Instances of loanword/native word textual variation in the manuscript transmission of Egils saga Skallagrímssonar and Gísla

This article is concerned with intrastemmatic variation in two sagas of the Icelanders, Egils saga and Gísla saga.1 The point of departure for this research is a comprehensive investigation of the interplay between loan words and native words in Old and Middle Icelandic (Tarsi [forth­ coming]).2 Here, the focus will be on a handful of case­studies which can accomodate a primarily philological nature and which involve elements


Introduction
This article is concerned with intrastemmatic variation in two sagas of the Icelanders, Egils saga and Gísla saga. 1 The point of departure for this research is a comprehensive investigation of the interplay between loan words and native words in Old and Middle Icelandic (Tarsi [forth coming]). 2 Here, the focus will be on a handful of casestudies which can accomodate a primarily philological nature and which involve elements 1 Intrastemmatic variation here is defined as the variation between a loanword and a syno nymic native lexeme in a given locus of a text in different manuscript witnesses preserving that very same text. The core idea of the present article was presented at GLAC 22 (2016), the annual conference of the Society for Germanic Linguistics, with the title "Loan words vs. Native Words in Old and Middle Icelandic: The coexistence and competition of (quasi) syno nymic couples in the texts and the lexicon". 2 The research project Loanwords and native words in Old andMiddle Icelandic (2016-2020) documents the coexistence and competition of loanwords and their respective native syno nyms across all typologies of Icelandic literary prosa in the period before the rise of purist attitudes in Iceland (end of 16th/beginning of 17th c.). In 2018 the project was awarded a twoyear grant from the University of Iceland Research Fund (Háskólasjóður Eimskipa félags Íslands). I wish to thank Andrew Wawn (University of Leeds) for having corrected my English and Dario Bullitta (University of Turin) for the fruitful discussion we had at the XX Seminario Avanzato in Filologia Germanica (16)(17)(18)  C version (K, Chesnutt 2006: 52 8-9 ) Egill hafði vapn sín. Suerð ok kesíu ok buklara· hann hafði ok vapn sin oll.
hann hafde vopn sïn øll, sverd og kesju, hjälm og skjølld· The loanword buklari is of particular interest, for it only occurs once in the whole text of M and is not found in any of the other fragments of the saga. It can thus be said that buklari is a hapax legomenon in the transmission of this saga. The question that arises is whether buklari can be considered an original reading in comparison to those others encountered and with particular reference to the C version, where the native word skjǫldr is used instead. 5 According to Jón Helgason's stemma (cf. Chesnutt 2005: 229), the B and C versions derive from a lost exemplar, which in turn derives from the same manuscript as the A version. In order to shed light on the present case, it is notable, firstly, that W preserves a text which is somewhat shorter than the versions in M and K. This is an important feature of W, which consistently shortens the text of the saga (cf. Finnur Jónsson 1886Jónsson -1888. The portion of text under discussion has infact features similar to the two examples adduced by Finnur Jónsson (1886Jónsson ( -1888 to make his point. The lectio in W can thus be considered as spurious for it follows a general tendency encountered in the B version of the saga letters (α-κ). Of these fragments, the oldest is θ, from 1240-1260 (A version). The other fragments are grouped as follows: η (ca. 1450-1475), κ (15th c.) (A version); β (ca. 1350), γ (ca. 1300-1325), δ (ca. 1300), ζ (1250-1300), ι (15th c.) (B version); α (16th c.), ε (ca. 1400) (C version). Among other manuscripts, Stock. perg. 7 4to (ca. 1300-1325) has independent value and preserves a text of the B class (for further discussion about the manuscripts of Egils saga see Finnur Jónsson 1886-1888: i-xxix). 5 OIcel. buklari 'buckler' is first attested in a þula believed to be from the 12th century (Skjald. AI: 667). The word is thought to derive from Middle Low German (bokelēr), where it is a loan from OFr. (escut) bucler, maybe from medieval Latin *(scutum) buccularium (cf. IeW, AeW, ÍOb, s.v. buklari). The word is not found in any other skaldic text, whereas its attestations in prosa are from the 13th century (cf. ONP, s.v. buklari and the remark by Gurevich 2017: 824). This suggests that either the word was inserted in the þula at a later time, or that the þula itself is somewhat younger.
OIcel. skjǫldr 'shield' is attested from the 10th century and is a word of native stock. The word has wellattested cognates in other branches of the Germanic language family: Got. skildus, OE scield, OS skild, OHG scilt 'shield'. The Germanic lexemes all go back to PGmc *skeldu-'shield'. and thus does not constitute a stylistic departure. The lost exemplar a, from which the B and C versions derive, can be said to have had a text which, if anything, was nearer to that of K. This is further confirmed by the fact that the same wording as in K, with the exception of the word ǫll, is preserved in the δ fragment (ed. Kjeldsen 2005: 147), which contains a B class text (cf. Finnur Jónsson 1886-1888: 131 and apparatus). As for the reading hjalm ok skjǫld in a (= Kδ), it may be noted that it has a particular rhythm when combined with the preceding sverð ok kesju. It could thus be suggested that the word hjalmr was added in a for the sake of completeness, for Egill did indeed wear a helmet, but also for rhythmical (i.e. textual) reasons. However, the opposite could also have been the case; namely, that the word hjalmr was indeed part of the original text but was omitted from the A version. There are thus elements in favor of either interpretation, although the latter seems more plausible in light of a comparison with the description of Egill's gear at the beginning of chapter 45, where, conversely, the shield is not mentioned. 6 We are thus left with chosing between buklari and skjǫldr. As we have no other A class witness to help shed light on this issue, relevant internal as well as external factors need to be considered. Firstly, it was noted that buklari occurs just once in the manuscript in which M is preserved, AM 132 fol. By contrast, the word skjǫldr occurs fortyfive times in M. Secondly, the word buklari occurs twice in the manuscript as a whole, once as a simplex (Egils saga) and once compounded with targa 'shield' (Njáls saga), whereas skjǫldr is used 197 times as a simplex and also twice in a compound with targa (Njáls saga) (see MvB, s.vv. buklari, skjöldr, törgubuklari, törguskjöldr). Thirdly, the word buklari occurs just three times in the whole corpus (cf. OlÍsl., s.v.): as the previously men tioned compound in Njáls saga, and as a simplex in Egils saga and Grettis saga, although in the latter saga it does not constitute a word pair as the buckler in question is never referred to as skjǫldr. However, data from OlÍsl. should be used with caution, for it treats the sagas as printed texts, not as manuscripts. Finally, it may be noted that buklari occurs frequently in narrative genres such as Kings' and Chivalric sagas (cf. ONP, s.v. buklari), whereas it appears only once in skaldic poetry (see footnote 5).
When all this evidence is considered, the question arises as to which Instances of loanword/native word textual variation of the two readings, buklari or skjǫldr, is more likely to be the original one. As has been shown, analysis of the textual tradition suggests that Egill's gear had to be listed in that particular locus of the saga (lectiones of M and a (= Kδ) but not W). Now, both buklari and skjǫldr would have been meaning ful choices. The loanword, albeit used ana chronistically in the saga narration, 7 was adopted in Icelandic before the purported date of composition of the saga. However, on closer scrutiny, buklari seems an implausible reading here, in the light of its infrequent use in the saga, in manuscript copies of the saga, and in the sagas of the Icelanders overall (cf. also the general com ment on this topic in West 1973: 51). The present analysis suggests that buklari may have substituted skjǫldr in that partic ular locus of Möðru valla bók. It can thus be proposed that skjǫldr is the original reading.

A few cases in Gísla saga
Gísla saga 8 includes a certain number of word pairs which alternate loan word and native word intrastemmatically. Also, the manuscript tradition of this particular saga has received much attention both in early and more recent times. 9 Guðni Kolbeinsson and Jónas Kristjánsson (1979) pro vide 7 That buklari was not in use in the North at the time when Egils saga takes place is reported by Hjalmar Falk (1914: 133), who is most probably the source of Sigurður Nordal's foot note in his edition of Egils saga in Íslenzk fornrit (ÍF II: 106 fn 1). Falk (1914: 133 fn 1) notes that the word is used anachronistically in Grettis saga but adds that bukl could have been acquired in Icelandic earlier than buklari. for the word ímunbukl 'shield' occurs in a lausa vísa by Grettir Ásmundarson. 8 Gísla saga is preserved in three different versions; S (longer version, Dan. større), M (shorter version, Dan. mindre), and B (fragmentary version, Dan. brudstykke). The main difference between S and M relates to the length of the introductory section. Scholarly debate about the relationship and origin of these three versions, and particularly of S and M, has long been part of discussion about this saga, and also, more generally, about philological practice itself (for an overview see e.g. Þórður Ingi Guðjónsson 2010 and Leth bridge 2013). The longer version (S) is preserved in two 18thcentury paper copies of a lost vellum manuscript, the socalled Membrana Regia Deperdita (Loth 1960). These copies, AM 149 fol. and NKS 1811 fol., were directly and independently made from the Membrana Regia Deperdita in two different periods, the former ca. 1700 and the latter ca. 1780. Accordingly, they both have independent value. The main manuscript for the shorter version (M) is AM 556a 4to (last quarter of the 15th century). The B version is preserved as four vellum folia (2r-5v) under the shelfmark AM 445c I 4to (ca. 1390-1425). 1 If either M or B agrees with S, that reading is to be considered original. 2 If each version of the saga preserves a different reading, it is likely that S preserves a more original text than do M or B. Moreover, M pre serves a more original text than B. 3 If M and B agree against S it cannot be determined which of the two read ings is original, for M and B stand for a lost manuscript Y against S. 4 It should be noted that y preserved a shortened text. Moreover B preserves a relatively shorter text than M. As a general rule, the more complete readings in S are to be regarded as original. 5 If the exemplars of S, AM 149 fol. and NKS 1181 fol., preserve different readings, these should be compared where possible to those in M and/or B, and the particular reading of S which is paralleled in M and/or B should be chosen.  (1956, middle) and Guðni Kolbeinsson and Jónas Kristjánsson (1979, right) Instances of loanword/native word textual variation At this point, it is interesting to test this set of rules, and thus the stemma from which they are derived (Fig. 1, right), against the stemmata pro posed in stead by Finnur Jónsson (1929, Fig. 1, left) and Jón Helgason (1956, Fig. 1, middle). The purpose of the following test is to see whether the intra stem matic variation between loanwords and native words can say some thing about the validity of the stemmata proposed for Gísla saga, i.e. whether it pro vides further evidence for one stemma over the others. The word pairs on which the present discussion focuses are ambáttþý 'slavewoman', buffeitkinnhestr 'buffet, slap', fríafirra 'to free' and kompánnfélagi 'companion, fellow, comrade'. The three stem mata are reproduced here as they appear in Guðni Kolbeinsson  In the case of ambátt and þý, we can unfortunately use only two of three manuscript witnesses. However, the good news is that S and M have in common a considerable part of the text under discussion. If the compar ison between S and M in this locus is tested against the three stemmata given above, the result will be the same, i.e. that þy hans is an addition in S. In particular, Finnur Jónsson's view that M preserves a text nearer to the original than S yields that þy hans was later added to S. In Jón Helga 10 OIcel. ambátt 'slavewoman' is first found in Icelandic in a 9thcentury kvaeði by the skald Þorbjǫrn Hornklofi. The word is undoubtedly a very ancient loan in Germanic, as it occurs already in Gothic (andbahts 'servant' son's stemma, however, the three versions of the saga are put on the same level, for he cannot find enough elements that confirm Finnur Jónsson's stemma. Jón Helgason argued that it was not possible to consider all the agree ing lectiones of B and S as secondary to that of M, as they were just as likely to be primary. In our case, we do not have B to compare with M and S. A shared portion of text between M and S ought to be con sidered original, and hence, the deviation of S from the lectio of M should be considered spurious. According to Guðni Kolbeinsson and Jónas Kristjáns son's pentalogue, rules 1 and 4 must be taken into account. Rule 4 suggests that more complete readings in S are to be considered as original. As regards the question of loanword and native word, rule 1 applies in that M agrees with S and employs the word ambátt, and thus any agreement of either M or B with S gives an original lectio. Following the given manuscript evidence, it is safe to assume that am bátt consti tutes an original reading here. The question is whether S actually pre serves the original lectio or whether the more concise reading of M is prefer able. A major trend in Gísla saga scholarship has been to consider the longer ver sion as secondary to the shorter version. In Björn K. Þórólfs son's 1943 edi tion, the editor goes as far as calling M and S, E and Y respectively, i.e. eldri 'older' and yngri 'younger', thus classifying the two versions accord ing to relative chronology of their witnesses. I want to focus here on a stylistic fea ture of the reading pre served in S, namely the synonymic dittol ogy þý hans eðr ambátt. 11 Syno nymic dittologies are also found elsewhere in the text of Gísla saga. A com parison between M and S concerning synonymic dit tol ogies which use the explanatory conjunction eða reveals that such a sty lis tic trait is not peculiar to either version of the saga. However, where the M version appears to have such a kind of dittology, the corresponding locus in S is, if not the same, regularly shorter. 12 This clashes somewhat with the fact 11 A synonymic dittology is a figure of speech in which two words of similar or identical semantic content are linked by the coordinative conjuction and (OIcel. ok, Icel. og) or the disjunctive/explanatory conjunction or (OIcel./Icel. eð(u)r, eða) in order to convey a certain meaning, along with a pleonastic stylistic reinforcement. Synonymic dittologies are one of the chief ways in which loanwords and respective native synonym may be paired in a given text. that the dittology þý hans eðr ambátt is preserved in S instead of M. This is not an isolated example, as S preserves the dittology sýruker ... eða stokkaker (Loth 1960: 13 19 ) whereas M has sýruker tvǫ (Finnur Jóns son 1929: 4 10 ). This last example, in particular, seems close to that of ambátt and þý, as it is the second element of the dittology to be added. Ultimately, it is impossible to establish which of the two readings is the original. It is pos sible, however, that it is the longer reading that has been augmented, and this may hold true also for the longer readings of M against those of S enu me r ated in normalized orthography in footnote 12. OIcel. kinnhestr 'slap' is a neoformation attested since ca. 1200 in Icelandic. The word is only found in North Germanic (ODan. kinhaest, OSw. kinhäster). The semantic relationship between the constituents of this word is obscure and requires further etymological inquiry. steder aendret af afskrivere" (Finnur Jónsson 1923: 454 fn 1). In chosing between the two, it is probable that Finnur Jónsson's decision depended on the idea that kinnhestr had to be more original because it is an endogenous word. A purely speculative argument in favor of buffeit can be made, as was pointed out to me by Klaus Johan Myrvoll in 2016, whom I thank for having drawn my attention on this specific issue. The argument is that buffeit could be considered an original reading because a Middle English loan (OIcel. buffeit < ME buffet < OFr. buffet, bouffet, Eyvindur Eiríksson 1977: 46) would be perfectly plausible for the period in which the saga is considered to have been composed (ca. 1250). In Jón Helgason's ap proach, instead, the witness of B is crucial because the agreement of two versions over the third gives the original reading or, at least, the most original that may be surmised. According to Guðni Kolbeinsson and Jónas Kristjáns son's method rule 1 applies and it yields the same result as Jón Helga son's; the chief difference is that the agreement in Guðni and Jónas' approach has necessarily to be between two versions, each of which has to be in a different branch of the stemma. Here, S and B agree against M, thus giving kinnhestr as the original reading. An argumentum ad absurdum can also be made in order to further verify this result. If buffeit mikit were the original reading, then kinnhestr had to be corrupted into the two other ver sions, S and B. This is clearly at odds with the philological principle in voked above for the similar case of buklari and skjǫldr. Here, also, buffeit is not a plausible reading for exactly the same reasons as in the case of buklari. The M version of Gísla saga is in fact the only instance of this word in the whole Old Icelandic corpus (cf. ONP, s.v. buffeit). 14 Instances of loanword/native word textual variation OIcel. fría 'to free' is first attested in Icelandic in works from the 14th century. AeW and ÍOb consider the word to be a loan from MLG vrīen 's.m.', whereas IeW (s.vv. prēi-and frí) claims that the word is native to Icelandic, along with the adjective frír. The relatively late first attestation of fría is suggestive of a loan rather than an inherited word. OIcel. firra 'to deprive, free, save, defend' is attested in Icelandic from the 10th century, in a drápa by Halfreðr Óttarsson. The word must be considered a neoformation since it is a derivative of the adjective firr 'further', comparative form of fjarri 'far'. Truth to tell, the attestation of cognate verbs in West Germanic (OE aferran, OHG firren) could also point in the direction of a common source for the North and West Germanic verbs. 16 OIcel. kompánn 'companion, fellow' is attested in Icelandic in works from the 13th cen tury. The word is considered a loan from MLG kumpān 's.m.', which in turn is borrowed An interpretation of this locus of the saga yields the same result with all three models examined in the present paper. In fact, as the reading felogum/ felaugum (SM) is opposed to kumpanum (B), both Finnur Jónsson's and Guðni and Jónas' stemmata (rule 1) suggest that such oppositions are part of two different branches of the tradition, no matter whether M or S is considered closer to the original text. Jón Helgason's main rule is rather that the agreement between any two out of three versions is decisive in establishing which variant reading is original.

Discussion and concluding remarks
In the present article, a textcritical method is used in order to evaluate intra stemmatic variation between loanwords and native synonyms in the manu script tradition of two sagas of the Icelanders, Egils saga and Gísla saga. Whereas in the case of Egils saga, only one word pair of interest was identified, 17 in the case of Gísla saga the text, albeit shorter, contains four examples of intrastemmatic loanword/native word alternation. The word pair buklariskjǫldr in Egils saga was interpreted as a single substi tution made in the text preserved in Möðruvallabók. In analyzing the material from Gísla saga, a comparison of different stemmata was carried out. The result is that in three out of four instances, the three from OFr. compai(g)n < Lat. companio, companium (IeW, s.v. kompán, kumpán(n), kumpáni; AeW, s.v. kompánn, kumpánn; ÍOb, s.v. kompáni, kumpáni). The Latin word, whose oldest attestation in from the Pactus legis Salicae (ca. 503,Du Cange,s.v. com panium), constitutes with all probability a structural calque of a Germanic word similar to Got. gahlaiba 'fellow, companion' and OHG galeipo 's.m.'. The Latin native word for 'fellow, com panion' was socius.
OIcel. félagi 'companion, comrade' is attested since the 12th century in Icelandic sources. The word is a neoformation and has no cognates in the Germanic language family out side North Germanic (ODan. faelaegh, OSw. faelagh, DR 68 runestone (11th c.) filaka) besides OE feólaga, which is a North Germanic loan (OED, s.v. fellow). 17 Truth to be told, in Egils saga also OIcel. kápa (< Lat. cappa, likely via late OE *cápe < ONFr. cape, cf. OED, s.v. cape) and ólpa (< PGmc *welp-/wlep-, cf. ME wlappen 'to wrap, fold' and NSaam. vuöl'po 'womans coat' < Old West Nordic, ÍOb, s.v. ólpa, olpa and Qvigstad 1893: 351) occur in intrastemmatic variation when compounded with OIcel. loð-'furred ' (cf. OIcel. loðinn 's.m.', Got. liudan 'to grow', OE léodan 'to spring, grow', OS liodan 'to grow' etc., < PGmc *leuđ-/luđ-). However, an intrastemmatic analysis of this word pair yields no conclusive result, as the locus corresponding to ÍF II: 232 13 is omitted in the C version (for the A version see Bjarni Einarsson 2001: 141 3-4 , B version (÷ W, locus preserved in AM 463 4to p. 96 33-35 , MS from 1664, see also Schwabe 2015: 8). Instances of loanword/native word textual variation stemmata yield the same result, although the choice between the two variants is based on a different classification of witnesses. In one case, that of buffeitkinnhestr, the stemmata by Jón Helgason on one side and Guðni Kolbeinsson and Jónas Kristjánsson on the other give the same result, whereas Finnur Jónsson's stemma cannot determine which of the two readings is to be considered original. Although in the case of the very limited phenomenon analyzed here the results given by the three stemmata are not too divergent, the classification proposed by Guðni Kol beins son and Jónas Kristjánsson certainly proves to be the better one. In fact, the phenomenon examined here invariably provides a positive test result for the pentalogue proposed by the two scholars.
Besides the more specifically philological considerations with which we have been dealing so far, some concluding remarks on the nature of the phenomenon under analysis are in order. On a general level, in docu menting the phenomenon of coexistence and competition of loan words and native words the former were all considered equally in the docu mentation phase, since priority was given to highlighting the contrast between loanwords and native words. Loanwords are a chiefly diachronic phenom enon, whereas on the synchronic plane all lexical material in a given language following its phonological and morphological rules is con sidered in much the same way, except when specific linguistic knowl edge allows for an evaluation of a word's provenance, e.g. when purist atti tudes are part of a conscious language policy (Gusmani 1981: 14-15). The situation is different with integral borrowings, i.e. loans which pre serve the phonological and in some cases also morphological structure of the source language. 18 This latter case is often associated with learned use, or at least language use strongly influenced by the prestige of the source language in the recipient language community, in particular with reference to written language. This case is not represented here, nor are integral borrowings in the sagas of the Icelanders very common. 19 In four out of the five cases listed above, the loanwords were adopted into Icelandic in medieval times, and can thus be said to be more or less contempo rary with the written media in which they are recorded. In the case of ambátt, however, the situation is different, as the loanword is a borrowing from Celtic into Germanic (Lehmann 1986: s.v. andbahts). A method o log ical consideration then arises, as to whether cases such as that of ambátt should be treated differently from those of loanwords roughly contem porary with the sources in which they are recorded. The answer seems to be a negative one. 20 From the overall excerpted data (Tarsi [forth coming]) it appears that the native lexical strategies underlying the expansion of the lexicon do not differ with respect to loanword chronol ogy. In order to exemplify this general principle, one may consider the typol ogies of native words corresponding to loanwords whose first attestation in Icelandic is in works composed before and after the twelfth century, i.e. before and after the oldest preserved Icelandic manuscripts. 21,22 In the former group, which is unexpectedly much smaller than the latter, are words such as ambátt 'slavewoman', biskup 'bishop', djǫfull 'devil', harri 'lord', klaeði 'clothes', kristinn 'Christian', messa 'mass', ormr 'worm, serpent, dragon', peningr 'coin, money, wealth' all of which are found to alternate with native words, but also other loanwords such as akkeri 'anchor', altari 'altar', eyrir 'ounce, money, property', kaupa 'to buy', kaerr 'dear' and a number of other loans attested in early skaldic poetry. 23 The native equivalents of the subgroup ambátt -peningr relate to all the four typologies mentioned above: 20 This consideration applies to the lexicon in general and not to specific lexical uses such as different registers used in a given text. This latter case could be that of Egils saga presented above, where a loanword, whose adoption is with all likelihood contemporary with the manu script in which a version of the saga is preserved, is present where the original text possibly had the corresponding native lexeme. 21 The loanwords presented here do not include those which show a hyponymic relationship with a native word. 22 Such a chronological division is, of course, arbitrary. In fact, there is nothing, in principle, that prevents presumably (or certainly) old loanwords from appearing in works dating from the twelfth century onwards. A representative example is OIcel. kirkja 'church', whose two oldest occurences in poetry and prosa respectively are in Einarr Skúlason's poem Geisli and in the Old Norse translation of the Elucidarius. Both works date from the midtwelfth cen tury the word may well have been adopted upon the introduction of Christianity (999). Other similar cases include OIcel. paradís 'paradise' (see furthermore the discussion in Tarsi 2016: 89-90, "OIcel. kristinn 12th c." read "11th c.") and OIcel. skrifa 'to write' (Tarsi 2019). 23 Albeit classified under a single label here, the acquisition of these loanwords spans a lengthy period of time, from the PGmc period until the eleventh century (the conversion to Instances of loanword/native word textual variation Tab. 1. A taxonomy of native words corresponding to loanwords attested in works composed before the twelfth century. As can be seen, all four strategies are represented also in this small group of loans. It would, of course, be possible to speculate about the produc tivity of neoformations in this period. Judging from the available data, it seems that the coinage of neoformations flourishes from the twelfth cen tury onwards, whereas before that time a stronger adherence to learned models is found. This result may of course be biased by the extent of early data available and by the textual typology in which they appear, which leads to the prevalence of certain semantic fields over others where stronger adherence to models was called for.