Speech , Texts , and Choices from the Modal System : Mood Distribution in Old English Sermons

The present paper investigates the alternation between the imperative and the subjunctive mood in speech representations in Old English. Both imperative and subjunctive mood form relatively fully inflected and robust modal categories in Old English, and a systematic alternation is assumed between both in speech and writing in the period. The imperative is formally restricted to the second person. Thus, it seems plausible to assume a functional restriction as well, that is, a restriction in usage as a deictic (modal) element with direct reference to a particular addressee. As such, its most natural habitat may be speech and speech representations in writing. By contrast, the subjunctive, with a full formal inventory for person and number, seems much less restricted in terms of functions, and may therefore occur in both speech and writing,—or even preferably in writing, considering the various levels of abstraction and detachment that writing allows. To my knowledge, there are no previous studies devoted to the topic. Sermons lend themselves to such an investigation for two important reasons: they frequently report speech, e.g. when preachers relate to the congregation Jesus’ words to his disciples from the Bible, and as speech-based texts, they provide frequent instances of direct speech to the congregation. This forms another level of speech representation, which may be contrasted very fruitfully to the speech reported (almost exclusively) from the Bible. The study is corpus-based with a selection from the extant Old English sermon material. The aims of this exploratory study are mainly qualitative. However, in quantitative terms, I will show that the distribution of imperative and subjunctive in Old English sermons follows the various levels of speech representation in texts.


Introduction
Compared to later stages of the language, Old English still shows a rather complex mood system in the inflection of both strong and weak verbs.Alongside an unmarked indicative that is commonly described as realis or factive, imperative and subjunctive are marked modal categories.While the imperative is used exclusively in the second person and marks directives, the subjunctive has a formal inventory for all three persons and is more multifunctional.Attested uses range from marking hypothetical and unreal contexts to mandative, i.e. directive uses.This directive function it curiously shares with the imperative, but Traugott remarks that "we must assume that there was a difference in meaning" (1992: 185), since imperative and subjunctive contrast morphologically.Further, she notes that the subjunctive was used in monastic and legal regulations, charms, medical prescriptions "and similar generalised instructions" (1992: 185).It is far from clear, however, what triggers this preference in these types of discourse, and which typical environments the imperative prefers by contrast.It is the aim of this paper to shed light on the distribution of both moods in Old English, and to investigate the assumed differences in meaning and use that both provide.
I will show that the distribution of imperative and subjunctive in Old English sermons follows the various levels of speech representation in texts.Intuitively, it may be assumed that imperatives, with their inherent selection of second-person subjects, are restricted to "face-to-face interaction", as far as we can define such a form of communication from the written records.Subjunctives, with their larger formal inventory and, supposedly, greater flexibility of use, mark more abstract and detached discourse.Contrary to what may be expected, however, subjunctives are not typically found in reported speech, but seem tied to the level of direct speech representation.
My point of departure is Old English homiletic writing.Homiletic collections and sermon discourse lend themselves to this investigation for two reasons: first, sermons and homilies frequently report speech events, for example, when preachers relate to the congregation Jesus' words to his disciples from the Bible.I will refer to this level in sermons as reports of speech events.Secondly, sermons as speech-based texts provide frequent instances of, originally or intentionally, direct speech to the congregation.This forms a second level of speech representation, namely that of direct speech events.So Old English homiletic writing provides a large and rather diversified selection of "speech" that is not found in other records of Old English to quite the same extent.
The analysis is restricted to occurrences of imperative and subjunctive mood in independent clauses, since it is here that the functional overlap occurs.Mandative subjunctives are also frequently found in nominal complement clauses after suasive verbs and related constructions (e.g.michel is nydþearf manna gehwilcum þaet he Godes lage gyme-it is very necessary for all men that they observe God's law; Wulfstan, Sermo lupi).Also, the subjunctive is often used in dependent clauses after reporting verbs (e.g.se preost cwaeð þaet an wer waere on Irlande-the priest said that there was a man in Ireland; cited from Quirk and Wrenn 1957: 83).However, since these are not syntactic environments or functions that the subjunctive shares with the imperative, they do not contribute to the question of functional overlap and I will not comment on them further.
In the next section I briefly outline forms, functions and the syntactic particularities of both moods as described in the relevant reference grammars and handbooks.I then present findings from an exploratory pilot study of the distribution of both moods in Old English homiletic writing, with particular reference to direct speech events and reports of speech events.This analysis is intended as a diagnostic overview of mood distribution.However, it reveals a striking complementary distribution along the lines of speech representation, which seems far too regular to be mere chance.In my conclusions, I will put these findings into perspective and argue for a fresh look at the modal system of Old English that systematically allows for speech/writing contrasts and, more fundamentally, the various communicative constellations of speakers and hearers which this creates and enforces.

Imperative and subjunctive mood in Old English: Forms and functions
In Old English, imperative and subjunctive mood are formally distinct from one another in person and number in both the strong and the weak declensions (see Table 1).In addition, they are distinct from the indicative, if not exclusively in form, then at least on syntactic grounds, i.e. in terms of constituents and constituent order.Only the present tense is relevant, since the imperative is restricted to the present.For reference to potentially ambiguous forms, the indicative is given, too.* Strictly speaking, not all endings given here are morphological inflections.Some derive from phonological processes that retain older roots and themes (or mutated forms of these), e.g.retention of -h in the imperative of contracted verbs.See Campbell (1959: 297-298, 322) or Lass (1994: 174-177) for more detail.
Potential cases of indistinctness arise in the plural, where the imperative is identical with the indicative. 1 Here, the syntactic environment usually disambiguates, since imperatives often lack subjects, a fact strongly corroborated by the present study (see Section 3).In the subjunctive paradigm, the first person singular is identical with the indicative.However, first-person singular subjunctives do not seem to have occurred in actual usage (cf.Mitchell 1985: 373 and references given there).
Thus, the formal inventory of imperative and subjunctive is rather robust, even though "scribal errors" or "late confusion" always seem a possibility (cf.Mitchell 1985: 377).Only in a number of cases do forms overlap, most notably in the so-called irregular verbs, where the paradigm is defective to begin with, e.g.witan or don (see Quirk and Wrenn 1957: 53-58 for an overview).These were excluded from the analysis.
In actual usage, this neat distribution is sometimes obscured, most frequently by phonological processes.We must expect levelled forms when subject-verb inversion occurs and a first-or second-person plural or dual pronoun subject follows the verb (we, ȝe etc.).This particular syntactic environment usually results in the levelling of the verbal inflection (e.g.helpe ȝe), which makes it impossible to decipher the verbal mood as derived either from helpaþ (imperative) or helpen (subjunctive), or even indicative helpaþ for that matter.In the analysis, these cases were discounted unless a post-verbal subject pronoun we/ȝe clearly follows -aþ or -en in the verb.Even then, cases of aþ + we/ȝe could be either imperative or indicative, but in most cases, pragmatic considerations help to disambiguate: whenever a directive is issued, I have opted for an imperative reading on the grounds that it is the imperative which prototypically encodes a directive and not the indicative. 2he corpus used for the analysis contains the Old English sermons compiled in the Helsinki Corpus (HC) plus additions from the Dictionary of Old English Corpus (DOEC), and amounts to roughly 60,000 words.The additional sermons were selected from the orthodox preaching tradition of AElfric and Wulfstan, and also from the more popular, often unconventional anonymous sermons, i.e. the Vercelli Homilies and from smaller, less principled collections (cf.Amodio 2014 for Vercelli and Tristram 1970 for the single anonymous sermons selected for this study).Table 2 gives an overview of the material used: Wulfstan's at times heavily idiosyncratic style influences the representation of subjunctives in the present study to a considerable extent.His homily Be Christendome (WHom 10c, HC) contains more than half of all third-person subjunctives found in the data, while the other forms are hardly affected.For this reason, the text file was excluded from the study.This dataset may seem relatively small and therefore somewhat restricted in the analytical potential it provides.But as stated in the introduction, this study is diagnostic rather than determinative.It aims at providing an initial overview of mood distribution in Old English, and with an overall frequency of roughly 7/1,000w (414 tokens in total) imperative and subjunctive mood can be considered high frequency items in sermon discourse, which justifies a relatively small corpus.It may be rewarding at a later stage of this project to expand the investigation and consult the York-Toronto-Helsinki Parsed Corpus of Old English Prose.With a much larger database and automatic retrieval of the relevant forms, the YCOE allows, but also demands, consideration of a much greater range of pragmatic factors than speaker constellations and representations of speech in the analysis.This may corroborate and further specify the results obtained in the present investigation.
Before I proceed with the analysis, a brief comment on the status of the material selected for the present study as "texts from speech" and "'speech' in text" may be in order.Among historical data, sermons can be counted as one of the earliest, most robust and widely-used texts from speech.Whether a sermon was preached first and then written down or was first written down and then disseminated in writing to be preached is a question that is impossible to answer.Luckily, it is not one that needs to be settled here in order to evaluate the results.It is undeniable that sermons and homilies are texts in the close vicinity of spoken language, and, arguably, even the closest we can hope to get when investigating the Old English period. 3owever, on this issue, evidence from Wulfstan suggests that his homilies were certainly intended for oral presentation (see Amodio 2014: 111).AElfric may have envisioned a more multifunctional use for his Catholic Homilies, including oral delivery for the first volume as well as silent reading for various audiences for the second (cf.Godden 1973; Lensing 2010: 81f.; Amodio 2014: 122). 4For yet other collections, for example the Vercelli Homilies, it may be hypothesised that they served the personal interests of individuals, most likely within monastic contexts.These may have been intended for private devotional settings or for recitation during the monastic night office (cf.Greenfield and Calder 1986: 74;Liuzza 2001: 240).

Distribution of imperative and subjunctive in Old English sermons
In the following discussion, I first consider the general distribution patterns of imperative and subjunctive in representations of speech.Afterwards, I provide a more detailed analysis of the preferred constructions and functions of both.
Generally, I distinguish direct speech events and reports of speech events as two relevant representations of speech in Old English sermons.Direct speech events are those where a potential preacher addresses his audience directly, as in example (1).
[Dear men, understand that Christ is the lord of the Christians and [that] all Christian men are counted as Christ's limbs.](Wulfstan, de Baptismate, HC; the translations given for the examples cited are mine.) Here, Wulfstan directs his congregation to learn and bear in mind (understandað) that Christ is Lord and that all Christians are members of one body in Christ, one of the fundamental concepts in Christian theology.Even though I label this direct speech event, it must be borne in mind that this cannot be taken as a verbatim rendering of Wulfstan's actual utterance from any one particular preaching occasion.Rather, this directive occurs in the shape of a direct speech event in the text, as a representation designed to be disseminated, possibly re-enacted, at other such preaching contexts for other preachers and to different congregations.
Not all directives at the level of direct speech events, however, work in this way.In example (2), it is not quite as clear who AElfric addresses: (2) Nu cwyð sum man on his geþance: eaðe mihte he arisan of deaðe for ðan þe he is god: ne mihte se deað hine gehaeftan.gehyre se mann þe ðis smeað.answare his smeagunge.Crist forðferde ana on ðam timan.ac he ne aras na ana of deaðe: ac aras mid miclum werode.
[Now some man may have this thought: easily he might have arisen from death because he is God: death might not have taken him captive.
[May] the man who thinks so hear the answer to his thought.Christ departed at this time alone.But he did not rise from death alone.But he arose with many other men.](AElfric, Homily for Easter Sunday, DOEC AECHom I, 15) Here, AElfric, in a homily designed for Easter Sunday, anticipates objections to the Gospel's report of Christ's resurrection and so provides potential preachers with a proper answer to such an objection.Even though such directives are at the level of direct speech events, i.e. directed straight to the congregation, they are not intended for the totality of listeners at all possible preaching occasions, as is the case in example (1).Instead, the directive is issued in the third person (se mann þe ðis smeað) and thus only provides for the possible presence of a doubtful mind.Consequently, such forms accommodate potential listeners from within the audience and even beyond.By contrast, reports of speech events usually occur within a narrative.So they take place on quite another level of discourse.Example (3) illustrated this; here, AElfric narrates the account of Christ's resurrection from the gospel of Matthew.
( It is important to note that such reports of speech events include direct speech, too, but are yet distinct from direct speech events in the present study, because they do not address the congregation.The distinction of direct speech event and report of speech events is thus based on the speaker-hearer constellations in the utterance.It is a distinction made on the level of discourse participants and not on the level of formal representation, as, for example, in the concepts of "direct" and "indirect" speech (cf.Quirk et al. 1985Quirk et al. : 1021)).
Table 3 gives an overview of the mood distribution in both types of speech representation.Direct speech events show a clear preference for the subjunctive, with an average frequency of 2.03/1,000w.In only one third of all instances where the preacher addresses his congregation, are imperatives used.Reports of speech events to the congregation, by contrast, largely rely on imperatives (2.1/1,000w), and hardly ever use subjunctive forms.There are only two tokens of subjunctives in reports of speech events, which, moreover, have a function different from those found in direct speech events, as we will see later.Overall, example (2) is thus a rather typical representation of direct speech events in sermons, while Wulfstan's request to his congregation to understand the basic doctrine in example (1) is the exception rather than the rule.Likewise, example (3) illustrates the default case of reports of speech events, with a more or less "verbatim record" of the original speech event.AElfric, on the other hand, as the most typical representative of the orthodox tradition, uses imperatives overwhelmingly in reports of speech events (76 percent of all imperative occurrences, 38 tokens; contrasted to 24 percent / 12 tokens in direct speech events).
In order to relate these findings more precisely to speaker-hearer constellations and to investigate the functions that imperative and subjunctive serve in these speech representations, a closer look at the referents and grammatical subjects in these constructions seems necessary.Apart from grammatical subjects in the first, second, and third person, I included the categories "zero subject" and "vocative".In zero subject clauses, there is no grammatical subject coded in the clause, but a second-person referent is usually implied contextually or is referred to in the surrounding co-text, for example in a reporting clause.In vocative constructions, the addressee is referred to by a direct form of address, e.g. a proper name, but is not grammatically coded as subject (see leofan men in example (1).While the formal inventory of the imperative is restricted to grammatical (or logical) subjects in the second person, the subjunctive paradigm extends across all three persons (see Section 2); however, attested forms in the data imply certain restrictions here, too, at least in the sermon collections investigated.Table 4 shows the distribution of imperative and subjunctive mood according to the subjects selected by the finite verb.I discuss the situation in direct speech events first, and then comment on reports.In direct speech events, subjunctives by and large occur with first-or thirdperson grammatical subjects only; third-person subjects occur slightly more frequently than first-person subjects (1.15/1,000w contrasted with 0.84/1,000w). 6There are only two instances of second-person subjects (0.03/1,000w); zero subjects and vocative constructions are not attested at all in the data.Examples (4) and (5) show the most common usages of subjunctives in direct speech events.
5 Figures in Table 3 and In the majority of cases, subjunctives are used to issue directives.But unlike imperatives, subjunctives either address a broad and rather nondescript audience, as in example (4), or a group of people of which the speaker is one, as in (5).In (4), the preacher exhorts the congregation to observe religious fasting.This command concerns each Christian and it occurs as a general obligation (each man [is required to] hold his fast).By contrast, instances such as in example ( 5) are often interpreted as the more polite form of a directive, because the speaker includes himself in the request, which lessens the distance between him and the congregation.
Examples ( 6) and ( 7) have similar pragmatic interpretations.Here, "pure widows" are singled out for the directive to praise and honour God, and "all men who have children" are requested to teach them the right way to heaven.But again, the directive is issued rather globally.The most plausible reason for these global or generalised directives is that homilists could not ascertain the presence of the respective group of people (i.e.widows and parents) at any single preaching occasion when they composed such model homilies.So it makes sense to include directives that convey instructions to a generalised, rather than a particularised, audience in the discourse to mark a direct speech event.
Fourteen tokens of subjunctives do not encode directives of this kind.All have third-person subjects, where the subject is non-agentive and/or inanimate or where God is the grammatical subject.These may be deciphered as wishes, i.e. expressive utterances, rather than directive utterances.In the first group, those with an inanimate or non-agentive subject, the subject (e.g.lof & wuldor in example ( 8)) cannot actively engage in matching the world to the speaker's words (cf.Searle 1976).Hence, the utterance must be taken as the speaker's wish rather than his order to bring about a particular state in the world.
[To him be praise and glory always in all the world's world without end, amen.](Wulfstan, de Baptismate, HC) The second group, subjunctive clauses that have God as subject, technically speaking constitutes a group of directive utterances.God is an agentive, animate entity in the Christian belief system, so we could expect him to engage in shaping the world according to the speaker's words (cf.Searle 1976).However, as sinful, mortal beings, it seems unlikely that Christian speakers are in a position to order God to do something, i.e. to fulfil the preparatory condition for a directive.These instances may more appropriately be interpreted as approaching God with the wish that a particular state obtain instead of requiring God to bring about this state.So these utterances constitute expressives more than directives too.7 (9) God ure helpe, amen.
[[May] God help us, amen.](Wulfstan, Sermo Lupi, HC) Subjunctives in the second person are extremely rare in the data.The fact that the formal inventory has a second-person form obscures the fact that this form seems hardly to have been used in independent clauses, as Table 4 indicates.Reference grammars also seem to exclude this construction in directive speech acts, but remain vague on the issue (e.g.Quirk and Wrenn 1957: 85).Traugott (1992: 185) notes that the "hortatory subjunctive does not occur in the first person singular" but that "it occurs in all other persons".Yet, her examples are all constructions with the first person plural.By contrast, Mitchell (1985: 378) says that "the evidence ... for 'hortative' use of the second person present subjunctive is not strong.But it is difficult to dismiss it completely and to deny that it was an OE idiom".This judgement is based on the observation of cases where the subjunctive disambiguates an otherwise indistinct form.Subjunctive constructions quite clearly have the potential to issue a directive to a direct, second-person referent, as the two tokens in Table 4 indicate.But a closer look at these two instances is quite revealing, and confirms Mitchell's observations.The first instance seems to prefer the subjunctive beon over imperative beoþ, because beoþ is morphologically indistinct from the indicative.Had beoþ been used, it would be unclear whether the character traits described are to be desisted (i.e. a directive reading) or whether the author claims that they are actually absent in the audience (i.e. a negative assertive reading: "you are not").
(10) Leofan men, [...]  If this example had regular imperative beoþ, an assertive reading as "ne ge ahwar ne beoþ" ("you are none of these things") would be likely, particularly since the subject precedes the verb in this clause, a pattern that is not prototypically associated with the imperative in either negative or positive clauses (see Mitchell 1985: 374-376).So in this case, the subjunctive disambiguates a potentially ambiguous utterance, all other things being equal.
The second instance occurs in a construction with an impersonal verb and the second-person referent is not grammatically coded as subject; impersonal tweogan ('to doubt') triggers an accusative þe and a nominal complement clause.So both instances constitute or remedy grammatical oddities, and it seems difficult, indeed, to see in them a proper alternative idiom to the imperative.Summing up, subjunctives in direct speech events commonly mark directive utterances, but instances of wishes (expressive acts, as it were) also occur.The most common construction is with a third-person subject.This marks a generalised directive to an unidentified audience or a general moral obligation.Homiletic collections were defined as a discourse type where such directives are in acute demand.First-person subjunctives can be considered more polite directives, since they lessen the distance between preacher and congregation, and mark the directive as a general morally desirable act too.Against this background, secondperson subjunctives appear as grammatical exceptions.
As for imperatives in direct speech events, Table 4 shows that they usually occur without a subject (freq.0.47/1,000w) or go along with a direct form of address (a "vocative" construction, freq.0.45/1,000w) as, for example, men þa leofestan in example (12), which forms a common opening phrase in sermons.
[Hear now, dear men, what the evangelist Lucas said about this present time, and also about the time to come, and how the Lord would come to the place where he would suffer.](Blickling Homilies, HC) Imperative clauses with a proper second-person subject are scarce (0.12/1,000w), even though non-reflexive as well as reflexive constructions are included in this category (cf.Traugott 1992: 184). 8All seven tokens occur in reflexive constructions and so lack a proper second-person grammatical subject, as in examples ( 13) and ( 14).Thus, this category is more fittingly described as zero-subject imperatives with a reflexive second-person pronoun referent.Nonreflexive constructions with a second-person grammatical subject are rare (five tokens, all in reports of speech events, see Table 4).Even though there are too few instances to be firm about this, explicit secondperson subject constructions thus seem to be a marked option of the imperative in direct speech events (cf.Suter 1955: 23 on this point). 9 The prevalent function of the imperative in direct speech events is a request to the audience to perform a mental act, i.e. to consider or to understand something, as was illustrated in example (12) (geherað).Half of all cases in representations of direct speech are based on verbs such as geþencan, gemunan, understandan and gehyran, in the sense of give ear to/listen attentively.So imperatives most commonly require the addressees to pay attention to or to consider a particular point of doctrine, even if other verbs are used, as in example (15).Summing up the findings so far, imperatives in representations of direct speech commonly mark direct orders to the congregation to think over a person imperatives" are claimed to exist in Present Day English (e.g. in Quirk et al. 1985: 829, Huddleston and Pullum 2002: 925 or Biber et al. 1999: 220).9 Only one token of a non-reflexive construction with a second-person subject occurs in a negative clause, which may explain the presence of an overt subject here (cf.Mitchell 1985: 376).
particular issue; less frequently, they require the audience to perform a physical act such as to rejoice in the love of God or to show mercy.By contrast, subjunctives normally function as markers of general moral obligations and as downtoners of direct requests, in which the author includes himself in the directive.In addition, subjunctives mark wishes, but this may be interpreted as a secondary function only.
Turning now to reports of speech events, the most noteworthy issue in Table 4 is the (almost) complete lack of subjunctive constructions in representations of reports.There are two tokens of subjunctives only, and on closer inspection, both mark wishes rather than directives: This otherwise unreserved reliance on imperatives in representations of reports of speech is highly suspicious and deserves detailed comment.
First of all, we may note that the distribution of subjects in imperatives is very similar to what was already observed for direct speech events.Usually, second-person subjects are lacking and the construction relies on vocatives or fails to acknowledge addressees altogether.Different from direct speech events, however, addressees can usually be identified from the context, as in example ( 18), where the referent is given in the reporting clause he cwaeþ to iohannem (see also example (3)).
[Then he said to John: Look now, here is your mother [standing].](AElfric, Assumption of the Virgin, DOEC AECHom I, 30) Five out of nine tokens in reports of speech events have an explicit, that is, non-reflexive, second-person subject pronoun.Again, this is different from direct speech events, where all second-person forms were used reflexively.Example (19) illustrates these cases, but note that here, too, the pronoun may be interpreted reflexively, i.e. "betake yourself", even though it is clearly nominative.The OED notes for hosanna that it is "an exclamation, meaning 'Save now!' or 'Save, pray!', occurring in Psalms 18:25" ("hosanna, int., n., and v.", OED online) and classifies the form as an indeclinable interjection.So the "translation" as hael us in an imperative construction is not only remarkable but quite an unexpected source for an imperative, also considering that it is usually given as osanna in other texts (the OED cites the West Saxon Gospels (c1000)).Taken together, these examples show that the peculiar accumulation of imperatives in reports of speech events cannot be discarded as mere translation practices.To better understand the linguistic situation in reported speech events, we would need to investigate other such speakerhearer constellations, that is, written data in which an "author" reports a speech event to the recipient of the text.One promising source may be administrative documents in which one party reports to the other the words of a third.A brief consultation of the Fonthill Letter (charters, DOEC Ch 1445 (HarmD 18)), whose sole purpose may be summed up as "x reporting to y the words of z", revealed no such relevant construction at all.All representations of speech were given as oratio obliqua.Taking recourse to the narrative plane, where fictional protagonists may provide the relevant speaker-hearer constellations, I also consulted Apollonius of Tyre, which yielded one report of a speech event that did not use an imperative, given in example ( 22 But this one swallow does not make a summer, at best it indicates that much more research is needed.For one thing, this one instance contrasts to many reports of speech in Apollonius which employ the imperative, as in the first half quoted here (nimaþ þas þing).So Apollonius also follows the pattern observed in sermons to some extent.Also, as a narrative text, speaker-hearer constellations do not quite match the ones investigated here for homiletic discourse, and other (pragmatic) factors may need to be taken into consideration.
To sum up, in the present data, reports of speech events rely exclusively on imperative constructions.In outer form, they are identical to imperative constructions in direct speech events, that is, they show the same patterns of subject selection.Concerning functions, imperatives encode all sorts of actions (to go, do, teach, baptize etc.) and show no focus on mental acts, which was observed to be the case in direct speech events.While translation issues are relevant for an explanation of these findings, Latin influence and Biblical discourse practices (or the specific selection of passages included in sermons) are not the ultimate answer to this clearly biased representation of speech, as a brief discussion of other data has shown.
The overall emerging picture, then, is that direct speech events and reports of speech events contrast in their use of both moods.Direct speech events, where the text indicates "face-to-face" interaction between preacher and congregation, are marked by both imperative and subjunctive constructions, with a preference for subjunctives.Subjunctives, as far as they are used to issue directives, encode general obligations to nondescript audiences, and this function is suited to the needs of detached (or "detachable") homiletic discourse.As such, directives of this type may even go beyond the confines of the momentary delivery of the sermon.Consequently, they seem quite suitable for texts whose audience and circulation patterns were open and, to some extent, unpredictable. 11mperative constructions in direct speech events always and exclusively address the totality of listeners directly and are often loaded with the fundamentals of doctrine.Most commonly they encode mental acts, asking the audience to understand a particular point of doctrine.
Reports of speech events overwhelmingly rely on imperative constructions.These serve to report the directive verbatim to the congregation and the original interactants are usually contextualised.This is most commonly done in the reporting clause.The directive is reperformed, rather than rephrased, in the report, which underlines the vivid and lively style of many of the homilies.

Conclusions
I have shown in the analysis that imperative and subjunctive provide a pragmatic contrast in Old English homiletic discourse, and that this contrast may be linked to representations of speech."Speech" in Old English sermon texts has proven to be very sensitive to different speakerhearer constellations.In constellations where direct interaction between preachers and their flock takes place, or is represented in the text as taking place, homilists consistently opt for a greater variety of forms than when they report the speech events of others.They consider direct and straightforward, as well as more global means of exhortation; these may even exceed the immediate preaching contexts.In representations of reports of speech events, by contrast, priorities are different.Here, the formal inventory is restricted to plain imperatives.Reported speakers and reported hearers are named, but the directives issued invariantly occur in the form of imperatives.This distribution is remarkably consistent in the present study, but requires some qualification.First, the dataset proved too small to investigate the obvious bias in representations of reported speech.A cursory look at other texts showed that speech representation in Old English is little understood so far.An analysis of other material that represents various layers of "speech" seems desirable, therefore; yet, many of the extant Old English texts do not contrast speaker-hearer constellations in the same way sermons do.Often, we lack the direct representation that is given in preacher-congregation constellations, or we lack reported representations, for example in narrative texts such as Apollonius of Tyre or the Lives of Saints, where all "speech" is mediated via the text and no direct interaction between composer and audience can be assumed.So even though we have automatic tools such as the YCOE to analyse mood distribution as such, we still lack a pragmatic framework to apply to much of the Old English material when it comes to speech representation.
Another point concerns other formal manifestations of directives.It is yet unclear how periphrastic constructions with nelle/nellaþ and uton, or even the emerging class of modal verbs, complement the picture of mood distribution outlined in the present study.Constructions with uton, in particular, may be seen as a prominent periphrastic alternative to firstperson subjunctives.In the data, they commonly occur as a summarising exhortation to the congregation at the end of the sermon.But it is far from clear whether they are part of the (seemingly larger) inventory of "speech" on the level of direct speech events only, or whether their use extends to reported speech events too.
The seminal research on explicit performatives and other directives in Old English by Kohnen (2000Kohnen ( , 2008) ) offers some clues as to a larger inventory of directives, of which verbal mood is one part.But from the perspective of the present paper, it is a very unfortunate decision to dismiss reports of speech events as "secondary items" and to claim that these "do not have the same significance as primary items [i.e.directives in direct speech events]" (Kohnen 2000: 311) based on the argument that they do not genuinely contribute to discourse practices in Old English.
So much remains to be done and it seems appropriate to emphasize again that this analysis is only an initial step in the investigation of mood distribution in Old English.It remains an open question whether the distributional patterns described here hold true, whether they are

Table 1 .
Imperative and subjunctive forms in weak and strong verbs (present tense indicative given for comparison)*

Table 2 .
The corpus used for analysis 3) Þa mid ðam þa gemetton hi crist.& he hi gegrette.& hi feollon to his fotum.Ða cwaeð se haelend him to.Farað.& cyþað minum gebroðrum.þaet hi cumon togeanes me on ðam lande galilea.þaer hi geseoð me.[Then they met Christ and he greeted them and they fell down to his feet.Then said the Saviour to them: Go and tell my brothers that they come towards me in the land of Galilee where they [shall] see me.] (AElfric, Homily for Easter Sunday, DOEC AECHom I, 15) Christ's encounter with Mary Magdalene after his resurrection and his request to tell his disciples to meet him is given in a verbatim report within the homily: ða cwaeð se haelend [...] farað & cyþað.The speech event is introduced with a reporting clause (ða cwaeð) and is then rendered, as if literally, to Mary Magdalene (farað & cyþað).Such representations of speech in texts are referred to as reports of speech events.

Table 3 .
Mood distribution in direct speech events and reports of speech events (freq.per 1,000w, raw figures and percentages) However, individual sermon collections show highly distinct distribution patterns.For example, in the Blickling Homilies, 95 percent of all imperative clauses occur in reports of speech events(19 tokens).This contrasts most clearly with Wulfstan's homilies, where only 32 percent (8 tokens) of the imperatives are embedded in reports of speech, while the majority address the audience directly (roughly 70 percent; 17 tokens).Considering Wulfstan's lively and vivid style, this comes as no surprise.

Table 4 .
Distribution of imperative and subjunctive mood according to person of selected grammatical subject (per 1,000w [abs.freq. in square brackets], zero=no overt subject, voc=vocative, i.e. a form of address in co-text) 5 Table 4 add up to 299, excluding Wulfstan's Be Christendome and 11 tokens which were morphologically ambiguous (see Section 2).6 Only two first-person subjects were excluded from the analysis because they resulted in levelled forms (see Section 2).The preferred usage of the subjunctive with the third person is thus authentic and not due to my methodological practice.(4) And aegwylc mann gehealde claenlice his faesten þas þry dagas oð ða nontide [...] [And [may] each man hold his fast carefully these three days until noon [...]] (anon.sermon, DOEC HomS 35) (5) [...] gecyrran we to drihtenes willan and geþencan, þaet he deað þrowude for us and he us of ðam nearwan þystrum alysde.[[...] [let us] turn to God's will and [let us] consider that he suffered death for us and that he saved us from the approaching darkness.](anon.Homily for fourth/sixth Sunday after Epiphany, HC)

ne ge ahwar ne beon, þaes
ðe ge betan magan, gewitan ne gewyrhtan aeniges morðres oððon manslihtas, stala ne strudunga, ac strynað mid rihte.[Dear men, [...] do not in any wise be what you may improve [i.e. to better], do not witness or undertake anyone's destruction or manslaughter, theft nor robbery, but strive rightfully.](Wulfstan, de Baptismate, HC) Vulgate: euntes ergo docete omnes gentes baptizantes eos in nomine Patris et Filii et Spiritus Sancti] (Mt 28:19) (AElfric's Homilies, HC)Here, the participle baptizantes in the Latin text is coordinated with the other two imperatives in the Old English translation to form a third imperative fulliaþ (not fulwande).In other cases, particularly with prohibitions, Latin nolite + infinitive-constructions are rendered as plain imperatives in the Old English text (e.g.ne forhtiaþ eow for nolite timere in the Vercelli Homilies), despite a periphrastic option with nellaþ.Yet other cases are even more intriguing.In the following example, hael us is the translation offered for "osanna in excelsis" from the Vulgate (Mark 11, 10).While the passage as such is a rather close rendering of the Latin text, the report of the speech event itself is only very loosely paraphrased.(21)Seo menigo þe þaer beforan ferde, & seo se þaer aefter fylgde, ealle hie cegdon, & cwaedon, Haelend, Dauides Sunu, þu eart gebletsad on Drihtnes naman, hael us on heanessum.[Thecrowd which went before and those who followed after, all of them cried out and said, Saviour, Son of David, you are blessed in the name of the Lord, heal us in the highest.] ):(22) He ... heom cwaeþ to: nimaþ þas þing mid eow þe me seo cwen forgeaf and gan we secan ure gesthus.[He...said to them: take these things with you which the woman gave to me and go we seek out our lodging.](Apollonius of Tyre, DOEC, ApT)