Old Norse-derived lexis in multilingual accounts: a case study

This article on Old Norse represents a fundamental departure from the previous literature on loaned material by examining multilingual documents written in Medieval Latin rather than in monolingual English, namely the Durham Account Rolls (DAR). The potential of this richer and more complex interplay between languages will be further addressed throughout the article, which assesses the different kinds of evidence available for establishing the relative plausibility for a word being derived from ON. Dance's (2013, 2018, 2019) taxonomy will be discussed and applied to multilingual material for the first time. The article concludes with some notes on the main semantic fields to which ON-derived lexis contributed within the multilingual lexical networks of the DAR.


Some notes on Anglo-Scandinavian contact in medieval England
The effects of the contact between speakers of Old English (OE) and Old Norse (ON) 1 in early medieval England still continue to fascinate researchers and the general public alike. The nature of the bilingual environment and culture that arose out of the continuous exchanges (linguistic and otherwise) between the newcomers and the native population are of extraordinary singularity in the history of English: the genetic proximity between ON and OE means that these two languages shared a significant number of cognates apart from evincing formal similarities. That is why a certain degree of mutual intelligibility has often been assumed (see, among others, Townend (2002) and Dance (2012) and the references therein) and the absorption of a large number of loanwords was facilitated. The most prototypical scenario for lexical borrowing involves requiring a new word to express a newly 'imported' concept or extralinguistic reality previously non-existent in the receiving community. In contrast to these 'need-based' loans, there was a transfer of material at a much deeper level: some prepositions and the third-person plural pronouns are argued to have been borrowed from ONsee Cole (2018) for an alternative, language-internal, explanation for some forms of themand native word forms took senses or meanings from Norse (e.g. dream). Only a situation of intense contact would have possibly catalysed the borrowing of such fundamental lexical units in a language, a phenomenon which would have derived, in Dance's words (2013: 43), from 'source-language-led "imposition" in communities routinely code-switching from Old Norse into Old English'. Yet the scarcity of firm evidence has often led some authors to reassess the role of contact-induced change in the explanation of linguistic change on a lexical and morphosyntactic level. Lass (1997: 209) would argue that 'in the absence of evidence, an endogenous explanation of a phenomenon is more parsimonious, because endogenous change must occur in any case, whereas borrowing is never necessary'. The World Loanword Database (Haspelmath & Tadmoor 2009a, 2009b, the largest compilation of cross-linguistic data available to explore lexical borrowing, has proven that the borrowing of core vocabulary is unexpectedly common, and that limitations on the kinds and degree of material subject to be borrowed between closely related languages are particularly elusive (cf. Haspelmath & Tadmoor 2009a, 2009bDance 2019: I.27). Regardless of our epistemological inclination, this finding should make us problematise Liberman's assertion that 'all other conditions being equal, tracing a word to a native root should be preferred to declaring it a borrowing ' (2008: xxvi). Contact can also be invoked in conjunction with potential native developmentswhich might have been actualised or accelerated because of the influence of the other languageso a scholarly polarisation between the advocates of internal vs external explanations for language change is not particularly helpful when thinking about multilingual scenarios; nor is the preference for language-internal accounts tantamount to dismissing any form of Scandinavian influence: from a lexical viewpoint, there has been general consensus on some fairly secure borrowings from Norse. Yet some aforementioned factors such as the typological proximity between the two languages and the paucity of textual witnesses means that 'it is often impossible to be sure whether borrowing or endogenous change is at work' (Dance 2019: I.29). Alluring as the Viking element has been ever since the nineteenth century, as Dance (2019) points out, in etymological research few directions have been given as to how to decide whether a particular lexeme exhibits ON influence when the amount (or lack) of information available can be utilised to argue both ways. The extent to which we should rely on negative evidence, that is, on unrecorded OE cognates, has divided scholars. Kolb (1965: 133) suggests not searching for such cognates too eagerly, whereas other authors embrace Norse input even if earlier native words (with the same meaning) are attested (see, e.g., Ringe 2004). Dance's (2013Dance's ( , 2018Dance's ( , 2019 rigorous typology will be utilised in the present article.

Aims, methodology and textual material
This article on Old Norse represents a fundamental departure from the previous literature on loaned material by examining sources written in Medieval Latin (ML) rather than in monolingual English: the Extracts from the Account Rolls of the Abbey of Durham from the Original MSS , edited by Fowler in three volumes (1898)(1899)(1900)(1901) and which I collated with the original manuscripts held at Durham Cathedral Library Archives. 2 Fowler's three-volume work is one of the largest edited collections of multilingual texts produced in Latin, 3 so it represents an ideal testing ground for the study of the influence of ON-derived lexis in texts which have not been interrogated from this perspective until now. Since the DAR are not like any other monolingual texts in Middle English (ME) and usually exhibit a certain degree of technicality, partly inherent to the nature of record-keeping itself, the vocabulary that surfaces in this material has an extra dimension of relevance in the multilingual context of late medieval and early modern England. 4 The potential of this richer and more complex interplay between languages will be further discussed throughout this article, but for now it is worth stressing that the taxonomical principles that underlie this examination are common to ME-based studies. I will, therefore, be drawing on classic reference works on ON-derived loanwords in medieval English such as Björkman's (1900-2) monograph and more recent research such as that carried out for the Gersum project (Dance, Pons-Sanz & Schorn 2019).
In order to examine the Norse element in the DAR, I shall briefly revisit some of the main criteria and kinds of evidence adduced for ON borrowings in the literature. In particular, Dance's (2013Dance's ( , 2018Dance's ( , 2019 typology and underlying theoretical basis and discussion will be followed. Important early contributions cited in Björkman's seminal work include Coleridge (1859), Steenstrup (1882), Skeat (1882), Brate (1885) and Kluge (1901). Björkman (1900) assesses the reliability and scope of his predecessors, 5 while foregrounding their treatment of ON borrowings, primarily on phonological grounds (see, e.g., Brate 1885: 4-30; Knigge 1885: 71-2). Coleridge (1859: 26) made use of a word's distribution in the Germanic languages, and Björkman (1900) himself concentrated on phonological tests but also considered other (less reliable) criteria. Dance propounds a twofold classification for the kinds of positive evidence that tends to be used for assessing Scandinavian input: 'structural' and 'circumstantial' evidence. Under structural evidence fall aspects relating to the formal structure of the language (e.g. phonology and morphology); and circumstantial evidence would 'derive from patterns of occurrence' (Dance 2019: I.36). He clarifies that chronology is not a reason per se for assigning ON influence; rather, the lateness of a word (900 is the terminus post quem for any visible lexical effect of Anglo-Scandinavian contact on written OE) is a prerequisite for a loan; attestations of lexemes earlier than the tenth century are deemed cognates, not loans. I will review in greater depth Dance's classification of evidence, which is primarily arranged into four main types (Types A, B, C and D), and I will illustrate each type with examples from the database. Broadly speaking, Type A relies on regular, formal evidence (phonology (A1), morphology (A2), and both phonology and morphology (A3)) to identify ON input. Type B signals that the Germanic root is not recorded in OE before Anglo-Scandinavian contact, so it examines the recorded use of that root across Germanic (Gmc) languages: the stem may be known only in North Germanic (NGmc) (B1) or in other Gmc languages as well (B2). In contrast, under Type C are lexical items which are indeed attested in the early, pre-contact, OE period, but an aspect of that lexeme (its derivational form (C1), orthographic/phonological form (C2), meaning (C3), the formation of a compound/ phrase (C4), or frequency (C5)) suggests ON influence. Finally, Type D indicates uncertainty: the etymology of the form may be problematic (D1) or the interpretation of the lexeme may be unclear in its Middle English (ME) textual environment (details of each type and subtype will be provided later in this article).
The lexis that is the focus of this article is a small part of a larger database of 1,598 lexical items (not counting orthographic variants of the same item) which were manually culled from the edited DAR. These rolls correspond to several departments: volume 1 comprises excerpts from the Cellarers', Hostillers', Almoners' (elemosinaria), Chamberlains' and Infirmarers' (magistrorum infirmarie) Rolls; volume 2, Commoners', Terrars', Stock-keepers', Sacrists', Feretrars', Bursars' Rolls and Marescalcia Prioris; and volume 3, Bursars' and Miners' Rolls. There are also some short extracts from Treasurers' books (1569-80) and miscellanea (1293-1542), mostly in monolingual English, which were beyond the scope of my research. All the ON-derived lexis here discussed is, therefore, attested in the rolls produced in ML. The Appendix lists all the spelling and morphological variants of the words in the DAR which are discussed in the body of the text. 6 The presence of the plural-forming suffix -ez in ON-derived lexis (e.g. crosez, flaggez, skelez, stopez) is rather common and may make us wonder whether the data herein analysed can be interpreted as Middle English or Anglo-French. Within the larger context of the DAR, where a variety of morphemes (-is/-ys, -es, -ez, or the zero morpheme plural) is used, my contention is that -ez is part of a large translingual repertoire that was available to scribes and that all these morphemes were used non-distinctively in multilingual texts, regardless of the etymological origin of the word. 7 The DAR material comes from an area beyond Samuels's 'Great Scandinavian Belt' 8 (1985: 269), so, in theory, a relatively lower incidence of ON influence should surface. A total of 160 lexical items (67 simplexes and 93 complex lexical units) have been tentatively classified as evincing ON influx (c. 10% of the total number of DAR vernacular items, although, out of that percentage, only c. 30% of them can be deemed rather secure (Type A) borrowings from ON). How representative this figure can be in the context of a text traditionally classified as 'Latin', in which there are few closed-class lexemes coming from the vernaculars (a few prepositions and the definite article from French), 9 is a question that remains to be answered in future research. For the purposes of this article, rather than analysing ON influence on the language of the DAR in purely numerical terms, I carried out a qualitative analysis. Many of these lexemes are classified under the ON-influenced headwords, so that any compounds/ lexical units are not counted as distinct lemmas. Other tentative borrowings were not included in the list because the evidence was either too tenuous (e.g. in capestonys 10 'the top stones' (cf. the OED ((1893), s.v. cope, n. 1 )) 11 'the Middle English forms might be from ON kápa, but this is an unlikely source') or it was suggested in the past but was ruled out (e.g. step in stepled 'a vessel/cauldron for soaking [something]', where the OED ((1916), s.v. steep, v. 1 ) indicates that it is 'of difficult etymology' and discards the possibility of it being a loanword from ON (Old Icelandic (OIcel) støypa) on phonological grounds). 12 The direct etymon is the main focus of investigation and, therefore, words such as til 'a cut of meat' and wyndas 'windlass' have been excluded from this list on the basis of their direct borrowing from French (even if they were ultimately borrowed from ON). So was brusket' (also bruskett') 'the brisket or breast of an animal': the OED identifies formal and semantic resemblances between brusket' and French brechet, whereas the Middle English Dictionary (MED) only considers its alleged NGmc parallels (OIcel brjósk 'cartilage', Danish (Dan) brusk). Palatalisation in the syllabic onset position is expected in Central French dialects (in contrast to the Insular variety), so it is more likely that brusket' comes from Anglo-French. The information provided by the OED and the MED has been collated with that included in Dance's volume 2 (2019) and the Gersum database, which also gives the proposed ON etymon, the OE cognate (if any), and phonological and morphological markers (if relevant). All of the terms in this article are regularly attested in ME.
While there might be discrepancies between the main etymologies given in the OED and the MED, some items do not attract much controversy. The classification below, 8 See Samuels (1985: 269) on the Great Scandinavian Belt, 'excluding the old kingdom of Bernicia in Durham and Northumberland'. 9 See Roig-Marín (forthcoming a) and the references therein for an overview of the use of the French definite article in this kind of multilingual text. 10 Only the first attestation in the DAR editions is here given. Capital letters are preserved as rendered in the editions unless the word is also attested in lower case, in which case the latter variant is preferred. 11 It is worth stressing that the OED is, in Durkin's words (2016: 392), a 'dynamically changing resource', which means that I will be citing the date of publication/revision for each headword used. 12 As per common practice, I will cite the OIcel reflexes as etyma of the ON loans.

OLD NORSE-DERIVED LEXIS IN MULTILINGUAL ACCOUNTS
nevertheless, is subject to further refinements. I have not made use of Dance's 'probability categories'unless they are adopted from the Gersum project or Dance (2019)but have indicated if the evidence for a tentative ON borrowing can be reinforced by the dialectal distribution of that particular word through the use of the letter 'c' (mainly confined to the north or east of England in the general ME lexis) or, just in one instance, 'b' (confined to the north or east in the toponymic record). The final section of this article considers the prevalent view on the semantic nature of ON-derived lexis in English and the specificities of the DAR data.

Type A
Systematic phonological and morphological parameters are well established in the literature as regular and reliable. 13 Indeed, as Dance puts it, 'their consistency is what allows comparative philologists to argue for English and Norse as distinct developments on the Gmc family tree in the first place' (2019: I.39). Dance further subdivides Type A borrowings into three subcategories: A1 phonological criteria, A2 morphological criteria, and A3 phonological and morphological criteria. A paradigmatic example of these formal, phonological, correspondences is OE /ɑː/ (corresponding to ME /aː, ɔː/) and ON /ei/. 14 If the English form happens to exhibit the corresponding ON sound instead of the expected OE-derived form, this would be indicative of a discontinuity between OE and ME generated by contact with ON. The OE cognate is, nevertheless, not needed to prove the influx of Old Norse precisely because of the systematicity of thesein this case, phonologicalfeatures. The probability of a native variant developing independently so that it would eventually converge with the ON form often seems unlikely; hence, we can speak of relatively 'secure' criteria within the philological paradigm. Yet, if there are no attested cognates in other Germanic languages apart from Old Norse, the distinctiveness of allegedly ON features may be at stake, and the possibility of independent convergence gains ground. 15 Unfortunately, there is only one instance of the A2 class in the DAR: the presence of the inflectional -t in twhertsawes 'crosscut saws'; -t is the suffix attached to adjectives to form adverbs in ON but not in OE: OIcel þvert adverb 'across', originally neuter of the OIcel adjective þver-r = OE þwerh (cf. OED (1912) thwart, adv., prep. and adj.). There are none representing A3, which is why I shall concentrate on the A1 group.
13 See Lass (1997: 123-39) on the need to establish correspondences which are '"lawful", statable in principle as particular instantiations of general rules'. 14 The fronting and raising of OE /ɑː/ in northern Middle English is accounted for in Roig-Marín (forthcoming b), focusing on northern Middle English spelling evidence in the DAR. It equally addresses the use of <ai> and <ay> for reflexes of OE /ɑː/ in northern late Middle English (see below). 15 Dance (2019: I.42) cites mynne, rake and rasse, among other examples of vocabulary which did not make the cut in this sense and are, therefore, classified as Type D.
Our understanding of certain phonological changes (A1) has been considerably refined since the nineteenth century, so that not all of them are regarded as equally reliableat least, under all circumstancesin present-day scholarship. Absence of palatalisation is a traditional index of ON influence. It had been long assumed that palatalisation had not been fully operative in Northumbrian, or that in initial position /k/ did not assibilate. 16 This belief, implying northern-exclusive developments, has been progressively invalidated ever since Gevenich (1918) proved prevalent /k/ > /tʃ/ in native northern toponyms. Yet its absence in non-initial position is much more problematic and is often seen as a possible native development, which is why in this article more secureinitial-positionenvironments will be considered. The reflex of PGmc (Proto-Germanic) */k/ is /tʃ/ in Old English but not in ON, as can be seen in castyng 'in ploughing, the method and operation of turning all the furrow-slices of a ridge in one direction, and those of the adjoining ridge in the opposite direction' (OED (1889), s.v. casting, n. 1.d.), cf. OIcel kasta, replacing the native OE weorpan in ME (Gersum, s.v. kest); kerr (Gersum, s.v. ker [A1bc]not in OE); keruyngknyves 'carving-knives' (OIcel kyrfa, OE cyrf); kydsape 'soap for treating sheep-louse' (OIcel kið); and kyrn 'a churn', a northern variant of OE cyrne, probably influenced by contact with ON (OIcel kirna). The other contexts in which we can perform a palatalisation test are PGmc */g/ > ON /g/, OE /j/ and PGmc */sk/ which also remains /sk/ in ON but palatalises in OE, resulting in ME /ʃ/. Gersuma 'a manorial rent' (< OE gaersama < OIcel gørsimi) and the garth ('yard') compounds (OIcel garð-r): bernegarth, connyngarth (×2), ympgarth, swynhousgarth, stakgarth (×3) and Wodegarth exhibit ON /g/. 17 It is worth noting that the native variant, with yard, is also present in three compounds, ymppeyard (×2), Wodyard (×5) and Hempyard. Both the native and the ON-influenced variants mostly occur in the fourteenth century, so there does not seem to be a diachronic motivation for the use of one form over another, especially in the cases of ymppeyard and Wodyard; rather, it is lexically conditioned. Sponegarn 'yarn produced by the process of spinning' (pronounced with /g/) also seems to exhibit partial input from ON (cf. OE gearn), and there is further evidence for the absence of palatalisation in the following lexemes starting with <sc/sk>: e.g. skelez 'vessels' (OIcel skjōla) (×7; ×1 with <sc>), 18 scale (pl.) 'scales', also in merowscales, Weyscill (all of them always spelt with <sc>), 19 scappes, also skepe (OIcel skeppa) (×7), sker and may skynnes. Scale (pl.), skelez and scappes are also in the Dictionary of Medieval Latin from British Sources (henceforth, DMLBS) (cf. DMLBS, s.v. 2 scala, skela and skeppa). Regarding the last two, skelez and scappes, they are written with <sk> and <sc> (see the Appendix for all the spelling variants recorded in the DAR), the former spelling allegedly representing an adapted Latinised version (nonetheless present in ME as well) and the latter being a reflex of the source etymon, directly adopted. These two graphemic versions are in ML as well as in ME. 20 Skelez is also spelt with <sch> (the spelling variant that the MED cites as northern '(N) schele' only because it is in the DAR, s.v. skē ̣le (n.)) in the records of the DMLBS (schalis and schelis with vowel alternation too), which evidences the flexibility of Medieval Latin. It might be contended that the early instances of these words were 'unadapted' and, therefore, contain <sk> and, as time progressed, <sc> replaced <sk> in ML texts. Wright (1998Wright ( , 2012 suggests the reverse phenomenon in the context of multilingual business writing which would show a progressive 'Anglicisation' of the lexis before there was a complete shift to monolingual English. The scarce <sk> data in the DAR prevent me from reaching any generalising conclusions, but the DMLBS shows considerable variation across space, time and possibly genre, rather than a steady trend (e.g. see skippis ((Ac. Milton) DCCant) as early as 1299 and skepcis (Househ. Bk. Durh. 133) from 1532).
Another consonantal change which may be more problematic in etymological descriptions is ON assimilation of /nk/ > /kk/ in drawkyng 'saturating quicklime with water'. According to the OED ((1897), s.v. drawk, v.), it should, in theory, be ascribed to Type D1 since its etymology is 'obscure', only 'possibly' related to ON (OIcel drekkja < PGmc *drankian 'to drawn') 'drench, drown, swamp, submerge'; yet, given the Northern English / Scottish distribution of the word (cf. the Dictionary of the Older Scottish Tongue [DOST], s.v. Drawk, v.; lemma not in the MED), the lack of a native source and/or closer Germanic etymon with this sense and phonological make-up, we can place it within this group.

Types B, C and D
The following types of evidence can be classified as less secure because they do not rely on formal criteria like Type A. The ultimate lexical source for these items is here the determining factor: the form source may not be attested in early OE but recorded in ON (in which case, the word would belong to Type B); it may also be found in early OE (Type C); or the form source may be obscure, not unequivocally identifiable (Type D). Generally speaking, the scale of probability ranges from the more secure Type B1 words to the unclear Type D items, but within these major groupings, there is also scope for a spectrum of individual cases, which may be equally placed on a continuum of likelihood as far as ON input is concerned (hence, the numeric subclassification).

Type B
An underlying premise with Type B words is that if a particular Germanic root is not recorded in early written OE, it may signify that it did not survive long enough to explain its use post Anglo-Norse contact. Frequency of use might play a part in a word's attestation, meaning that very specialised or low-frequency items might be absent in extant OE writings, but this should be carefully assessed on a case-by-case basis. A further distinction can be made between Type B1 (when the root is only known in 21 Swayn is always derived from Old Norse in the literature (see Gersum, s.v. swaynes and the references therein), but it is worth acknowledging that etymological /ɑː/ could also be graphemically represented as <ai> and <ay> in late northern Middle English (see Roig-Marín (forthcoming b) for details and its problematisation). Yet the <ai>/<ay> digraphic spelling of swayn is by no means unique to the north of England and it is in fact attested in the first half of the fourteenth century (thus, earlier than the proposed dating of this innovation in the north) in such texts as The Metrical Chronicle of Robert of Gloucester (c.1325 (c.1300) Glo.Chron.A (Clg A.11)), whose Linguistic atlas of late mediaeval English (LALME) linguistic profile locates it to Gloucs., in the south-west of England (see MED, s.v. swein and its other attestations).

OLD NORSE-DERIVED LEXIS IN MULTILINGUAL ACCOUNTS
NGmc) and Type B2 items (when these lexemes are indeed present in Gothic and/or the continental West Gmc languages). The information has been mostly gleaned from both the OED and MED, but the classification below is not exclusively reliant on the MED or the OED entries, some of which are still in the process of being revised and do not give enough details or may wrongly assume that a given lexeme unproblematically comes from ON.
Less straightforward B2 lexical items (nevertheless, probably ON-derived) are kagges, Ripp's and snypys. Kagges in the DAR has to be read as 'kegs', 'small casks or barrels', thereby fully corresponding to ON kaggi. Nonetheless, there seem to be cognates in Dutch kaag, Low German kag with the meaning of 'fishing-boats'. That word was also borrowed into French (cague 'fishing-boat' and caque 'a herring-barrel'), so the theory that 'ships, or boats, and casks, or tubs, often go by the same name' is echoed in the OED etymology of the word ((1888), s.v. † cag, n. 1 ), which is relatively uncertain (the MED just limits its etymology to give the ON etymon but does not mention any possible cognates in other Germanic languages), so the word is tentatively included here. Ripp's 'baskets for fish' is 'probably' cognate, according to the OED ((2010), s.v. rip, n. 1 ), with German regional (Low German) rep, OHG href, ref 'basket' (MHG ref, German (now regional) Reff 'pannier'). The OED also draws attention to the lack of additional supporting evidence by means of the word's regional distribution, since 'the (post-medieval) currency of the word in south-eastern counties outside the Danelaw area is surprising in a word of Scandinavian origin'. The development of lexemes throughout time can be varied and complex, so this can only be taken as a lexicographical curiosity. Snypys 'snipes' has an even more complicated origin: the OED links it to ON, but the lexeme seems only to be attested as part of compounds (Icel mýrisnípa, Norw myr-, strandsnipa) and the connections with MLG and LG (also older Danish) snippe/ sneppe, German dialect schnippe, MDut sneppe (Dutch snep), OHG snepha (snepfa) and snepho are not clear. Yet, because the word is not found in OE and the localised attestations in the MED suggest a northern/east midland distribution, it is included here rather than along with the etymologically obscure Type D words.

Type C
Type C items vary greatly in so far as an early OE source form has been identified; the revealing feature is a sense, usage or word form which might be rare or unparalleled in earlier OE. Type C items are, therefore, classified according to that defining characteristic, which can fall under the following linguistic levels or parameters: derivational morphology (C1), phonology (C2), semantics (C3), complex word-formation processes (C4) and frequency (C5), arranged by category below. 27 Note the OED claims that it is 'unknown elsewhere in Germanic' although it does cite the possible connection to the ablaut series to which OHG kracko belongs (cf. OED (1893), s.v. crook, n. and adj.).

OLD NORSE-DERIVED LEXIS IN MULTILINGUAL ACCOUNTS
Within Type C are also those lexical items whose (direct or indirect) source forms come from another language (e.g. Latin or French). The question is, again, whether the feature under consideration could be tentatively explained through an unattested (endogenous) descent or through a direct borrowing from that third language rather than from ON, a probability that may vary widely from item to item.
There are two Type C2 lexical items in the DAR which could be claimed to have arisen endogenously: cart and stoth. Cart (in Cartbod. (×2), Cartfleykkes, Carth'neys (×3), cartrapes (×5), Cartesadle, (×2), cartsadiltrees, cartsilver, Cartstrakes, Langcart', Stankart (le)) with metathesis (< PGmc *kratt-) is not present in OE (craet) or other 12 AMANDA ROIG-MARÍN West Germanic cognates (e.g. OHG kratto). As stated in the Gersum database (Gersum, s.v. kart (labelled CC2)) and the OED ((1888), s.v. cart, n.), there is no overlap of the metathesised and unmetathesised variants, and the unmetathesised form is not attested in ME, which may be revealing in assessing the extent to which ON could have influenced (not necessarily replaced) the native form. On the other hand, in stothes 'ornamental studs'also in dorestothez (×3)there is final-position /ð/ rather than /d/. Nonetheless, unlike girthys (Type A1, discussed above), there is a cognate in OE, so the word could have been a variant of the native studu, perhaps reinforced by ON, rather than a direct borrowing (OIcel stoð). The fact that most attestations of stooth are located to the north and east midlands is, again, not sufficient proof. Type C3 lexical items (coddis, thyxtyll, mosse, ferybote and spon (in sponbrod and sponyarn)), 28 based on semantics, are sometimes more slippery: spon represents a relatively secure Type C3 item because OE spón only had the sense of 'chip', not the utensil; coddis 'bolsters or bearings of an axle' may have had an OE source codd 'bag' and/or OIcel near-cognate koddi 'pillow'. The MED identifies sense 4, both 'a metal "cushion", such as a bearing of an axle or a bell' (4b) and, more generally, 'a pillow' (4a) as only northern (MED, s.v. cod n.(1)), and the OED devotes a separate entry to this northern usage (OED (1891), s.v. cod, n. 2 ), although it is noted that it shares its root with cod, n. 1 'bag'. The connection between 'a bag' and a pillowbeing a stuffed bagbecomes patent, but the fact that ON koddi (also Old Dan kodde) had the primary meaning of 'pillow' might have triggered semantic narrowing and the development of that particular sub-sense in the English lexeme; thyxtyll 'a kind of ax', in thyxtyll goug, probably comes from OI þexla (cf. Norw teksla, Dan teksel, with the same meaning in OHG dehsala), since it has a cognate in OE (OE þīxl) but with a different meaning 'beam or pole' (MED, s.v. thixel n.); 29 likewise, the double sense of mosse (×3) is linked to ON (OIcel mosi both 'moss' and 'bog, moorland') since it seems that OE mos only meant 'bog' (Gersum, s.v. mosse (CCC3ab); Dance 2019: II. §259). It has been claimed that both meanings might have already had currency in early OE, so a native-only historical trajectory is also possible; and fery in ferybote is another instance of partial multiple inputs: it is, in part, a borrowing from Scandinavian (OIcel ferja, OSw faeria, ODan faerie, all meaning 'ferry boat') as well as the outcome of the conversion of the verb ferry 'to transport or carry' ( ferian in OE and OSax, and on the more specific sense of 'by boat', see OIcel ferja, Gothic farjan, and MLG vēren, MHG vern) into a noun.
Bollez was tentatively included as Class C3c although it is a more dubious case: bollez (×2) is used as a term for a dry measure in the north of England and Scotland, so it was hypothesised that this sense might derive from ON (OIcel bolli) rather than OE bolla 28 Spon 'shingle' in Sponbrod 'a nail for fastening shingles' is a much less likely loan from ON: it is classified as a 'CCC3' word (see below and Dance (2019: II. §263) s.v. sponez 'spoons' (CCC3a)), so the likelihood of it being derived from ON is very low. 29 Unlike the MED, the OED ((1912), s.v. thixel | thixle, n.) mentions that no cognate in OE is recorded: 'known c1300, not yet found in Old English'.

OLD NORSE-DERIVED LEXIS IN MULTILINGUAL ACCOUNTS
'bowl', but the argumentation seems to rest on its northern distribution (OED (1887), s.v. boll, n. 2 ). The closest sense recorded in the Dictionary of Old Norse Prose is 'liquid measure, measuring cup of specific size' (ONP, s.v. bolli sb. m., sense I.2) whereas in the Dictionary of Old English (DOE) no measure-related meanings are attested. Because of the highly unspecific and broad nature of terms such as bolle 'cup' (also the content of the vessel, etc.), this sense could have been developed in parallel to ON, where its usage seems to be more limited to 'liquid measures'.
C4 and C5 are represented by only two lexical items and one respectively: (C4) axiltre (×6) and blandcorne and (C5) crosez. For the Type C4 stems there is a ME compound or combining form which is not in early OE although it has parallels in Scandinavian languages: the native equivalent to axiltree 'an axletree' (only recorded from early ME onwards) is axtree (OED (1885), s.v. † ax-tree, n.). Because axil was also in OE (OE eaxl), 30 there might have lexical substitution through contact with ON (cf. the OIcel compound öxul-trē). Blandcorne is attested in Scandinavian languages, both as a simplex and as a compound, OIcel bland 'mixture' and dialectally in Swedish, blandkorn (MED, s.v. bland-corn n. 'ON; cp. Swed. dial. blandkorn & E dial. (Yks.) blend-corn.'). There are also OE cognates bland and gebland, but, as Dance (2019: II.230) acknowledges, these OE stems are 'rare and confined to poetry', so its use in blandcorne could be tentatively derived from ON (also note this is consistent with circumstantial evidence of type 'c' for both axiltre and blandcorne). Crosez is classified as 'FC5b' (Dance 2019: II. §292; Gersum, s.v. cros): both the late OE cros (exclusively attested in the toponymic records) and ME cross have Latin crux as their ultimate etymons, but the direct source forms are less well-established; there are two possible sources, Old Irish cros and OIcel kross, and there are other cognates in AF and ML (crosce, crosse and cros, respectively) which might also have contributed to the development of the ME word if they are not borrowings of the ME lexeme themselves. Apart from the hypothesis of multiple inputs, Dance (2003: 417-18; 2019: II.246-7) and Durkin (2014: §4.1 n. 6), among others, are more inclined towards accepting the less complex explanation, which is to assume that there is just one source 'a Hiberno-Norse word borrowed from OIr cros' (Dance 2019: II.247, see also the extensive bibliography on this word listed under fn. 955).

Type D
The etymologies of Type D items are the most difficult to track down of the four categories. They are often classified as 'obscure' or 'difficult' in two main respects: the etymology of the word has not been agreed upon, although its form and sense are relatively clear (Type D1), or even the interpretation of the word in context can be highly debatable (Type D2). Type D1 is represented in the DAR by cloukis, sowmys and steyned clothes: cloukis, most likely, 'clutches', is a word with a complex history (OED (1891), s.v. clutch, n.1): the word is first attested in ME (clōke) and Scots (clūke), and only in the seventeenth century do there seem to be records of a Southern palatalised version (clooch). The reconstructed OE stem would have been *clōc, clōce or, alternatively, *clūc or clūce, following the paradigm of ME brōke < OE brūcan. The ancestor Germanic root would then be *klûka-or klûkôn-(whence OE clycc(e) an). Like the MED (s.v. clō ̣ke n.(2)), the OED notes the possible influence of the verb clicchen (< OE clyccan) on the palatalised forms. If we apply Occam's razor againas we did with crosezit would not be far-fetched to presuppose that the ME word clōke might have been adopted from another Germanic, nominal, root rather than assuming multiple unrecorded native developments and a word-class change (verb > noun). Yet an ON cognate does not seem to have been identified either, only a surviving form in Swedish (see MED, s.v. clō ̣ke n.(2), '?ON; cp. Sw klyka "a clamp, fork"'). Sowmys 'chains or ropes' has multiple possible sources: not only ON (OIcel saumr 'nail, seam') but also AF (OF some, soume denoting a 'pack-saddle'), even if there are semantic differences which are left unexplained; 31 an Anglo-Latin variant of LL sagma 'pack-saddle' (DMLBS, s.v. 2 salma, sauma, somma, suma, 1 summa) is also attested, which adds more complexity to the possible layers of influences that converged in the adoption of the ME word. The best semantic analogue, thus, seems to be the ON lexeme, and the supporting circumstantial evidence is here represented by the northern distribution of the word (cf. OED (1913), s.v. soam, n.; MED, s.v. soume). As indicated in the Gersum database (s.v. stayned), steynedclothes 'worked in colours or embroidered clothes'to be distinguished from stevynd, which in the DAR interestingly only co-occurs with Latin pannus/-i 'clothes'can be semantically traced to two forms: ME stainen < aphetic version OFr desteindre 'to remove colour' or OIcel steina 'to colour, stain' (on the same root as steinn (OE stān)), whose semantic development is complicated but seems to bear resemblance to other Germanic languages (cf. OHG staimbort ?'painted shield'), including OE ā-stǣnan 'to adorn (something) with precious stones' (DOE, s.v. ā-stǣnan).
As Dance (2013Dance ( , 2018Dance ( , 2019 fleshes out, the focus of this taxonomy is on etymological evidence and not on the outcome of contact. These two different ways of conceptualising ON input may interact in fruitful ways, but Dance's categorisation offers a more dynamic approach: the different types may encompass manifestations of language contact which would belong to different categories in traditional typologies (e.g. direct loanwords and loan shifts). Beyond structural evidence, other factors can come to play a role in assessing the etymology of a word. Dance brings to the fore two kinds of 'circumstantial' evidence, namely the Germanic distribution of a word in the form of cognates in closely related languages, represented by the West Germanic family (Middle Low German, Old and Middle High German, Middle Dutch, Old Frisian and Old Saxon) and the word's geographical distribution. The geographical confinement of a word to the north and the Danelaw area has often been identified as a marker indicating ON influence: while Björkman (1900) does consider a word's use as an auxiliary criterion, Miller (2012: 99) seems to go too far when claiming that this is 'the safest way to recognise a Scandinavian loan', especially if this distribution endures through time and still applies to Modern English. A correlation between such an important area of Scandinavian settlement and a greater incidence of ON input is not far-fetched and has indeed been demonstrated. 32 The labels that Dance proposes to distinguish the kind of documentary evidence coming from the north or east of England are 'b' for toponymic recordings, and 'c' for the general lexis (see Dance 2019: I.62). Nevertheless, as Dance points out, the evidence adduced to claim ON influence based on geographical distribution is often contentious (2019: I.56-7). The context of use (e.g. in a literary text) and the size of the corpus examined are just a few of the factors that may skew the representativeness of the material employed if these are not properly accounted for. The Gersum project contains a few of the items in the DAR database, but the context-dependent cues and semantic interpretation of them did not necessarily match that found in the DAR. This, for instance, applies to Type D2 sloknyng (B1 in the DAR) 33 and one Type CCC3 lexeme, bord, which cannot be among the ON-derived words because of its sense: rather than meaning 'table' (or derived meanings), senses which are rare in OE and may evince ON input (cf. ON borð), in the DAR it always occurs in complex lexical units (e.g. in bordnayles) and has its prototypical meaning, which can be traced to OE bord.
Likewise, homographs within my database will also be kept distinct: band in hatband is OED (1885), s.v. band n. 2 (from French), not the ON loanword (bandeʒ, also in belybandes and dorebandez; see below). Arguably, a few more words than those classified into the different types show inputs from Scandinavian in various ways: for instance, saez (×3) (also in watersay), another Scottish and northern term for 'a wooden tub', seems to derive from OE on the basis of the dating and contexts in which it is attested in early OE. Nevertheless, as the OED ((2015), s.v. sae, n.) remarks, the northern/eastern distribution of the lexeme and its ME variant soe possibly reflects ON-influx rather than an uninterrupted continuity with the native word. In the following section, I will overview the main semantic fields to which ON-derived lexis contributed within the multilingual lexical networks of the DAR by using the Historical Thesaurus of the OED (HTOED). 32 See Kaiser (1937: 178-278), Xandry (1914), Thorson (1936), Kolb (1965) and Samuels (1985) on the 'Great Scandinavian Belt' and the more noticeable number of ON-origin grammatical words (Samuels 1985: 274-5;Dance 2003: 289-91). The results are dependent on the amount of the localised evidence surveyed, which was less accessible in the past than it is nowadays thanks to the online MED, LALME, LAEME (A linguistic atlas of early Middle English) and corpora. 33 Sloknyng is here considered Type B1 because 'le Sloknyng' (Rott. Bursar. 1488-9, 651) 'extinguishing' in the DAR can be derived directly from ON (cp. Icel weak verb slokna 'to be extinguished'), whereas the meaning of slokes in Gawain is much more remote from its meaning in Scandinavian languages and, thus, more difficult to account for (see Gersum, s.v. slokes).
The significantly larger numbers of Scandinavian-origin lexemes in Middle English, extant in a much wider range of text-types and registers, better allow us to assess the realms of Anglo-Scandinavian contact: not only were Scandinavian-specific or technical terms adopted and/or adapted into English but also basic lexical units used in everyday situations, sometimes already having synonyms in OE. 34  A significant proportion of the ON-derived vocabulary has to do with building, 36 cultivation, ploughing, farming and animal husbandry. Equipment, either containers or tools represent 27 lexical items out of the total (160). I will now illustrate some of the most frequent categories under which these lexemes sit in the HTOED: 37 • society » farming » animal husbandry » general equipment: belybandes 'bands which pass round the belly of horses', latbrodes 'goads', tedyr 'a tether' and sowmys 'trace-ropes'; 34 Apart from the borrowing of new lexical material, the form, phonology or meaning of OE words was also sometimes remodelled to match the ON cognate (a well-known example of a process of phonological substitution in native material through contact is sister (cf. OE sweoster, swuster and OIcel systir) (see also in Gersum under half-suster and sister-sunes)). 35 Avermalts could also have been borrowed as a unit later on from another Germanic language, MLG hāvermolt (cf. earlier oat-malt). The OED ((2015), s.v. haver-malt n.) dates its first attestation to 1569, but it is attested two centuries earlier in the DAR (Rott. Bursar. 1388-9, 596). 36 Some names of materials also belong to this larger occupation domain of building (e.g. gaddis 'a metal bar or rod', flaggez 'flagstones' and wandes 'rods or sticks'). 37 Some labels in the HTOED seemed to be unnecessarily complex (e.g. the categorisation of slaughter under 'the world » life » death » killing » killing of animals'), so I have adapted and simplified the categories for the purposes of this article.

OLD NORSE-DERIVED LEXIS IN MULTILINGUAL ACCOUNTS
• plants (plants » particular plants), brakennez 'a variety of ferns', Scrabbez 'crab apples' ('food plant'), Lyng 'heather', and the compounds lynghouse and lyngthake; places to store materials, animals, or to cultivate plants, including bernegarth 'a barn', Impgarth 'a garden where seedlings are cultivated', swynhousgarth and Wodegarth; • natural phenomena or features of the landscape (under 'the earth' section of the HTOED, although scattered under different categories), dam, banke, kerr 'a marsh' and mosse.
Other miscellaneous lexical items include 'parts of a building' (inhabiting and dwelling » inhabited place » a building » parts of building), bandeʒ 'hinges of a door', dorestothez 'doorjambs', gabell 'gable or façade of a building', wadmale 'a kind of woollen cloth' and other -ing forms (some related to building as well) such as drawkyng 'saturating with moisture, as flour or quicklime with water', flyttyng 'removing', kyrvynge 'carving', ryvyng 'splitting of wood' and wyndyng 'wattling, enclosing with wattle-work'.
It is worth noting that not all the ON-derived words are classified in the HTOED (only, e.g., the separate units in dorecrokez, eldyngpan, twhertsawes and windowclathe), and a few others are semantically tagged with a sense different from the one in the DAR (e.g. flywinges as either (a) the wing of a fly or (b) bookbinding (see quot.), OED (1897), s.v. fly-wing n. rather than with the figurative sense of 'some kind of small nail' (MED, s.v. flīe n.(1), 3. (c) fle-wing, fli-)). Likewise, the two different kinds of borrowings preand post-twelfth century are present in the DAR (gersuma and male were already borrowed into OE) although other than those, there are not many cultural concepts. Rather, as expected given the nature of these rolls, vocabulary from everyday activities such as building, farming, ploughing and animal husbandry, including some specialised terminology for pieces, tools or instruments is the general trend of ON-derived lexical items in the DAR.

Final remarks
All in all, this article has concentrated on the challenges and opportunities in assessing the different kinds of evidence available for establishing the relative plausibility of a word being derived from ON in the DAR. It has drawn on Dance's taxonomy -Type A relying on formal (phonological, morphological, and phonological and morphological) features, and from Types B to D, the evidence is less secure in ascending alphabetical orderand applied it to the DAR multilingual material, an enterprise which had not been undertaken beyond the analysis of monolingual English sources. This lexical compilation also brings to light the many ways in which scribes would express plurality irrespective of the language of origin of the morphological material involved (see, e.g., gonnys, gonnez, gun, Gonnis, Gunnis), which underscores the particularly permeable language borders found in ML texts produced by multilingual scribes.
The semantic distribution discussed in section 3 points towards several semantic subdomains (most notably equipment, either containers or tools) as being particularly prone to making use of vernacular, ON-origin, vocabulary. The data available are, however, conditioned by the scribes' expressive necessities and frequency of usage in