PATTERNS OF LINEAR CORRESPONDENCE IN THE GOTHIC BIBLE TRANSLATION: THE CASE OF THE ADJECTIVE*

By examining the realisation of the adjective in Gothic against the original Greek, this paper explores the borderline between literalness and idiomaticity in the Gothic translation of the Bible. As a general rule, the Gothic translation follows the linearity of the Greek model very closely. Occasionally, however, Gothic strays from the Greek, the most typical types of deviation being concerned with the use of different grammatical material, function and word order. Where Gothic innovates elements unattested in the original or deviates in word order, it likely asserts its native properties. It is argued that, as a whole, the Gothic text presents in itself a mixture of truly idiomatic and marginally acceptable language.


INTRODUCTION
It has become a well-worn cliché, whenever a discussion of Gothic syntax is attempted, that the Gothic text follows its Greek original so closely 1 that hardly anything can be said about the native structure of Gothic. Thus, Bennett (1980, 127) writes with reference to the language of the Gothic Bible and the Skeireins: "Both documents, in fact, contain so many Greek syntactic features that they are all but useless for the study of Germanic syntax." In contrast, Peeters (1985,(76)(77) concludes that the Gothic translation is fully idiomatic and presents in itself "a priceless tool for the historical and comparative study of any other Germanic language". His thinking is predicated mers apply. 1 Literalness was a defining characteristic of all early Bible translations (Nida 1964, 12;Falluomini 2015, 66). The normative nature of biblical texts, which had a claim to divine authority, demanded literalness as a way to preserve the word of God and guard it from heterodoxy (Šarčević 1997, 25). The Gothic translation aiming at a word-for-word rendering of the Greek-even if it sometimes came at the expense of grammatical correctness or smoothness of phrasing-was therefore no exception. on the notion that, where the Gothic translation closely follows the Greek model, this is due to the two languages being structurally comparable; any mutually irreproducible aspects of the text were carefully considered and addressed by the Gothic translator, resulting in Gothic deviating from the Greek. This way, Peeters deems Gothic to be entirely idiomatic in both its allegiance to and deviation from the Greek.
Much research on Gothic grammar, especially syntax, has been based on exploring the evidence of Gothic deviations from the Greek model (and pre-Vulgate Latin translations of the Bible) or identifying the properties of Gothic that are independent of the Greek-for an overview of the basic differences between Greek and Gothic, with many references to further literature, see Falluomini (2015, 66-91); Miller (2018) offers the most up-to-date and linguistically comprehensive discussion. Recognising the incompatibility of the above approaches, in what follows I operate on the assumption that even the most literal translation cannot be performed without some regard for the structural requirements of the target language. In an effort to test this assumption, I give a descriptive account of the structural linear resemblances and differences between Greek and Gothic, with a focus on the realisation of the Gothic adjective.

GOTHIC DEVIATIONS FROM THE SOURCE TEXT(S)
Prior to addressing the actual data, a word must be accorded to the directionality of linguistic comparison conducted in this paper as well as the comparative sources used. Since this research represents a component of a larger project on Gothic adjectives, the point of departure in this discussion is Gothic material. The investigation is based on a corpus of 2,056 examples of adjective tokens from the Gothic Bible, collected on the basis of Snaedal's (2005) Concordance. In an effort to distil from the corpus of evidence instances of Gothic that might showcase its authentic linguistic properties, care has been taken to compare the Gothic not merely with the Byzantine Greek Bible, which served as the basis for the Gothic translation, but also the Alexandrian text and the pre-Vulgate Latin translation ("Vetus Latina"), whose elements the Gothic text is also known to contain (for details, see Falluomini 2015, 92-129) The Gothic citations are taken from Streitberg (2000). The Byzantine citations follow the Majority Text edition of Robinson and Pierpont (2005), used as the primary comparator; the secondary comparator was the Critical Text of Nestle et al. (2012). For the pre-Vulgate Latin translations I consulted Jülicher (1963Jülicher ( -1976 as well as the online Vetus Latina Database (VLD). Where a pre-Vulgate witness was lacking, the Vulgate (Weber & Gryson 1994) was consulted instead.
For the purposes of this investigation, only those deviations that have an immediate impact on the adjective or its relations within the clause have been considered as relevant. It is impossible to name every kind of deviation, as a detailed classification may go down to individual examples. As a result, the classification given here covers the most typical patterns. At the most general level of description, Gothic deviates from the source text(s) in material, function and word order.

(I) Differences in material
Deviations in this class encompass every instance of adjective use in Gothic that does not correspond to an adjective in Greek. This difference in lexical class may or may not concatenate with differences in syntactic functions or relations.

(i) No functional/relational change
The examples in this group represent lexeme-to-lexeme correspondences that share a sameness of syntactic function while being rendered by an element of a different lexical class. The following kinds of correspondence are typical.

Participle → Adjective
(1) a. Greek  Examples (1)-(3) illustrate three ways in which Gothic innovates adjectives by substituting them for words belonging to other lexical classes in the Greek, with no functional changes being invoked. In (1) the Greek present participle of the verb zaō 'live' and the Gothic adjective qius 'living, alive' are postnominal attributes; in (2) the relevant Greek and Gothic elements are both copula complements; in (3) the relevant Greek and Gothic elements are both complements of the preposition 'against'. The Gothic innovations in the above examples are only of trivial interest, however, as they merely illustrate that identical semantic concepts can be distributed across the lexicons of different languages in different ways-i.e. an adjective of one language need not be realised by an adjective in another, and so on. What is more important, however, is that all these innovations occur without any changes in the syntax, as the relative linear ordering of all the key elements in the translation remains the same.

(ii) Change in function/relations
Similar to the above examples, deviations in this class encompass instances of adjective use in Gothic that do not correspond to adjectives in the Greek. However, unlike (i), where the syntactic function of the adjective in Gothic coincides with that of the corresponding element in the Greek, here the change in word class concatenates with differences in the syntactic function of the elements and, consequently, their relations with other elements in the clause. The following are the most typical types of rendering. As examples (4) and (5) illustrate, Gothic replaces what is a single verb or participle in Greek with an adjective and a copular verb, the adjectives being assigned the function of predicate complements to the copulae. Example (5) is particularly interesting, as in it Gothic innovates a relative clause where Greek has a postnominal attribute. Although renderings with the opposite ordering of the adjective and verb are also attested, the pattern illustrated in (5), with the verb following the adjective, is by far the most common one. This type of deviation is valuable, as it illustrates a tendency for verbs to appear clause-finally in Gothic. 3 In the Latin version, example (4) matches the Greek, which suggests that the copula is a Gothic innovation; the Latin version of (5) has absconditum fuit lit. 'hidden was', which leaves open the possibility that the Gothic copula in this verse is not original. What examples (6)-(8) share is that in them the Gothic adjective translates a noun or an element of a noun in the Greek; in each case, the change in lexical class is accompanied by the assignment of a new syntactic function. Thus, in (6) the Greek possessive noun phrase turns into a postnominal attributive adjective in Gothic. In (7) Gothic spells out a Greek compound, in which the element kalli-represents the adjective 'good' that modifies the head element elaia 'olive tree'. The point is, however, that in Gothic the attributive relationship is effected at the level of the phrase, with the adjective being assigned the relevant morphosyntactic features through agreement. Example (8) represents a paraphrase realised through a change of lexical class and syntactic function. Unlike (6), in which the Greek genitive noun is the source of the Gothic adjective, in (8) the source of the Gothic adjective is the head of the Greek noun phrase, and the Greek possessive genitive becomes the head of the Gothic noun phrase. What is important is that all these changes are effected without compromising the relative ordering of the elements.

Verb/Participle
In the pre-Vulgate Latin version (as well as the Vulgate), example (6) reads gloriam hominum 'glory of men', and (8) reads altitudinem terrae 'depth of earth'. The changes in Gothic cannot therefore be ascribed to Latin influence in these verses. The Latin version of (7) matches Gothic in its departure from the Greek model: in bonam olivam 'into a good olive'. This, however, does not necessarily mean that the change in Gothic is inauthentic, as neither Gothic nor Latin appear to have had a compound to match the Greek one. Thus, to spell out the Greek compound as a noun with an attributive modifier was the only available solution in both languages.
Adjective → Adjective + Noun (9)  In example (9), Gothic translates a Greek predicative adjective by means of a noun phrase involving an attributive adjective. As a whole, however, the Gothic noun phrase is also predicative. Whereas the Gothic translation spells out the Greek compound (theo-'God' and pneustos 'inspired'), neither the lexical class affiliation nor the functions of the components coincide between Greek and Gothic. Similar to the previously discussed deviations, the changes in Gothic are realised within the same slot assigned to the adjective in the Greek, and the ordering of the elements in the Gothic noun phrase matches that in the Greek compound. In the Latin version, the Greek compound is translated as divinitus inspirata 'divinely|ADV| inspired|PTC|', which indicates that the Gothic version is authentic. As with (7), the Greek compound being rendered with two words is dictated by the internal constraints of Gothic and Latin independently of each other. In this example, the Greek adverb esōthen is rendered by means of a (substantivised) adjective in Gothic, which is assigned all the relevant morphosyntactic features and, in conjunction with the definite article, performs the role of the subject of the clause. As before, no changes in the syntax have occurred. Although there is some disagreement between different pre-Vulgate witnesses, at least one version matches the Gothic in its use of an adjective: sed interior renovatur lit. 'but internal is renewed'. In (11), a composite Greek structure is rendered by means of an adjective in Gothic. In particular, samafraþjis* 'of the same mind' is a compound predicative adjective that translates a Greek participle with a preposed direct object. Significantly, the Gothic compound incorporates in its structure the notional elements of the Greek composite: 'one' + 'think', where 'one' implies 'the same'-hence Gothic sama-(however, cf. Miller 2018, §7.11). The Latin translation idipsum sentientes 'feeling/thinking that very same thing' is unlike Gothic, and is structurally similar to the Greek version: idipsum (< id 'it/that' + ipsum 'self') + sentientes 'perceiving|PTC|'.

Composite structure → Adjective
The existence of more authentic sama-compounds in Gothic (samalauþs* 'as much; equally great', samaleiks* 'agreeing together, consistent') that do not calque the Greek would seem to suggest that they were an idiomatic means of expression. If this is correct, then the Gothic departure from the Greek in this verse is truly remarkable, as it speaks to the translator making a Gothic-centred decision in aligning two Gothic compounds in the interest of smoothness, rather than opting for samasaiwalai, þata samo fraþjandans lit. 'of the same soul, thinking the same' (cf. attested þata samo fraþjan 'to be thinking the same' in Romans 15:5, Philippians 2:2). However, since this rendering was effected within the same slot occupied by the Greek composite and did not compromise the linearity of the verse, the end justified the means.

(II) Lexical innovations
In a relatively small number of instances, Gothic innovates lexical material (as individual items or, sometimes, components of compound words) that does not exist in the Greek version. In (12), the Greek noun halusis 'chain' is matched by a composite structure of a noun and a postposed attributive adjective (dat.) naudibandjom eisarneinaim 'iron chain' (lit. 'iron constraining bond') in Gothic, with the adjective being unattested in any Greek or Latin version. It is highly probable that this Gothic rendering (also attested in Mark 5:3) 4 represents an attempt at descriptive translation where the lexical resources of Gothic were insufficient. That this translation was ad hoc, and perhaps even somewhat awkward, also follows from the omission of the adjective eisarneina* upon second mention of 'chains' in the same verse: having clarified the sense of naudibandi* as a metal chain (rather than merely a bond or constraint), the adjective became superfluous in the anaphor, the onefor-one rendering being syntactically more preferable.
Outside the interpolation of the adjective, Mark 5:4 showcases more interesting examples of descriptive translation. In particular, the two successive mentions of the noun pede 'fetter' receive complex descriptive renderings: (dat.) eisarnam bi fotuns gabuganaim 'irons bent around feet' and ana fotum eisarna 'irons on feet'. Similar to the translation of Greek halusis 'chain', pede 'fetter' also appears to refer to a lexical gap in Gothic (so Falluomini 2015, 86). 5 The first rendering is syntactically rather more 4 In 2 Timothy 1:16 naudibandi* occurs on its own, without an attribute. In Luke 8:29, the Greek noun halusis 'chain' is rendered as eisarnabandi* 'chain' (lit. 'iron bond') in Gothic. There are 7 attestations of the noun bandi in the Gothic texts (Mark 7:35, Colossians 4:19, Luke 8:29, 2 Timothy 2:9, Philippians 1:14, 17, Philemon 13); in all cases it translates the Greek noun desmos 'bond', with a broad contextual sense range of 'bond, constraint, impediment, imprisonment'. Thus, on its own bandi evidently required specification when translating Greek halusis 'chain'. However, it is clear that each of the above Gothic renderings of 'chain' was ad hoc and neither was therefore fully adequate. 5 It is possible that these lexical gaps represent cultural realia that refer to items unfamiliar to the Gothic people. Falluomini (2015, 86) points out that in Luke 8:29 (dat.) pedais 'fetters' is rendered by means of the compound fotubandjom 'foot bonds'. As a whole, the above incompatible differences in translation technique (compounds in Luke vs. periphrastic renderings in Mark) are suggestive of different translators at work on different parts of the Gothic text. On the possibility that multiple translators were involved in the Gothic Bible project, see Ratkus (2018). complex than the anaphoric second one. As with the translation of 'chain', having explained the precise reference of eisarni 'irons', the translator opts for a lighter, albeit internally (syntactically) more awkward, second rendering. The modification of the second rendering is additionally motivated by the need not to interfere with the linear OV sequence of eisarna gabrak 'broke the irons', which would otherwise have been disrupted and made the clause less intelligible.
In (13), the Greek predicative adjective taipeinos 'humble' is rendered in Gothic with the adjective hauns 'humble, lowly', but with the difference that Gothic inserts a copula (im 'am'), which follows its adjective complement. Linguistically, the result is comparable to (5) above, which suggests a tendency of clause-final verb placement. However, although most Vetus Latina manuscripts match the Greek in having no copula, one Vetus manuscript (Cod. 77) and Jerome's Vulgate attest humilis sum lit. 'humble am'. As a result, one cannot be fully confident that, in this instance, the inserted Gothic copula is an authentic Gothic-centred operation. 6

(III) Word order
Although, as a general rule, Gothic tends to abide by the linearity of the Greek model, occasionally Gothic deviates from the Greek in the arrangement of elements at various levels of sentence structure. Compare examples (14) and (15). Examples (14) and (15) illustrate two types of change in word order that are significant for the present description: the placement of an adjective with respect to its head noun and the placement of a copula with respect to its adjective complement. In either instance, however, the reason for the change in linearisation is unclear. In particular, example (14) shows that Gothic not only eliminates the Greek polydefinite structure (as is to be expected), but also places the adjective in pre-position to the noun. Latin influence cannot be implicated, as the Latin Vulgate attests a postnominal adjective (VLD has no Vetus witnesses): Spiritum Sanctum Dei lit. 'Spirit Holy of God'.
It is possible, though unverifiable, however, that the change was internally motivated by the avoidance of stacking two postnominal modifiers, aimed at linear adjacency of concepts that belong together (ahman gudis ahman þana weihan gudis lit. 'spirit that holy of God').
In (15) the copular verb is moved from post-position to pre-position to the adjective. There appears to be no obvious motive for this change, which goes against the pattern discussed with reference to examples (5) and (13). As with example (14), Latin influence is ruled out by the fact that the linearisation of the Latin version matches the Greek model (utilis est lit. 'beneficial is'). Based on the evidence of the verse, the only conceivable explanation for the change in Gothic is harmonisation of the second mention of 'is useful' with the linearisation of the first one. The rhetorical nature of both of the above verses, with the direct address of (14) and rhetorical repetition of (15), may have been amongst the circumstances that conduced to changes in their linear structure.

IMPLICATIONS AND CONCLUSIONS
Based on the assumption that a translation, however literal, can serve as a valuable source of information on the properties of the target language, in this paper I set out to compare aspects of the linear structure of Gothic against the background of the Greek model. More specifically, I performed a basic classification of the how adjectives in Gothic deviate from the ways in which they are realised in the source text(s). Of the 2,056 examples examined, 1,636 match the Greek, which amounts to 80% of the total, with 420 examples (20%) departing from the model in one of the above ways.
Although the share of deviations is substantial, the figure of 20% is not meaningful in and of itself, because much depends on the type of deviation at issue. Most of the deviations have to do with different grammatical material used in the Greek version to represent what Gothic renders by means of an adjective. The difference in material may or may not entail a difference in syntactic function or relations. The least controversial type of deviation is the one illustrated in examples (1)-(3), where Gothic adjectives are matched by a participle, a noun and a pronoun in the Greek version, with no impact on the element's syntactic role within the clause or phrase. As noted above, this type of deviation refers to trivial differences of lexical distribution between languages: what is a noun or an adjective in one language need not be realised as the same word class in another.
Occasionally, the use of different grammatical material involves a change in syntactic relations, as illustrated in (4) -(11). Regrettably, most of the patterns illustrated do not seem to supply a window of inference into the idiom of Gothic syntax owing to the principle of structural equivalence carefully observed in the Gothic translation: formal deviations are tolerated within the slot assigned for a given word, every effort being made not to stray from the linear structure of the original. In particular, the change in syntactic relations illustrated in examples (5)-(11) has no effect on the relative ordering of the elements, which matches that of the Greek model. This evidence confirms, yet again, that the strategy underlying the Gothic translation technique had the lemma as its main focus: "The unit of translation was the word, and what subjectivity of treatment the Goth allowed himself was thus confined, in the main, to individual renderings" (Friedrichsen 1926, 246). The only exception is example (4), where the use of different grammatical material combines with an innovated copula, supplying an inference into aspects of Gothic verb syntax. The authenticity of this find is further confirmed by the evidence of lexical innovations and changes in word order, as illustrated in (13) and (15).
Syntactically, the most informative types of deviation are lexical innovations and changes in word order, as shown in (12) and (14). Similar to the above, they are effected within the slot assigned to the corresponding element(s) in the Greek model. However, the difference is that the adjective can be innovated on either side of the noun being rendered, the translator being unconstrained by the linearisation of the original. The same assumption of idiomaticity applies to instances of change in word order, where the translator departs from the linearisation of the original in favour of the idiom of the target language. The scarcity of these examples, however, does not permit sweeping conclusions and calls for a large-scale study of Gothic linearisation aimed at a statistically better informed judgment. 7 That the early Latin translation(s) of the Bible had some degree of influence on the Gothic translation has always been known. 8 In order to allow for the possibility of Latin influence, the above discussion makes references to instances of Gothic deviations matching the Latin version of the Bible. However, as has been noted, there is no reason to assume, where the Gothic and Latin versions deviate from the Greek in similar ways, 7 The few extant studies of Gothic linearisation (e.g. Koppitz 1900, Lenk 1910, Cuendet 1929, Smith 1971 suffer from a combination of limited or selective sampling of data and incorrect or limited comparative discussion; full-scale and methodologically adequate comparative studies of Gothic linearisation are nonexistent. 8 For a discussion of Latin influence on the Gothic text, see Friedrichsen (1926, 169-211;1939, 172-231;1961, 66-89), Burton (2002), Falluomini (2015, 105-114), Miller (2018) and the references there. For arguments against wholesale assumption of Latin influence, see Ratkus (2018). that the deviations are not independent of each other. This reasoning applies equally to differences in material, lexical innovations and differences in word order.
Finally, it is perhaps unnecessary to say that Bennett's (1980) and Peeters's (1985) irreconcilable approaches to Gothic comparative studies, as discussed above, each represent an extreme point on a continuum ranging from the Gothic text being syntactically vacuous to fully idiomatic. Although it is evident that Gothic follows the Greek model very closely, the considerable overlap between the original and the translation is a measure of the limits to which the possibilities in the target language can be pushed. In other words, the Gothic text faithfully following the Greek model conflates what was truly idiomatic with what was marginally acceptable. Any statistically wellsupported patterns of innovation and deviation from the Greek model likely represent native properties of Gothic that go beyond the limits of what was marginally acceptable. The result is an aggregate of linguistic patterns that cannot be easily teased apart in a linguistic description and each of which must be addressed on a case by case basis.
Thus, although it would perhaps be correct to assume that a native speaker of Gothic would have, for the most part, had no trouble understanding the word-forword Gothic translation of the Bible, 9 it does not automatically follow that it is a fully authentic and linguistically adequate sample of the language (cf. Stutz 1966, 48;Falluomini 2015, 67). Nor does this mean, however, that the Gothic translation is unusable as a source of evidence on Gothic syntax.
This assumption of intelligibility draws on the premise that the Gothic translation was a homogeneous text, consistent in its use of translation technique and exhibiting a uniform linguistic fabric and exegetical standard. This, in turn, assumes the existence of a single translator or a meticulous editorial process. However, as argued by Ratkus (2018), the variations in translation technique as well as the linguistic and exegetical inconsistencies observed across the Gothic Bible implicate several translators. Recognition of this important factor highlights some irregularities and discrepancies in the translation of the same concepts or the occasionally awkward use of grammar (for instance, see the discussion under "Lexical innovations" in II) as circumstances emanating from individual creative decisions or even differences in the degree of care exercised by different people. It is therefore plausible that some individual passages of the Gothic text may have been more difficult for a native speaker of Gothic to make sense of than others.