESMA
With regard to Durham, the extension of -es is widespread in this gloss, no example of the etymological inflection being attested in the genitive of these nouns. Twenty-eight out of a total of 31 occurrences (90.32%) take the innovative ending -es, as against two instances of abbreviated forms (6.45%) 10 and one example of uninflected genitive (Rodríguez Ledesma 2018: 618). Only one of these nouns is attested in the nominative/accusative plural in Durham, bloedsvng, and takes -as: bloedsvngas (1x) (DurRitGl 2 (Thomp-Lind) 126.1).

Nouns ending in -ness 11
In his study on Lindisfarne, Ross gives 26 tokens of innovative -es for the genitive singular of these nouns vs 43 of -e and 3 of endingless forms (1937: 71). He comments on this variation and offers a possible explanation for the limited extension of -es: 'there was a tendency to avoid forms in -es in these words with s in the stem ' (1937: 73). Table 2 gives the list of nouns belonging to this class which are attested in the genitive singular in Lindisfarne and the number of occurrences of the different inflections. My figures are very similar to those given by Ross (1937). As can be seen in the table, forms with the etymological inflection are dominant: most nouns take -e in the genitive singular, and those which present variation tend to favour -e (cyðnise, geliornise, smirinise), the exceptions being soðfaestnise and unrehtwisnise, which favour -es. The noun hernise stands out, since only forms in -es are attested in the genitive in the gloss.
Some of the innovative forms gloss Latin genitives ending in -is, so there is the possibility of influence of the original. Thus, the noun soðfaestnisse is attested five times in the genitive: the four instances ending in -es gloss L. ueritatis, whereas the example in -e renders L. iustitiae: (4) gast soðfaestnises spiritum ueritatis 'the spirit of truth' (JnGl (Li) 14.17) (5) cuom forðon to iuh in weg soðfaestnise uenit enim ad uos iohannes in uia iustitiae 'for [John] came to you in the way of justice' (MtGl (Li) 21.32) The noun unrehtwisnise, which also favours -es, renders another Latin noun whose genitive ends in -is (iniquitatis), as in LkGl (Li) 13.27: alle ða wyrcendo unrehtwisnises, L. omnes operari iniquitatis 'all the workers of iniquity'. However, the noun which stands out in this group, hernise, since it has only variants in -es in the genitive, renders a Latin form ending in a vowel: Table 2. Genitive singular of nouns ending in -ness in Lindisfarne In some cases there is no internal gender agreement within the noun phrase: the noun takes the etymological inflection but is modified by the historically masculine/neuter form of the demonstrative: 12 (7)  The spread of -es is not so advanced as in the previous group, probably for phonetic reasons, as pointed out by Ross (1937). That may also explain why no example of -as is attested in the nominative/accusative plural of these nouns in the gloss, the few forms found taking the etymological inflection ending in a vowel: smiriniso (1x), telnisse (1x), unrehtwisnise (1x).
Comparison of these findings with those obtained from Durham (Rodríguez Ledesma 2018: 622) confirms that the spread of -es to these nouns has advanced further in Durham, where examples of innovative -es outnumber those of the etymological ending: 25 out of a total of 64 (39.06%), as against 9 of -e (14.06%). 13 This was already pointed out by Ross, who remarked that in Durham the genitive singular in -es coexists with that in -e, and that 'here, in contradistinction to his practice in L[indisfarne], Aldred has no feeling against -es ' (1971: 56). Only two of these nouns are attested in the nominative/accusative plural, and both take the etymological inflection ending in a vowel: eftseleniso (1x), hernisso (1x).

Remaining ō-stem nouns
As can be seen in table 3, forms in -es are dominant in the genitive of these nouns in Lindisfarne, followed by forms with the etymological inflection and finally by uninflected forms. Most nouns (12 out of a total of 20) take -es in the genitive singular, and those which present variation (6) tend to favour the innovative inflection (lufu, sibb), the exception being cneoreso, which favours a vowel. 14 The spread of -es has advanced further than in nouns ending in -ness, but not so much as in those in -UNG.
As in the previous group, some of these forms, both those ending in a vowel (9) and those ending in -es (10), are modified by the historically masculine/neuter form of the demonstrative: 12 For similar instances in Durham, see Rodríguez Ledesma (2018: 624-5). 13 The remaining are tokens of abbreviated forms (Rodríguez Ledesma 2018: 622). 14 Maybe because of phonetic reasons (cf. nouns ending in -ness).

Feminine i-stems
Although few nouns belonging to this class are attested in the genitive in Lindisfarne, this gloss exhibits generalization of innovative -es to all of them. However, they retain the etymological inflection ending in a vowel in the nominative/accusative plural: tido (1x), tide (1x); uorulda (1x).
Four of these nouns are attested in the genitive in Durham: ae (2x), maeht (4x), tid (1x) and worvld (4x), and they take -es (8x), except for one instance of uninflected vorld and two of abbreviated forms (Rodríguez Ledesma 2018: 632). In Durham the number of tokens is much lower than in Lindisfarne, but in both glosses there is generalization of -es in these nouns, no example being attested of the etymological inflection. 15 In this class, therefore, the innovative -es inflection is already well established in the paradigm, showing no synchronic variation. With regard to the nominative/accusative plural, these nouns retain the etymological inflection ending in a vowel in Durham, as is also the case in Lindisfarne: dedo (4x), maehto (1x), tido (3x); worvlda, worvldo, worvlde (see Lindelöf 1901).

Feminine root-stems
Only two nouns belonging to this class are attested in the genitive singular in Lindisfarne: naeht 'night' and burug 'city'. The former takes innovative -es: naehtes (3x), but the latter shows variation: burge (1x) and uninflected burug (1x). By contrast, the five instances of nominative/accusative plural take -as (burgas). This seems to be the only noun in the gloss which has adopted the s-plural but not the s-genitive. 16 In Durham there are five tokens of genitive singular naehtes, whereas burug is not attested in any of these cases.

Kinship r-stems
As shown in table 5, there is widespread generalization of -es to nouns belonging to this declension, with the exception of swoester, which takes zero inflection in the only example attested in the gloss. The Lindisfarne data seem to reveal a different picture from that described by Adamczyk for this declension, according to which 'the spread of innovative inflection appears to have been affected by the gender, with the feminine nouns resenting [sic] the adoption of the masculine genitive sg. -es inflection' (2014: 67-8).
In Durham only two of these nouns are attested in the genitive: faeder (17x) and moder (1x). With regard to faeder, abbreviated forms are dominant (10x), followed by forms in -es (6x) and finally by zero inflection (1x). The only instance of the genitive of moder is abbreviated moder' (cf. Rodríguez Ledesma 2018: 633). In the nominative/accusative plural, forms ending in a vowel or abbreviated forms are attested: broðro (4x), broð' (57x), bro' (15x), br' (1x); dohtoro (1x).  Table 6 gives the list of weak nouns which are attested in the genitive singular in Lindisfarne and the number of occurrences of the different inflections. They have been separated according to their gender: feminine nouns are listed first, and then masculine nouns. As can be seen in the table, -es is the dominant ending in the genitive singular of n-stems in Lindisfarne, although not all nouns behave in the same way. Grammatical gender does not seem to condition the choice of inflection: forms in -es are dominant with both feminine and masculine nouns, despite the fact that some nouns take only the etymological inflection.

Neuter a-stems
The etymological inflections of these nouns in the nominative/accusative plural are -u for short stems and zero for long stems. Campbell (1959: 144) explains that -u in final unstressed syllables is lost after a long stressed syllable (word) or a short stressed syllable and another syllable (weorod), but is preserved after a short stressed syllable ( fatu) or a long stressed syllable followed by a short syllable (heafodu). Nonetheless, many analogical forms are found, such as heafod on the pattern of weorod (1959: 146). Nouns ending in a liquid or a nasal often develop a parasitic vowel in the uninflected forms: tungol, waeter, tacon. The nominative/accusative plural of these neuter nouns should have zero inflection, since the final -u should be lost after the long syllable before the parasitic vowel was added (*tunglu > *tungl > tungol), but forms in -u are also found (tunglu, waeteru, tacnu), normally with loss of the parasitic vowel (1959: 226). 20 Although u-apocope after a heavy syllable is the rule in early texts, the situation is different in late West Saxon and, especially, in late Northumbrian. Hogg & Fulk (2011: 108) state that all disyllabic neuter nouns are inflected in the nominative/ accusative plural in North Northumbrian, which seems to indicate that the synchronic rule of apocope had been lost from the inflectional morphology of a-stems. The loss of this rule is confirmed by the emergence of heavy-stemmed monosyllabic nouns ending in a vowel in the plural, such as wordu, and may have been motivated by a desire to distinguish singular and plural forms (2011: 92). Two other factors in the loss of this rule may have been the replacement of -u by -a as the regular inflection for the plural, and the existence of phonologically regular -u after a long syllable in old ja-stems (witu) and feminine nouns ( yrmþu) (2011: 108-9). a They include ja-and wa-stems such as cynn and cnew respectively. b No example of etymological -u is attested in Lindisfarne, the usual vowels being -o and -a. c Originally this was an os-stem that was assimilated to the a-stems (DOE s.v. cild). d Like cild, lomb was originally an os-stem that was assimilated to the a-stems. Table 7 gives the list of neuter a-stems which are attested in the nominative/accusative plural in Lindisfarne and the number of occurrences of the different inflections. Some of them are classified by Hogg & Fulk (2011: 14-15) amongst the 100 most frequent lexemes: bearn, bebod, gebed, folc, godspell, heafod, hus, land, þing, weorc, wif, waeter, word, wundor, yfel. The results of my study show two changes in progress in the nominative/accusative plural of neuter a-stems in Lindisfarne: the spread of the masculine s-form plural marker, which is, however, restricted to ten items, and the extension of a vowel to nouns with a long stem whose etymological inflection was zero, a change that is much more widespread in the gloss.

NIEVES RODRÍGUEZ LEDESMA
My findings confirm and refine Hogg & Fulk's description given above: although most disyllabic neuter nouns are inflected in the nominative/accusative plural in Lindisfarne, some of them take zero inflection: bispell (2x), timber (1x). Nonetheless, the rule of apocope seems to have been lost from the inflectional morphology of a-stems, since monosyllabic nouns with a long stem tend to add a vowel in the plural. Not only are forms such as husa/huso, wifo/wifa, woerca/woerco or wordo/worda attested in the nominative/accusative plural, but they are much more frequent than uninflected hus, wif, woerc or word (see table 7). Only in the case of scip 'sheep' does the etymological form outnumber those ending in a vowel. 21 My results are consistent with those obtained by Newman (2011) in his study of disyllabic neuter a-stems. His corpus comprises six late Old English texts (three West Saxon, one Northumbrian, one Mercian and one Kentish), and he concludes that there is 'extensive irregular vocalic-plural marking in these neuters in LOE [late Old English]' (2011: 12). Newman also finds evidence of the spread of the masculine s-form plural marker: the s-plural forms tacnas and tunglas, both attested in LG [Lindisfarne Gospels], indicate an increase in the productivity, and likely the incidence of the masculine s-form plural marker in Late Northumbrian, just as the n-plural form tunglan, attested in CH [AElfric's Catholic Homilies] (I), points to similar circumstances regarding the weak n-form plural marker in Late West Saxon. (2011: 12) 22 As mentioned above, only 10 nouns out of a total of 73 have variants in -s in Lindisfarne: cild, eher, godspell, hol, nest, ric, sedl, sword, tacon, tungul. In only three of these, s-forms outnumber those ending in a vowel: godspell, hol, sedl. Influence of the Latin lemma or of neighbouring words may account for some of these forms. Thus, in DOE ehras is considered an anomalous masculine form in (12), and the inflection -as due perhaps to the influence of the alternative gloss croppas or the feminine accusative plural ending of the Latin lemma spicas (DOE s.v. ēar 1 ):  Table 8 gives the list of neuter a-stems which are attested in the nominative/accusative plural in Durham and the number of occurrences of the different inflections. 28 As in Lindisfarne, there is limited evidence of the spread of -s to the plural of neuter a-stems in Durham (only one token), and widespread extension of a vowel to disyllabic nouns and heavy-stemmed monosyllabic nouns whose etymological inflection was zero. These findings are in line with those obtained for feminine nouns, kinship r-stems and n-stems (see sections 2, 3 and 4), and show that the spread of -s in the plural is an incipient change in progress in late Northumbrian, not so advanced as the extension of -s in the genitive singular.
6 Masculine a-stems with vocalic-plural forms In both Lindisfarne and Durham there are some masculine a-stems which take a vowel in the plural instead of, or besides, etymological -as. This is the case with the nouns eswice, gebearscip, hehstald, hlaf, hrof and hunger, as illustrated by the following examples: Masculine a-stem hlaf behaves slightly differently from the previous nouns, since it presents variation. In Durham only one token is attested, lafo (DurRitGl 1 Influence of the Latin lemma cannot be argued to account for these forms, since in many cases they render nouns ending in -s, such as L. simulationes (17), uirgines (19), fames (21) or panes (23). These forms seem to illustrate a further step in the generalization of a vowel as plural marker in the glosses, after its spread to heavy-stemmed neuter a-stems which etymologically had zero inflection in the plural (section 5).

Conclusions
This article has shown the extension of -es from the a-stems to all noun classes in Lindisfarne regardless of gender and declension, a clear example of interparadigmatic analogical change. In this light, Lass's observation that in Old English 'anything in -es is a genitive singular (and not feminine) ' (1992: 104) has to be modified: that may be the case in early texts or in West Saxon, but in late Northumbrian this inflection is used as a genitive singular marker of feminine nouns as well.
Although forms in -es are found in the genitive singular of all noun classes, the spread of this inflection advances at different speed in the different declensions. At one end there are three classes with a widespread extension of -es: feminine i-stems, nouns ending in -UNG and kinship r-stems. In these classes the innovative genitive -es inflection is already well established in the paradigm, showing little synchronic variation. At the other end of the scale are positioned nouns ending in -ness, the only group which favours the etymological inflection, probably for reasons of euphony, as pointed out by Ross (1937: 73). Comparison with Durham shows similar results: as in Lindisfarne, nouns ending in -UNG and feminine i-stems present widespread generalization of -es and little or no synchronic variation. The main difference between the two glosses concerns nouns ending in -ness: the extension of -es to this class has advanced further in Durham, where they outnumber forms with the etymological ending (see table 9).
The widespread extension of -es to feminine i-stems and kinship r-stems may be due to the fact that both are minor declensions and therefore contain few nouns, although they tend to have a very high token frequency. This factor, however, does not seem to inhibit the operation of analogical processes in Lindisfarne and Durham. It may seem strange that feminine i-stems were not attracted to the productive feminine declension, the ō-stems, but the reason may be that these nouns were already undergoing a change in progress at the time, gradually adopting the genitive inflection of the a-stems, as shown in table 9.
Declension, therefore, has been shown to play a significant role in the extension of genitive singular -es in both Lindisfarne and Durham. The Latin lemma may have had some influence, since many of these forms gloss Latin nouns belonging to the third declension whose genitive singular ends in -is. However, as shown above, the -es inflection is also used in rendering Latin forms ending in a vowel. Therefore, influence from the Latin original may have reinforced an existing trend in the language, rather than being fully responsible for it.
As is the case in Durham (cf. Rodríguez Ledesma 2018: 638), the frequency variable does not seem to play a crucial role in Lindisfarne. The -es ending is found and even favoured not only by low-frequency nouns, but also by those with a high token frequency, such as the i-stems ae, woruld, or the n-stems eorðo, hearta and lichoma. 29 This seems to run counter to the hypothesis that 'high-frequency forms are resistant to change on the basis of the structure of other forms and patterns' (Bybee 2015: 102) 30 and may indicate that the spread of -es is quite advanced in the glosses, at the stage of nearing completion.
Demarcation is not significant either, since innovative forms are not confined to certain sections of Lindisfarne that have been identified in previous literature (Brunner 1947/48;van Bergen 2008). 31 The noun lichoma, for example, is attested 10 times in the genitive singular: 9 tokens end in -s and one in the etymological vowel (MtArgGl (Li) 1). 29 These nouns are frequent not only in Lindisfarne, but also in the whole corpus: a search in DOEC reveals c. These results are not in line with those obtained from the study of accusative/dative syncretism in Lindisfarne, which show that this process is not even throughout the gloss, with the section corresponding to John having the lowest degree of syncretism in a statistically significant way 32 (cf. Fernández Cuesta & Rodríguez Ledesma 2020). The spread of -es to feminine nouns in Lindisfarne is interpreted by Lass as evidence of gender loss, a process which began in the north: 'there are fluctuations as early as the tenth-century Lindisfarne Gospels, where, for instance, endung "ending" appears as masculine (abstract nouns in -Vng are historically feminine)' (1992: 107). However, I disagree with this interpretation, since in that case those feminine nouns with -es in the genitive singular would be expected to take -as in the nominative/accusative plural, something not supported by the data, as shown above. Rather, I favour Jones' contention that the earlier gender/case indicative function of the -es inflection has been 'waived in favour of one purely of case' (Jones 1967: 104). Sandred (1991: 5) also remarks that -es genitives such as byriges or burges 'are no proof that OE burg was masc. in the language of the writer' and Adamczyk (2018: 135) explains that the spread of -s in the paradigm of feminine nouns 'did not necessarily entail a change in gender affiliation'. 33 Sometimes there is no internal gender agreement within the noun phrase: the noun takes the etymological inflection, but is modified by the historically masculine/neuter form of the demonstrative or adjective: lytles geleafa (MtGl (Li) 14.31, 16.8), godcundes stefn (MkArgGl (Li) 2). These examples are not so frequent in Lindisfarne as in Durham, but seem to indicate that -es no longer signals gender, but just grammatical function, and can therefore be added to any element in a noun phrase to unambiguously mark genitive case (see Rodríguez Ledesma 2018).
With regard to the other change in progress, the evidence for the spread of -s as plural marker is very limited in both Lindisfarne and Durham, showing that this is an incipient change. Most of the nouns which take -es in the genitive singular in Lindisfarne and are attested in the nominative/accusative plural in the gloss take the etymological inflection ending in a vowel: nouns ending in -UNG ( faestnunga, groetengo, mercunga), remaining ō-stems (byrgenna, ceastra, cneoreso, laro, stowa, synna) and weak nouns ( faermo, hearta, witgo). There are only occasional occurrences of forms in -as with nouns ending in -UNG (ebolsungas, gitsungas), the kinship r-stem fader ( fadoras), and the weak nouns witga and stearra (witgas, steorras). The spread of -s as plural marker is also very limited in Durham, where -as forms are attested only with nouns ending in -UNG (bloedsvngas).
The results of the study of neuter a-stems are in line with these findings and show that the spread of -s to the plural of these nouns is an incipient change in late Northumbrian, 32 It has been argued that Aldred could have based his gloss to the gospel of John on a previous translation made by Bede (see Roberts 2016: 53). 33 See also Hogg & Fulk (2011: 141).

CHANGES IN PROGRESS IN LATE NORTHUMBRIAN
restricted to very few items (see tables 7 and 8). There is, however, another change in progress in the nominative/accusative plural of these nouns: the extension of a vowel to heavy-stemmed neuters whose etymological inflection was zero, an analogical process that is much more widespread in the glosses and results in the elimination of allomorphic variation within this paradigm and the loss of the rule of apocope.
My findings support Ross's observation that in late Northumbrian there is a clear distinction between these two analogical processes: whereas genitive singular -es is extended to practically all classes, nominative/accusative plural -as is 'somewhat abnormal where it was not etymologically justifiable' (Ross 1937: 101 footnote 236). The differences in the pattern of dissemination of these two inflections may have to do with their type frequency: genitive -es is the etymological inflection of both masculine and neuter a-stems and, therefore, its type frequency is much higher than plural -as, found only with masculine a-stems. In the plural, moreover, there are two patterns in competition in late Northumbrian: forms ending in a vowel and forms ending in -as. A vowel (regardless of its quality) is the etymological inflection of a large number of nouns: ō-stems, i-stems, u-stems and light-stemmed neuter a-stems and, because of the loss of /n/ in final unstressed position in late Northumbrian, is also found in weak nouns. In the glosses, it is also extended to heavy-stemmed neuter a-stems. Numerically, then, vocalic-plural forms outnumber those ending in -as in late Northumbrian. As the type frequency of this pattern became higher in these varieties, it had the power to attract more members to it, reinforcing its productivity and extension throughout the plural, as shown by the fact that forms ending in a vowel are also attested for the nominative/accusative plural of masculine a-stems, such as hlaf, hrof or hunger. My findings, therefore, support Fertig's claim that 'analogical change favors whatever patterns and pattern types happen to be most common in a language, in terms of type frequency ' (2013: 121). 34 Lindisfarne and Durham show similar results with regard to the development of these analogical processes: in both the extension of -es as a genitive singular marker is much more widespread than -as as a plural marker, and the pattern of dissemination across declensions is the same. Both glosses provide evidence that these linguistic changes that have generally been ascribed to Middle English were already in progress in late Northumbrian, hence their importance for the history of English.