The Periphrastic Subjunctive in the Old English Multiple Glosses to the Lindisfarne Gospels

The recessive nature of the subjunctive as a formal category in Old English is witnessed in the use of alternative grammatical structures other than inflectional subjunctives in contexts of non-fact modality. An increasing analytic reliance on grammatical devices signalling nonfact modality (e.g. gif ‘if’, sua hua ‘whoever’, etc.) both fostered and facilitated the occurrence of the indicative in such contexts. The modal verbs magan, *sculan and willan served as fully independent verbs in Old English, but even during the Old English period there appears to have been a ‘modern’ tendency to use modal constructions involving a (subjunctive) modal + infinitive construction, instead of an inflectional subjunctive, with little (or no) underlying sense of non-modal notional meaning.1 In the Old Northumbrian (ONbr) interlinear gloss to the Latin text of the Lindisfarne Gospels (British Library, MS Cotton Nero D.iv; henceforth Li), the increasing lack of direct correspondence between contexts of non-fact modality and the subjunctive in Old English is attested in the widespread tendency for present-indicative forms in -s and -ð to supplant subjunctive forms, e.g. 7 swiðe bebead him þaette hia ne aewades l mersades hine L et uehementer comminabatur eis ne manifestarent illum ‘And he very much commanded them that they should not make him known’ MkGl (Li) 3.12, or for indicative forms to alternate with subjunctive forms, which in ONbr ended in -a/-e/-o in both

The recessive nature of the subjunctive as a formal category in Old English is witnessed in the use of alternative grammatical structures other than inflectional subjunctives in contexts of non-fact modality. An increasing analytic reliance on grammatical devices signalling nonfact modality (e.g. gif 'if', sua hua 'whoever', etc.) both fostered and facilitated the occurrence of the indicative in such contexts. The modal verbs magan, *sculan and willan served as fully independent verbs in Old English, but even during the Old English period there appears to have been a 'modern' tendency to use modal constructions involving a (subjunctive) modal + infinitive construction, instead of an inflectional subjunctive, with little (or no) underlying sense of non-modal notional meaning. 1 In the Old Northumbrian (ONbr) interlinear gloss to the Latin text of the Lindisfarne Gospels (British Library, MS Cotton Nero D.iv; henceforth Li), the increasing lack of direct correspondence between contexts of non-fact modality and the subjunctive in Old English is attested in the widespread tendency for present-indicative forms in -s and -ð to supplant subjunctive forms, e.g. 7 swiðe bebead him þaette hia ne aewades ł mersades hine L et uehementer comminabatur eis ne manifestarent illum 'And he very much commanded them that they should not make him known' MkGl (Li) 3.12, or for indicative forms to alternate with subjunctive forms, which in ONbr ended in -a/-e/-o in both the singular and plural, e.g. gif gie habbas ł haebbe leafo L si habueritis fidem 'If you have faith' MtGl (Li) 21.21 and þaette gie eta 7 drincga [...] 7 gie sittað ofer heh sedlo L ut edatis et bibatis [...] et sedeatis super thronos 'That you may eat and drink [...] and sit on high thrones' LkGl (Li) 22.30 (see Cole, 2014 for extended discussion). 2 Periphrastic modal verb + infinitive constructions also occur in Li to translate the Latin subjunctive mood, as in 7 sohton ða hehsacerdas 7 ða uðuuto huu hine mið facne gehealdon ł maehton hia gehalda 7 ofslogon ł hia maehton ofslaa L et quaerebant summi sacerdotes et scribae quomodo eum dolo tenerent et occiderent 'and the chief priests and the scribes sought how they might with wile lay hold of him and they might kill him' MkGl (Li) 14.01. 3 The glossator's reliance on structures other than the inflectional subjunctive is possibly all the more surprising given the requirements of the glossing process to render the Latin as atomistically and faithfully as possible in the target language.
The present paper examines the glossator's use of the modal + infinitive construction in Li in relation to that of the inflectional subjunctive and indicative simple verb forms. The strategy of using modal + infinitive structures in the translation of Latin subjunctives in Li has long been noted (Bosworth & Toller 1898, magan), yet data drawn from the ONbr glosses are not included in any of the major studies on Old English modal verbs (cf. Gorrell 1895;Standop 1957;Ogawa 1989). Modal + infinitive constructions in Li occur frequently (though not exclusively) in multiple glosses whereby a single Latin lemma is rendered using at least two Old English glosses, separated by Latin vel 'or' (abbreviated to ł in the manuscript). Multiple glosses conveniently facilitate the comparison of the forms that occur in identical contexts of non-fact modality in ONbr. Certain properties of the multiple glosses 2 The abbreviations used in this paper to refer to the Lindisfarne Gospels (Li) and the West Saxon Gospels ( may also function as a diagnostic for evaluating the status of modal + infinitive constructions in Old English.

The periphrastic subjunctive in Old English
The extent to which modal + infinitive constructions functioned interchangeably with inflectional forms as periphrastic expressions of mood is a vexed question that has received a great deal of attention in the literature (Gorrell 1895;Standop 1957;Krzyszpién 1980;Mitchell 1985;Goossens 1987;Ogawa 1989). The need for caution in too readily interpreting modal verbs as grammatical circumlocutions for the inflectional subjunctive has been reiterated in the literature: modal verbs often present little loss of primary meaning and occur in the subjunctive under the same circumstances that trigger its use with other verbs (Mitchell 1985(Mitchell , § §2971-2980. The terms 'modal' and 'modal auxiliary' are in themselves problematic, given that these verbs functioned as independent lexical items in Old English and the issue of whether magan, *sculan and willan expressed mood at this early stage is controversial. I follow Mitchell (1985, §991) in using the label 'modal auxiliary' for want of a better term, but in full awareness of the potential prolepsis involved in employing the terms 'modal' and 'auxiliary' with regard to the function of these verbs in Old English. From a historical point of view, it has been argued that in the initial (Old English) stage of the history of the subjunctive and the 'modal auxiliaries' in English, modal verbs with an infinitive were treated as grammatical equivalents to inflectional subjunctive forms (Gorrell 1895). The growing tendency in the language to use auxiliary constructions was triggered and propagated by the breakdown in the formal distinction between the indicative and subjunctive inflectional forms of verbs. Gorrell's examination of the frequency of Old English 'modal auxiliaries' in relation to that of inflectional subjunctives in indirect discourse indicates a striking increase in the use of the periphrastic construction with a distinct tendency to occur when the corresponding inflectional forms of verbs would prove ambiguous (1895: 458).
Observations in the literature that the use of a periphrastic subjunctive was fostered by the breakdown in the formal distinction between the indicative and subjunctive mood is particularly pertinent to the ONbr gloss. One of the main characteristics of the ONbr texts is the advanced state of morphological simplification across the verbal system caused by various processes of reduction and levelling, including the proliferation of the northern present-tense marker -s at the expense of -ð, and the early loss of final -n, most notably in the infinitive and present-plural subjunctive, but also in the preterite-present plural verbs and preterite indicative and subjunctive (see Cole 2014). The preterite subjunctive and indicative are more often than not indistinguishable from each other as the preterite plural subjunctive shows preterite-indicative -on endings rather than -en and indicative forms occur with subjunctive -en, which suggests the coalescence of the endings in [-ən].
The view of the history of the modal auxiliaries as a simple chronological development whereby modal + infinitive constructions increasingly occur as a subjunctive equivalent as the Old English period progresses is belied, however, by the textual variation that the incidence of modal verbs exhibits in the Old English period. Ogawa's (1989) survey of Old English texts indicates a varying incidence in the use of the periphrastic construction across different text types and time periods that challenges what he terms the 'substitution theory', the notion of the history of the English subjunctive as characterised by the use of modal auxiliaries to compensate for the loss of inflectional morphology. His analysis of the meanings of the modal verbs in subjunctive contexts does not show the degree of loss of semantic meaning which would be expected if the verbs were being used as auxiliaries. Instead modal verbs are employed to convey a particular shade of meaning as required by the context: their use adds a nuance that is not explicit in the inflectional subjunctive. The suggestion that the periphrastic construction is not a mere grammatical alternative to the inflectional subjunctive but expresses instead a semantic nuance not found in the simple subjunctive is in line with Standop (1957: 169) and Krzyszpién (1980). The latter argues that periphrastic magan + infinitive and the inflectional subjunctive were not wholly semantically interchangeable. The use of one form or the other was determined by a difference in meaning: inflectional subjunctives denoted general non-fact modality whereas the periphrastic magan + infinitive expressed a particular aspect of non-fact modality, objective possibility. When magan itself occurred in the subjunctive inflectional form "general non-fact modality was 'superimposed' on the narrower meaning of objective possibility" (Krzyszpién, 1980: 51).
Ogawa's findings, in particular, emphasise the relevance of text type and stylistic factors in determining the occurrence of the modal verbs; there is a preference for poetry as opposed to prose to favour the use of modal verbs across the entire Old English period, probably as a stylistic poetic device that added emphatic detail (Ogawa 1989: 231-232;Gorrell 1895: 458). In prose, modal verbs are favoured in argumentative religious and philosophical writings and homiletic literature rather than in narrative prose. Text type and subject matter are also found to affect the individual incidence of each modal verb, e.g. magan and *sculan are common in didactic and religious writings where the latter commonly serves to emphasise commands, whereas narrative prose works show a preference for willan (Ogawa 1989: 235). Interestingly, the West Saxon Gospels (Corpus Christi College Cambridge MS, 140; henceforth WSCp) exhibit an "almost entire neglect of the periphrastic [modal] forms" (Gorrell 1895: 458). The low incidence of modal verbs in WSCp is attributed by both authors to the translator holding "slavishly" to the Latin original (Gorrell 1895: 458; see also Ogawa 1989: 225, 236). Verse translations in the Paris Psalter and the Kentish Psalm reflect a similar reluctance to employ periphrastic forms (Ogawa 1989: 236), which tells in favour of a close degree of dependence on the Latin source inhibiting the use of periphrastic [modal] forms. The avoidance of periphrastic forms involving modal verbs found in WSCp does not, however, find a parallel in the interlinear ONbr glosses in Li; despite the glossarial nature of the text type, the glossator's language is not as subjugated to the demands of atomistic glossing as might be expected. 4

Modal + infinitive and the subjunctive mood in Li
Li is consistent with general Old English usage in its employment of the modal + infinitive compound instead of an inflectional subjunctive form of the verb in indirect discourse after verbs of thinking and believing and in the employment of willan with expressions of promise (Gorrell 1895: 449-455), exemplified in (1a) and (1b), respectively. Such usage overlaps broadly with the future-in-the-past employment of the preterite subjunctive (Mitchell 1985: §646) found in (1c). It should be noted that (1d) is the only instance in which *sculan is used periphrastically to translate a Latin subjunctive in Li (Kotake 2006: 44). Ogawa's (1989: 235) observation that *sculan commonly serves to emphasise commands would seem to find exemplification here. 5 (1)a. wiste forðon huoelce uoere seðe salde hine ł ualde hine sealla ' The construction magan + infinitive is used in Li in both the present and preterite in clauses of purpose, a usage that is not peculiar to Li, but which differentiates it from WSCp where magan is never used in purpose clauses: there is a preference for willan instead (Ogawa 1989: 236). In the case of (2a) and (2b), the inflectional subjunctive gesii alternates with (subjunctive) magan + infinitive. Behre (1934: 92, fn. 1) states that the subjunctive form of magan + infinitive is the only way of unambiguously expressing subjunctive mood in the first-person singular given the lack of formal distinction between the first-person present-indicative and subjunctive. In ONbr gesii is a subjunctive form (Ross 1937: 133), which effectively eliminates morphological ambiguity as a motive for including maege + infinitive alongside gesii in the double gloss. Standop (1957: 60-61) suggests that magan expresses a different kind of uncertainty from the subjunctive of a simple verb. Similarly, as previously mentioned, Krzyszpién (1980: 51) argues that subjunctive magan + infinitive is not a mere circumlocution for the subjunctive but contributes semantic precision by expressing a particular aspect of non-fact modality, that of objective possibility. The inclusion in Li of both an inflectional subjunctive and maege + infinitive would seem to corroborate this view. Behre's hypothesis may hold nonetheless in preterite contexts where the simple verb form is formally indicative but indistinguishable from the subjunctive in speech (example 2c).
(2)a. c. monigo forðon he gehaelde ðus þaette hia raesdon on him þaette hine hie gehrindon ł hrina maehtaes 'for he healed many, thus that they pressed upon him that they might touch him' multos enim sanabat ita ut inruerent in eum ut illum tangerent MkGl (Li) 3.10 Clauses of indefiniteness involving sede and sua hua, sua huelc 'whoever', suaehuaelc 'whatever', etc. generally required a subjunctive verb form in Old English, although the indicative became increasingly common as the period progressed (Visser 1963(Visser -1973. Indicative and inflectional subjunctive forms both occur in indefinite contexts in Li, e.g. in suahuelcum hus gie inngae L et in quamcumque domum intraueritis 'in whatever house you enter' LkGl (Li) 9.4; on sua huelcne hus gie ingaeeð L in quamcumque domum intraueritis 'in whatever house you enter' LkGl (Li) 10.5. There is also one instance of a double gloss consisting of both an inflectional subjunctive and an indicative, e.g. sua huelc iuer haebbe ł haefeð friond L quis uestrum habebit amicum 'whoever of you has a friend' LkGl (Li) 11.5. The Li glossator also avails himself of a further strategy in double glosses and renders the Latin subjunctive using a (subjunctive) modal + infinitive construction. In double glosses present indicative ł welle + infinitive is the particular combination of grammatical forms that is employed in Li to translate the Latin subjunctive mood in clauses of indefiniteness. This strategy is illustrated in (3) In double glosses, therefore, there is a preference for the modal periphrastic construction, as opposed to an inflectional subjunctive, to occur alongside a present-indicative form in clauses of indefiniteness. Clearly, both forms were acceptable grammatical alternatives in this context, although it should be noted that the subjunctive modal form welle is used and willan here retains an implied element of volition, intention or acquiescence. Given that the indicative also occurs in this context, the use of welle + infinitive may be an attempt to vivify the sense of modality inherent in clauses of indefiniteness as to the person, place, time referred to in the clause. The modal's primary meaning has not been eliminated, but such usage provides an insight into an intermediate stage in the development of the English periphrastic subjunctive.
The apodosis of a hypothetical condition in ModE requires a preterite modal: in fact its occurrence in this environment serves as a test of modal status (Denison 1993: 313). In Old English the preterite inflectional subjunctive generally occurred in this context, e rally 'if God were your father indeed you loved me' Jn (WSCp) 8.47, where lufedon, though formally preterite indicative, is to be understood as a subjunctive. In Li, however, preterite forms of willan + infinitive occur in the apodosis of a hypothetical proposition, compare, gif god faeder iuer woere gie ualde lufiga uutedlice mec 'if God were your Father, you would indeed love me' L si deus pater uester esset diligeretis utique me Jn (Li) 8.42. Visser (1963Visser ( -1973, §1607, §1672) records such usage in Old English with should and might, but cites no examples of would for Old English, apparently dating the emergence of would in the apodosis of a hypothetical proposition to the early ME period. Instances of wolde in the apodosis of a conditional would appear to exist in Old English, however: Ogawa (1989: 131) identifies an example of wolde used with an inanimate subject in AElfric's Lives of Saints (AELS 31.672) that "may well serve as a pure expression of an imaginary event in the past". The examples in (4) illustrate the occurrence of walde in the apodosis of a hypothetical condition in Li. Here willan serves to express an intermediate idea between simple volition and an imaginary past result: it cannot be said that walde is entirely non-volitional in nature, but it does approximate 'modern' usage in conveying the probable past results of an unreal past condition. The examples also succinctly exemplify the lack of formal opposition between the indicative and the subjunctive in the preterite that may well have contributed to the glossator's decision to signal non-fact modality more explicitly using a periphrastic form.
As previously noted, periphrastic subjunctives occur frequently in double glosses in Li, which, given the preference for multiple glosses in Li is not in itself significant. 7 Nonetheless, the double-glossing translation technique provides a unique insight into what type of grammatical units co-occurred in identical contexts. Multiple glosses in Li generally involve two items, though triple and even quadruple glosses also occur (see examples 1d and 3f in the present paper, in which the triple glosses gehened ł geniðrad ł geteled and loses ł fordoes ł forfaeras translate Latin contempnatur 'scorn, despise' [3sg pres.subj.pass.] and perdiderit 'lose, destroy' [3sg perf.subj.act.], respectively). The multiple glosses vary with regard to the type of information that they supply; some provide lexical alternatives for a single Latin lemma, involving the use of synonyms or near-synonyms, e.g. berað ł bringeð L adferte 'bring!' JnGl (Li) 21.10. Other double glosses provide grammatical alternatives for a single Latin lemma, e.g. geseað ł geseas L uideritis 'you see' LkGl (Li) 21.20, ne habbas ł nabbas L non habent 'they do not have' MtGl (Li) 14.16, or supply both a grammatical and a lexical alternative, e.g. gae we ł wutu[m] geonga L eamus 'Let's go!' MkGl (Li) 1.38.
In double glosses translating the subjunctive mood the preferred strategy in Li is to juxtapose an indicative with a (subjunctive) modal + infinitive, rather than an inflectional subjunctive. It would be an oversimplification to infer that the occurrence of the periphrastic construction alongside simple forms in double glosses is in itself proof that the modal + infinitive structure is a mechanical substitute for the inflectional subjunctive. Double glosses in Li clearly provide alternatives that are equally acceptable in a given context but they convey differing nuances. Double lexical glosses generally involve an item that introduces a semantic nuance, such as berað ł bringeð L adferte 'bring!' JnGl (Li) 21.10, cited above, or a contextualised nuance. Pons-Sanz (forthc.) cites the rendering of L puella as dohter ł maegden at MkGl (Li) 5.41. The Latin lemma puella would normally be translated using OE maegden 'girl, maiden', but here, Pons-Sanz argues, OE dohtor 'daughter' is included because the girl referred to is the daughter of the leader of the synagogue mentioned a few lines previously. Even alternative grammatical glosses, such as geseað ł geseas L uideritis LkGl (Li) 21.20 and ne habbas ł nabbas L non habent MtGl (Li) 14.16, which appear to present simply morphological variants, provide an additional form that is more idiomatic or colloquial, such as the dialectal ONbr second-person plural form geseas alongside geseað 'you see' at LkGl (Li) 21.20, or nabbas instead of ne habbas at MtGl (Li) 14.16. Similarly, the double glosses under scrutiny involving 'modal auxiliary' constructions are a complex case that do not necessarily involve grammatically interchangeable forms with no difference in meaning. Careful analysis suggests that the periphrastic modal subjunctive is a grammatically acceptable -if somewhat stylistically different -alternative to the inflectional subjunctive.
With regard to the ordering of the items in double grammatical glosses in Li, i.e. whether the glossator chooses to place a term in first or second position, Kotake (2006) identifies a considerably consistent pattern in the ordering of the alternative grammatical glosses, including those that translate Latin verbs forms conveying future tense and subjunctive mood. In the majority of cases, the double glosses consist of a simple form in first position followed by a periphrastic construction. He attributes this ordering preference to the glossator placing the Old English grammatical category that is morphologically closest to the Latin first, followed by the "more morphologically marked" periphrastic construction (2006: 44-46). I interpret the grammatical unit that most deviates from the Latin original as the more idiomatic or colloquial item in the doublet. The translation technique of double glossing provides the glossator with the scope to include one Old English translation that is atomistic and closely parallels the Latin morphologically and another that distances itself from the original Latin text and provides a more concrete or colloquial rendering. Such an interpretation both corroborates and finds support in Ogawa's contention that the modal verb phrase probably reflects colloquial Old English usage. Ogawa notes the tendency for modal verbs to occur more readily at the colloquial rather than literary level of Old English prose and in direct speech and suggests that "the colloquial level of style favours the use of modal verbs in its endeavour to make description vivid and concrete" (1989: 237-238). The colloquial sphere of usage is also a locus of grammatical change where the 'modern' periphrastic future and subjunctive forms would be expected to make their first appearance.

Summary
The data in Li provide further insight into an intermediate stage in the history of the periphrastic subjunctive whereby modal verbs were used initially to emphasise a particular aspect of non-fact modality that an inflectional subjunctive could only do more generally and which indic-ative forms left unexpressed. The notion of the history of the English subjunctive as characterised by the use of modal auxiliaries to compensate for the loss of inflectional morphology only tells part of the story of the English periphrastic subjunctive: stylistic factors were of considerable importance. In a succinct summary of the Old English state of affairs, Ogawa (1989. 223-234) highlights that the varying distribution of modal verbs across text types points to no clear tendency for them to form the 'periphrastic subjunctive' when the corresponding simple verb form would be ambiguous with respect to mood. […] Although the distinction between the modal verb construction and the simple verb form is not always easy to explain, the former can be generally shown to stand for clearer, more concrete expressions, emphasizing and specifying, by the appropriate choice of a relevant modal verb, a particular nuance of desired relationship as the context requires it The status of the modal verb construction as a clearer, more concrete expression that emphasises and specifies is also in line with the double glossing technique found in Li. Double grammatical glosses generally provide one translation that is morphologically closer to the Latin and an additional gloss that is more idiomatic or colloquial, in this case, the modal verb construction.
Nevertheless, the type of grammatical structures that alternate in double glosses alongside modal verb constructions also suggest that the breakdown in formal opposition between the indicative and subjunctive played a role in the history of the subjunctive and the 'modal auxiliaries' in English. Double glosses such as þaette ic gesii ł þaet ic maege sea L ut uideam MkGl (Li) 10.51 indicate that maege sea is not a mechanical grammatical circumlocution for the subjunctive form gesii. But in cases such as ne ðe ue gesaldon ł nalde ue gesealla L tradidissemus JnGl (Li) 18.30, the lack of formal opposition between the indicative and the subjunctive in the preterite may well have fostered the use of the periphrastic structure as a more distinctive means of signalling non-fact modality; after all, walde + infinitive is the only way of unambiguously expressing non-fact modality given the coalescence of the preterite indicative and subjunctive. It is also important to bear in mind that even in the present where unambiguous subjunctive forms were retained for longer, modal + infinitive constructions co-occur mainly with indicative forms in double glosses rather than the inflectional subjunctive. The lack of direct correspondence between contexts of non-fact modality and the subjunctive (inflectional or periphrastic) in Old English and the increasing occurrence of present-indicative (or of morphologically ambiguous preterite) forms would also have fostered the use of periphrastic subjunctive forms. The glossator's preferred strategy in Li of juxtaposing an indicative with a (subjunctive) modal + infinitive, rather than an inflectional subjunctive militates in favour of Krzyszpién's (1980: 62) view that one of the functions of modal verb compounds was to serve as "a more distinctive and productive means of signalling non-fact modality" [be it a particular aspect of non-fact modality] left unexpressed by indicative forms.