For whom the bell tolls, or: Why we predicted the death of the mandative subjunctive

Tanja Rütten
University of Cologne


In this paper, I suggest that the life, death and revival of the mandative subjunctive that have been “predicted” in recent research appear in a different light under re-consideration of three points: 1) the range of permissible alternatives for inflected subjunctive forms, 2) the inventory of triggering expressions, and 3) the evolution of subjunctives in independent clauses. I show that methodology, co-text and conceptualisations of the linguistic category “subjunctive” have direct and sweeping consequences on the respective mappings of its change, its growth rate and our ability to correctly predict direction, speed and completion of the change. With hindsight, I thus hope to sharpen our senses for identifying unaccounted, and even unaccountable, factors in predictions of ongoing and future variation and change.

1. Introduction

Our predictive potential seemed rather poorly developed when we foresaw the death of the English subjunctive in the early and mid-twentieth century (e.g. in Harsh 1968). Since then, numerous studies have concerned themselves with a description of the astonishing revival, especially of the mandative subjunctive. Studies include synchronic, diachronic as well as diatopic perspectives. [1] While various change phenomena are known to resist the statistical exponential S-curve growth rate for various reasons (see Nevalainen, this volume), the question why the mandative subjunctive eventually resists the seemingly flawless S-curve pattern it had originally produced has hardly been voiced, let alone answered convincingly. In other words, the history of the mandative subjunctive shows how difficult it is to extrapolate from what (we think) we know about linguistic change so as to predict future developments.

In this paper, I suggest that we have neglected some vital clues for predictions about the life, death and revival of the mandative subjunctive: firstly, the range of alternatives allowed as variants. Commonly, only modal verb constructions and indicative forms are considered proper variants of the subjunctive – but does that do justice to the synchronic variation of the construction and the linguistic practices of speakers? For example, in Early Modern English mandative constructions of the type “I desire (you) that p”, subjunctive forms in the nominal complement (e.g. I desire that he come) occur alongside modal verb constructions (I desire that he should come) and indicative forms (I desire that he comes). Additionally, to-infinitives (I desire you to come), gerunds and nominalisations (I desire your coming) and even non-verbal alternatives (I desire your presence) are found. I will show that a broader selection of alternative variants will lead to different predictions about the evolution of the subjunctive.

Secondly, I will consider the inventory of triggering expressions: here, research often proceeds from preconceived eclectic lists of suasive verbs that trigger the subjunctive and alternative forms (e.g. to desire, to command, to suggest, to insist). Often, these lists of verbs are compiled from previous studies or taken from modern grammars and handbooks and vary in size quite considerably. For example, Moessner's (2010) study on the decline of the subjunctive is based on 36 triggering expressions that occur in the Middle English part of the Helsinki Corpus (2010: 153). Johansson & Norheim (1988: 29), predicting a potential increase of the mandative subjunctive in British English, analysed 22 triggers that include verbs and adjectives (e.g. “It is important that p”). Leech et al. (2009), predicting a revival led by (spoken) American English, focus on 17 verbs, adjectives and nouns as triggers (2009: 53). By contrast, Kastronic & Poplack (2014: 78), investigating North American English speech, collect 240 triggering expressions.

This variability and the inclusion of various parts of speech in individual analyses have greater consequences for our perception and prediction of the development of the subjunctive than is commonly acknowledged in the literature. Taking Visser's (1966: 827–843) historical collection of verbs that trigger mandative subjunctives as the basis, I will argue that diachronic studies which are based on triggering expressions may result in somewhat deceptive observations.

My final point concerns the exclusive focus on dependent constructions: next to nothing is known empirically about distribution patterns and functions of the so-called hortative subjunctive in non-dependent constructions (i.e. wishes or commands of the type “retyre we to our Chamber” from Macbeth II.ii, cited in Rissanen 1999: 279, as a hortative subjunctive or “first person exhortation”). I will discuss how such uses of the subjunctive complement our understanding of its historical development and in how far our predictions need to take such linguistic environments into consideration as well.

Taken together, these three points provide valuable clues that we may have been led astray in our predictions of both the demise and the return of the mandative subjunctive. I will argue that the growth, exchange rate and the eventual completion viz. reversal of the change appear radically different under only slightly different methodological and conceptual prerequisites.

In the next section, I briefly illustrate the evolution of the mandative subjunctive since late Middle English as it occurs in the relevant studies. In section three, I consider the range of variant forms, the inventory of triggering expressions and the hortative subjunctive in independent clauses as three vital clues for a reassessment of the change at hand. Section four summarises this discussion. I will argue that it is indeed exceedingly difficult to predict for whom the bell tolls, i.e. to predict the death of a linguistic variant, because linguistic co-text has the amazing ability to change change-phenomena.

2. For whom the bell tolls: The case of the mandative subjunctive

The mandative subjunctive occurs in Present-Day English only in the third person singular present tense, where it is marked by the absence of the suffix -s, i.e. where the base form of the verb is used (Quirk et al. 1985: 155), for example in the utterance I desire that he come. With the exception of the verb to be, where the subjunctive is invariant be in the present tense (i.e. distinct from am, is and are, respectively), there is no marking in any other person to indicate non-indicative. This formal inventory had largely been established by the Early Modern English period.

Utterances such as I desire that you/they come are often considered subjunctive or at least ambiguous between indicative and subjunctive (e.g. in Övergaard 1995). However, they have no morphological basis upon which we might consider them so. The verb in the complement is clearly in the base form (they come). The “quasi-subjunctive” or rather mandative reading of such sentences may result from an extension of the semantic scope of the matrix verb desire to the complement that you/they come. Leech et al. (2009: 70) note that such constructions are less exhortative than those with a modal verb or a proper subjunctive. Notwithstanding this pragmatic mandatory reading of such utterances, I consider such instances grammatically unmarked, i.e. in the base form, and treat them as indicative, not subjunctive or ambiguous forms in the present study.

The development of the subjunctive, particularly the mandative subjunctive, has received much scholarly attention from a synchronic, diachronic, and diatopic perspective. Research can be divided up into two groups. The first group of studies investigates the more remote history of the construction, the second group its more recent development. Research of the more remote history, i.e. Middle English as well as the Early and Late Modern English periods, quite clearly shows that the inflected subjunctive is losing ground against a periphrastic rival with should. For example, utterances such as I desire that he come are increasingly replaced by utterances such as I desire that he should come, where the modal verb lexically marks the mandatory force in the complement.

Auer (2006: 40) notes that the demise of the inflected subjunctive has been felt and lamented by grammarians as early as the eighteenth century. Fillbrandt (2006), Moessner (2007, 2010), Auer (2006, 2008), and Grund & Walker (2006) all sketch the gradual loss of the subjunctive in various dependent contexts to the advantage of modal verb periphrases, particularly should. This demise seems robust enough in quite a range of material. In a much-quoted statement, Henry Fowler (1965: 595) pronounced the subjunctive “moribund” at the beginning of the twentieth century, and by mid-century it was thought “non-existent” in certain genres where it had been used previously, e.g. in drama (cf. Harsh 1968: 99).

It is noteworthy that amidst this unanimous diachronic assessment, Los (2007) makes the observation that the decline of the subjunctive may be related to more fundamental changes in verb complementation patterns after verbs of persuading, urging, commanding and permitting (cf. Los 2007: 41). In a study on the rise of the to-infinitive, she observes that finite complements in a that-clause are increasingly replaced by non-finite clauses with the to-infinitive, and that the mandative subjunctive slowly lost this “competition” (Los 2007: 43); hence, utterances such as I desire that he come are replaced by I desire you to come. However, her findings have hardly been mirrored in the choice of variant forms for the mandative subjunctive yet. While Leech et al. (2009: 61) acknowledge that “alternative patterns that avoid the choice between subjunctive, should-periphrasis and indicative altogether (such as non-finite clauses) would result in much lower frequencies of the subjunctive”, they do not consider any such variant in their quantitative analysis of the mandative subjunctive.  

The second group of research papers investigates the more recent history of mandative subjunctives, i.e. the twentieth century and “contemporary English” (e.g. Leech et al. 2009). Against the background of the steady demise just outlined, recent developments predict a revival of the mandative subjunctive, in particular in spoken, informal discourse (cf. Leech et al. 2009: 60) and argue for an “active retirement” of the construction (Peters 1998: 101). Again, these findings are based on a direct comparison of the inflected subjunctive with periphrastic should and argue that the change, or reversal, is consistently found in the major varieties of English (see, among others, Johansson & Norheim 1988 and Övergaard 1995 for American and British English, Hundt 1998 for Australian and New Zealand English, Peters 1998 for Australian English, Schneider 2005 for outer circle Englishes and Leech et al. 2009 for a contemporary comparison of American and British English usage).

Only very recently has this view been questioned (e.g. in Rütten 2014 and Kastronic & Poplack 2014). Kastronic & Poplack (2014) quite rightly ask whether should and the indicative make proper variants in the first place (in a strictly technical variationist sense of variant), that is, whether there is no change in referential meaning in either of these variants compared to the mandative subjunctive. Referring to Labov's Principle of Accountability (Labov 1972), i.e. to account for all non-occurrences of a variant in addition to all actual occurrences, they also claim that, ideally, all instances where the subjunctive could potentially occur should be counted. However, they note that this is an unattainable aim, since it is impossible to know what these contexts are if the subjunctive does not occur there at least once (Kastronic & Poplack 2014: 72). Against this background, they observe that the subjunctive was “sparse and sporadic as far back as Early Modern English, and has remained that way ever since” (Kastronic & Poplack 2014: 72).

Why is it that we find in the mandative subjunctive a seeming reversal of the change and such contradictory predictions? I will now provide a more detailed discussion of factors that possibly constrain our predictive potential in this case.

3. Re-assessing variant forms, triggering expressions and conceptualisations

In this section, I will reassess the history of the mandative subjunctive by considering methodological preconditions, the influence of co-text and our conceptualisations of the subjunctive.

3.1 Methodological preconditions: Variants of the mandative subjunctive

In a recent study (Rütten 2014), I investigated the evolution of the mandative subjunctive against a larger set of variants in Early Modern English religious discourse, one of the strongholds of mandative constructions (see Moessner 2007: 224). The study was conducted with the Early Modern English sampler of the Corpus of English Religious Prose (COERP). [2] One of the unsurprising results of that study was that mandative subjunctives are steadily losing ground against other variants occurring in a similar function and co-text in religious discourse, too. What is surprising, though, is that all other variants decreased likewise in number, and that it is not the should-variant but infinitival forms that are the greatest rival of the mandative subjunctive. From these observations, we may predict that there are fewer linguistic co-texts for the mandative subjunctive in the first place, and that it is not the subjunctive which dies out, but the relevant linguistic co-text which is lacking increasingly. I will return to this point presently.

Apart from mandative subjunctives as illustrated in example (1), I included the common alternatives with a modal verb (example 2) and indicative forms (example 3) in that study.

(1) MANDATITVE SUBJUNCTIVE: Fourthly on the other side, he chargeth vs, that wee bee contented with that portion of goods which the Lorde giueth vs: (Dering, catechism, 1597, COERP)
(2) MODAL VERB: Now mine exhortation vnto you all, is, that you would be most diligent and faithfull, as in generall to take heed vnto the whole flocke ouer which the holy Ghost hath made you overseers... (Hill, preface, 1616, COERP)
(3) INDICATIVE: Qu. What is the meaning of this commaundement? An. The Lord God straightly chargeth vs in this first commaundement, that we worship God alone, which worship standeth in foure poynts. (Dering, catechism, 1597, COERP) [3]

In addition to these, I investigated several non-finite and also non-verbal alternatives, following upon Los' (2007) observation that non-finite complements have been used increasingly since Middle English (see examples 4 to 6). It is important to note that infinitival constructions (example 4), gerunds/nominalisations (example 5) and prepositional/noun phrases (example 6) do not occur in dependent structures, as subjunctives, modal verbs and indicative forms do. They are not, strictly speaking, an identical linguistic option to finite verb variants. Yet, they constitute regular and frequent alternatives for contemporary authors and it seems reasonable to discuss them alongside the subjunctive. Note that most examples (1–4; 6), despite the range of variants and sentence structures, use the same triggering expressions charge and exhort.

(4) INFINITIVAL CONSTRUCTIONS: Christe charged Peter thre times to kepe well, & norishe his shepe. The pope chargeth moche more to kepe well his moneye: and as for the shepe, he shereth and punissheth with infinite exactions. (anon., controversial treatise, 1534, COERP)
(5) GERUND/NOMINALISATION: I am forbidden all false and evil speaking, lying and slandering, railing and reviling, rash censuring and condemning others: And to this end I am forbidden all tale-bearing, and much medling and talking of other mens lives... (Sherlock, catechism, 1656, COERP)
(6) PREPOSITIONAL PHRASE: Then he generally exhorteth all to faythe & to vnitie in the same, and to compassion fraternal, loue, ge~tylnes, mekenes... (Matthew, sermon, 1535, COERP)
NOUN PHRASE: First, God forbiddeth all adulterie and vncleannesse in our bodies. (Paget, catechism, 1579, COERP)

This selection of alternatives is certainly justified by the linguistic behaviour of contemporary speakers, who used any of the following utterances after a suasive matrix verb:

Matrix verb (e.g. The pope chargeth us / he exhorteth / I am forbidden...)  


+ dependent clause with finite VPsubj

~ that we bee contended
(see example 1)

+ dependent clause with finite VPmodal aux

~ that you would be most diligent and faithfull (see example 2)

+ dependent clause with finite VPind

~ that we worship God alone
(see example 3)


+ infinitival complement

~ to kepe well & norishe his shepe
(see example 4)


+ gerund/nominalisation

~ all false and evil speaking
(see example 5)


+ NP / PP

~ to faythe & to vnitie
~ all adulterie and vncleannesse
(see example 6)

Table 1. Variant forms in mandative constructions.

Table 2 shows the distribution of these variants in the sixteenth and seventeenth centuries in five religious genres. Quite consistently, the mandative subjunctive, along with should-periphrasis, indicatives and even the non-finite and non-verbal variants (to- and bare infinitives, gerunds and prepositional/noun phrases) declines from the early sixteenth to the late seventeenth century in constructions with a (matrix) suasive verb.

  Finite variants Non-finite / non-verbal variants
Period subjunctive modal verb indicative infinitive gerund PP/NP*
1500–1549 1.3 4.2 0.8 10.2 0 3.1
1550–1599 1.7 4.0 0.7 10.7 0 2.6
1600–1649 0.4 3.4 0.9 8.4 0.6 2.1
1650–1700 0.6 3.1 0.7 9.8 0.3 1.9

Table 2. Mandative subjunctive and alternative constructions in five religious genres in Early Modern English, freq. per 10,000w; adapted from Rütten (2014).
* Prepositional and noun phrases are taken together as non-verbal variants.

Within two hundred years, figures for subjunctives are cut in half; likewise, modal verb periphrases and infinitival constructions decline, though not as drastically (from 4.2 to 3.1 in modal verbs and from 10.2 to 9.8 in infinitival complements). Indicative and gerund constructions as well as noun phrases/prepositional phrases are rather insignificant alternatives, but they also decrease. [4] This indicates that the decline of the subjunctive that has been observed in numerous previous studies must be linked to more fundamental changes. These may either be found in the context, i.e. a less directive discourse style in religious instruction, greater consideration of the congregation's face wants etc. or in the co-text, namely a preference for non-finite complementation patterns, as suggested by Los (2007).

Against the background of the religious upheavals brought about by the Reformation, changes in discourse practices, a less directive diction, as it were, cannot be denied. This may result in the replacement of straightforward exhortation such as in example (7) with more indirect, face-saving utterances as in (8), at least for some time: [5]

(7) INDICATIVE: I charge you in the feare of God that you do not mistake that which is said (Smith, sermon, 1591, HC; cited in Kohnen 2007: 144)
(8) NOUN PHRASE: This I say on purpose to recommend to men a nobler exercise for their wits (Tillotson, sermon, 1671, HC; cited in Kohnen 2007: 144)

However, Table 2 also shows that infinitives were by far the preferred complementation pattern in all periods (10.2 in the early sixteenth century and still 9.8 in the late seventeenth) and that even after 1650, their numbers remain much higher than those of modal verb constructions (3.1/10,000w). This fact is completely ignored when one considers only finite variants. While the diachronic loss of the subjunctive attested in previous studies is confirmed here, this study testifies to a more fundamental change concerning verb complementation. What, then, are the consequences of this result for the expected “revival”?

There are two possible answers: either the loss that precedes the revival or the revival itself constitute a “phantom change”. Analysing the loss of the mandative subjunctive as a phantom change means identifying the change as originating elsewhere. It is not the morphological subjunctive that succumbs to periphrasis with a modal verb, but it is the co-text that is wanting in the first place. If, increasingly, non-finite complements are selected by the matrix verb, there is simply no syntactic environment for the subjunctive (or any other finite variant) to appear. The loss of the subjunctive is real enough, but it is only a secondary repercussion, caused by changes on a higher grammatical level, not by an inherent preference for periphrastic constructions, as previous research suggests. The change is less morphological but more morpho-syntactic (and maybe even pragmatic). This, of course, emphasises the need to inquire into other syntactic environments of the subjunctive, and I will do so presently.

For our predictions about the mandative subjunctive, this re-interpretation of the development in Early Modern English means that the revival comes even more unexpectedly and in a more pronounced manner. Also, it must be interpreted as even more far-reaching in the language system, since the revival of the category involves a return to finite complements, according to the account given above.

On the other hand, the revival itself could be the “phantom change”, because in correlating the two variants (subjunctive and should) it is curiously overlooked that should itself undergoes change. Leech et al. (2009: 87) discuss ongoing change in the central modal verb system and come to the conclusion that uses of “putative, quasi-subjunctive should” (i.e. of the type It is right that women should control their desires; cited from Leech et al. 2009: 86) have declined since the mid-twentieth century, which is exactly the time frame for which the revival of the mandative subjunctive was postulated. Counting two and two together, the revival of the subjunctive may strike one as a “home-grown” change. It may not be the subjunctive that returns but modal should that declines – a correlative change that has not been added in the equation so far. While I cannot provide a numerical analysis of this issue here, the argument is relevant in so far as it shows very clearly how strongly change-phenomena depend on the variants considered and how sensitive they are to co-text. Possibly, we did not allow for correlative changes in our predictions.  

3.2 Triggering expressions

Another issue relating to the influence of co-text is the set of triggering expressions, at first sight a methodological problem: depending on the number of triggers we consider, there may be fewer or more tokens of the subjunctive in individual studies. The range of triggers allowed in studies is impressive enough; it ranges from 17 (in Leech et al. 2009) to 40 (in Moessner 2007) and an unprecedented 240 in Kastronic & Poplack (2014). One reason for this discrepancy is the sheer lack of a definite and comprehensive set of triggers in the English language. I will discuss this issue in relation to Visser's historical list of mandative matrix verbs from his Historical English Syntax (1966) and show that this is a problem which simply cannot be tackled and is thus a largely unforeseeable factor in predictions about the mandative subjunctive.

Visser (1966: 827–843) lists 174 verbs that are attested to have taken a mandative subjunctive as complement on at least one occasion in the history of English (nota bene: methodologically, this is contrary to Labov's Principle of Accountability again, since this list offers no verbs where the subjunctive is possible in principle, but only those where the subjunctive is actually attested). [6] Table 3 provides a quantitative presentation of these verbs according to the periods in the language at which they were attested in the relevant construction.

  Allocated to period New items in period Total attested in period Examples
Old English 108 - - - 108 seon, will/would, wish
Middle English 27 37 64 require, pray, suppose
Early Modern English 32 15 47 sentence
Present-Day English 11 21 32 insist, propose, recommend

Table 3. Triggering expressions for mandative subjunctives in the history of English, based on Visser (1966: 827–843).

This overview shows that Old English has by far the largest number of triggers, and the further we proceed in time, the fewer triggers there are, from a total of 108 attested in Old English to 32 in Present-Day English. In addition, triggers are highly period-specific. Most of the items given in Table 3 spring up and die out within one and the same period. For example, out of the 108 Old English triggers, only 27 are still used in Middle English. While another 37 new items stock up the inventory to a total of 64 Middle English triggers, 81 Old English triggers died out during the transition to Middle English.

The same is true for the transition from Middle to Early Modern English, even though figures are not as drastic here. Half of the Middle English triggers are still in use in Early Modern English (32 out of 64), and another 15 newly occur, resulting in an inventory of 47.

For Present-Day English, figures decrease even more to merely 32, of which only 11 were retained from Early Modern English. Most of the triggers that are considered typical of mandative constructions in Present-Day English are from 1800 or later according to Visser's list. Examples include demand and resolve (1800), propose (1806), vote (1820), beg (1861), urge (1905), suggest (1917), move (1929), insist (1931) and recommend from as late as 1948. Only eight triggers from Visser's list are attested throughout all periods, among them see, will, would and wish. Example 9 illustrates one of these (see), which forms the sturdy exception in the changing diachronic inventory of triggering expressions:

(9) MANDATIVE SUBJUNCTIVE: ... and I pray you, sweet Husband, bring up this Child in good letters, in Learning and Discipline, and above all things, see that he be brought up and instructed in the exercise of true Religion. (Stubbes, biography, 1591, COERP)

The period-specific distribution of triggers puts considerable constraints on any study that takes triggering expressions as its starting point. For example, eight out of the seventeen triggers that Leech et al. (2009) investigate are from 1800 and later in Visser's list. Since it seems that nearly half of the triggers they examine are relevant for PDE only, this restricts their explanatory power considerably.

The issue of triggering expressions does thus not only pose a methodological problem, viz. the discrepancy of the number of triggers in individual studies, which Kastronic & Poplack (2014: 78) rightly call a “methodological infelicity”. More severely, it puts heavy constraints on any diachronic claims we wish to make about mandative subjunctives. In principle, we would be comparing two (or more) individual linguistic systems, whose inventory of triggers was highly period-specific. Thus, any long-term diachronic study is necessarily skewed, unless we restrict the analysis to the eight triggers that have been extant from Old English onwards. [7]

One way of dealing with this problem is to start from the actual texts and work “bottom-up” instead of running an eclectic list of triggers through the dataset. The relevant triggers must be extracted manually from the dataset. However, not even a bottom-up approach guarantees an exhaustive inventory of triggers. Quite contrarily, it would most likely result in a very restricted set of triggers, namely that relevant for the discourse type(s) under investigation. The inventory of triggering expressions is unstable at best and certainly no reliable factor in predictions about constructions involving the mandative subjunctive.

3.3 Conceptualisations of the (mandative) subjunctive

My third point concerns the development of the subjunctive in independent clauses. All too often, the subjunctive is discussed as a form which is exclusive to grammatically dependent contexts. In Present-Day English, the subjunctive mood is said to be restricted to “set expressions” and to mark a “formal and rather old-fashioned style” in independent clauses (Quirk et al. 1985: 157–158; Biber et al. 1999: 918; Huddleston & Pullum 2002: 925). This situation seems relatively unchanged since at least the Early Modern period (cf. Rissanen 1999: 228), but it is noteworthy that empirical studies on this point are lacking – quite contrary to the bustling activity evolving around mandative and adverbial subjunctives. The function of the subjunctive in independent clauses is described as hortative, so quite like that of mandative subjunctives in dependent clauses, for example in utterances of the type A curse vpon him, die he like a theefe (Shakespeare, Pericles,; cited in Rissanen 1999: 279).

In order to relate our conceptions of mandative subjunctives in dependent clauses to hortative subjunctives in independent clauses, I conducted a pilot study of c. 50,000w in a corpus of Old English sermons compiled from the Old English section of the Helsinki Corpus and the Dictionary of Old English Corpus (DOEC). Old English has a mood system where the subjunctive is still marked in all persons in the present tense and so allows a rather precise description of the functions of the subjunctive mood in independent clauses. I use the religious domain as a testing ground again, because it is assumed that the subjunctive is particularly frequent in the genres of the religious sphere (see Moessner 2010: 166). The results are given in Table 4.

Relevant forms occur with an overall frequency of 2.1/1,000w and are attested in all but one of the text files investigated. It turns out that out of the total occurrences of 119 subjunctive forms, 104 (87.4%) were used to mark directive speech acts. In only 15 cases (12.6%), subjunctives mark wishes, e.g. God forgeafe þæt... [may] God forgive that...; (Ælfric, ÆCHom I, 30, DOEC). This overall distribution is, by and large, reflected in the individual text files. Examples 10 to 12 illustrate both usages:

(10) SUBJ_wish: Him sy lof and wuldor.
[To] him be praise and glory. (Anonymous, Homily, HomS30, DOEC)
(11) SUBJ_hortative: Biddan we georne urne Drihten þæt...
[let us] earnestly ask our Lord that...; (anon., Blickling Homilies, HC)
(12) SUBJ_hortative: Gehyre se mann þe ðis smeað andsware his smeagunge.
[Let] the man who thinks this hear the answer to his thought. (Ælfric, Homiliy for Easter Sunday; CH I, 15, DOEC)

Example (10) illustrates part of the standard doxology terminating a sermon and is found in similar formulaic usage in Present-Day English, too, e.g. in God bless you or God save the Queen. In example (11), the preacher includes himself in the directive, which in Present-Day English is normally done with a periphrastic construction involving let. In example (12), on the other hand, people who doubt the account of Christ's resurrection are admonished to hear the arguments in favour of it. The formula used is rather vague (any man who doubts this point of doctrine) and is given in the subjunctive with a third-person referent. Thus, it provides for the possibility of an objection, but does not appear as a direct command to the congregation at large. It may rather be described as a generalising or abstract(ing) directive function.

Text files

Subjunctive forms




Total (55,180w)

87.4 [104]

12.6 [15]

100 [119]

(files: COWULF3, COWULF4 HC; Hom6 DOEC; size: 12,580w)

85.7 [30]

14.3 [5]


(files: COAELHOM, HC; ÆCHom I, 15, 30, 40, ÆHom 12, DOEC; size: 13,733w)

84.2 [16]

15.8 [3]


Blickling Homilies
(file: COBLICK HC; size: 10,670w)

95.5 [21]

4.5 [1]


Homily 6th Sunday Epiphany
(file COEPIHOM, HC, size: 1,610w)

100 [7]



Vercelli Homilies
(files: HomS 1 HomU 8 HomU 9 HomS 3, DOEC; size: 8,988w)

88 [22]

12 [3]


(file: HomS 48, DOEC; size: 1,492w)




(file HomS 35, DOEC; size: 2,284w)

100 [5]



(file: HomS 30, DOEC; size: 1,810w)

50 [1]

50 [1]


(file: HomS 45, DOEC; size: 2013w)

50 [2]



Table 4. Subjunctives in independent clauses in OE sermons according to function in the text (percentages and [number of tokens]; overall frequency: 2.1/1,000w)

As can be seen from Table 4, the hortative function is the default function of the subjunctive in the sermon material investigated. It occurs either in the form of first-person exhortations, in which the preacher includes himself (as in example 11), or it expresses the type of vague, generalising exhortation illustrated in example (12). Wishes do not figure prominently at all in constructions with the subjunctive.

In light of Present-Day English usage, these findings are striking. Judging by the descriptions in reference grammars, hortative subjunctives seem to have all but disappeared. The subjunctive is consistently described as marking wishes or as a petrified construction, e.g. God save the Queen or be that as it may (cf. Quirk et al. 1985: 157–158; Biber et al. 1999: 918). Hortative subjunctives are hardly ever mentioned and their history remains somewhat obscure so far.

However, from the results presented in Table 4 and the descriptions in modern grammars, we might, with some confidence, predict the death of the hortative subjunctive in independent clauses, too, and claim a replacement by periphrastic let in constructions with a first-person referent. Both developments (decline and replacement) mirror the evolution of mandative subjunctives in dependent clauses, and there is a certain appeal in this analogy.

But what has become of generalising directives with a third-person referent, i.e. utterances such as the following?

(13) SUBJ_hortative: A curse vpon him, die he like a theefe.
(Shakespeare, Pericles,; cited in Rissanen 1999: 279).

Rissanen claims that such cases may be interpreted as either hortative subjunctives or alternatively as imperative constructions with a third-person referent (1999: 279), and notes that the issue is largely terminological. But this explanation for the situation in Early Modern English is unsatisfactory for two reasons. Firstly, it indicates an overlap of grammatical categories that deserves more commentary. Secondly, Old English did not have a morphological form for third-person imperatives, neither are such forms attested in Middle English. So when, where and why did subjunctive and imperative merge?

The data presented in Table 4 may shed some light on this issue. Table 4 shows a preponderance of hortative subjunctives, particularly with third-person referents, and we may claim that these are the actual ancestors of “third-person imperatives” (comp. gehyre se man in example (12) with die he in example (13)). Under this perspective, all third-person imperatives are, in fact, historical subjunctives. Thus, the hortative subjunctive appears much more robust and seems to persist quite productively until the present day in such forms as Someone open the door, a generalised directive in the form of a subjunctive, quite like in Old English. In turn, descriptions of subjunctives as fossilised forms relevant only for set expressions such as God save the Queen appear as incorrect predictions once again from this perspective.

This, admittedly, brief discussion shows how our preconceived ideas of the grammatical category subjunctive and its behaviour in other syntactic environments, i.e. in dependent clauses, may affect our perception of change phenomena. We will have to investigate in more detail in how far the historical development of the subjunctive and the imperative overlap and when (and for what reasons) the reanalysis of hortative subjunctives as third-person imperatives occurred. For the present purpose, the discussion has shown that linguistic predictability is affected by relying too closely on analogues and by gaps in the empirical documentation.

4. Conclusion

The aim of this paper was to shed light on factors that influence predictions of language change. I have chosen the subjunctive as an example, because its development has received much scholarly attention and is yet elusive to a considerable extent. So it presents a good test case for factors that influence linguistic predictions. Equipped with a rather traditional understanding of grammatical categories and a pragmatic, usage-based approach to data evaluation, I have shown that factors which affect predictions of linguistic change fall into two categories: those that are foreseeable and those which require a crystal ball.

To a considerable extent, alternatives for mandative subjunctives are predictable. Yet, a wider or narrower range of alternatives will affect the outcomes of the study, and the extent of the influence of the number and type of variants seems largely unforeseeable, and unpredictable. For the mandative subjunctive, I have included non-finite and non-verbal forms in addition to the finite set of variants that is commonly considered. This showed that it is not the morphological form that is disfavoured per se, but that the change may involve more fundamental changes in verb complementation patterns. This result is surprising in light of the received wisdom of the mandative subjunctive, but it seems quite sound – and it is unpredictable from the mere choice of alternatives. We do not know beforehand and cannot predict which set of alternatives will result in a faithful trajectory of the change.

Along these lines, the fact that individual alternatives for any one linguistic form will change with time, too, is predictable, if not self-evident. But just how the change of one variant will affect the behaviour of the other(s) that are investigated is unforeseeable once again. I have argued that the assumed rivalry of should-periphrasis and inflected subjunctive, which intuitively is another instance of the general drift towards analyticity in English, has obscured the fact that the central modal verb system underwent considerable change simultaneously, at least since the mid-twentieth century. The “revival” of the subjunctive may be caused by the demise of should, rather than by a strengthening of the subjunctive itself. Such correlative changes seem plausible enough, but are unpredictable. We do not know how exactly the life of one variant will bear upon the development of the other. So we cannot predict the future life of the subjunctive, since we do not know how permanent the “demise” of putative should is, or which other alternatives future speakers might prefer.

The same is true for predictions about the subjunctive in independent clauses, in particular for hortative subjunctives with a third-person referent. Had we considered its development in isolation or in analogy to dependent context only, we probably would have foretold its decline with residual forms in fixed formulae of the type God bless you. Again, it is related forms (imperatives) which tell a different story, and whose influence is, again, much less predictable and in this case completely unexplored.

Yet other factors are entirely unforeseeable. The most relevant for the evolution of the subjunctive is the lack of a comprehensive set of triggering expressions. This is a crucial point, since most studies of the mandative subjunctive rely on just such a set. While it seems obvious that divergent numbers of triggers in individual studies make it difficult to compare the results of these studies, the diachronic effects are even more severe. I have shown that the number of triggers not only decreased since Old English, but was highly period-specific, too. Consequently, there are not only fewer co-texts for the mandative subjunctive, but they are highly distinct in each historical period, too. While this result must be taken with caution, since these observations rely exclusively on Visser's eclectic list of verbs, it is indicative of a rather fundamental problem: we simply cannot say which verbs will trigger mandative subjunctives (or finite complements in the first place) in the future. Since the inventory of triggers seems highly flexible, we might predict that it is going to change, and that the triggers which are considered typical today (e.g. to command, to insist) may not be used in the same way in the future. But we have no way of identifying a comprehensive, stable set.

At the end of the day, then, the subjunctive may not be “so unimportant as is sometimes suggested” (Quirk et al. 1985: 155). At the same time, the sheer number and range of factors encumbering precise predictions show that, in the words of seventeenth-century cleric and poet John Donne, we may never be “sen[t] to know for whom the bell tolls” (John Donne, Devotions vpon emergent occasions) when it comes to predictions of the future life of linguistic forms.


[1] For diachronic approaches in the more recent history of the subjunctive, see Fillbrandt (2006), Auer (2006, 2008), Grund & Walker (2006) and Kastronic & Poplack (2014). For diatopic studies, sometimes combined with a diachronic perspective, see Johansson & Norheim (1988), Övergaard (1995), Hundt (1998), Peters (1998), Schneider (2005), Crawford (2009) and Leech et al. (2009).

[2] For the corpus see COERP in the Sources section below; also Rütten et al. (2008) and Kohnen (2010).

[3] Note that I do not consider such forms ambiguous, since they are morphologically unmarked, regular indicative verbal forms; semantically, however, the mandative scope of the matrix verb (chargeth in the present example) may well extend to the complement. See section 2 for an overview of the formal inventory of the subjunctive.

[4] Gerunds are only attested in the relevant construction from the seventeenth century onwards and figures are really too low to make any sensible claims about their role as alternatives.

[5]  On this point see Kohnen (2007), who claims that “only during the Early Modern period a decline of the frequency [of directive speech acts] could be linked to a tendency towards more polite manifestations” (2007: 158).

[6] The question whether these verbs actually “trigger” the subjunctive, as is commonly tacitly assumed, or whether the list exists simply because Visser selected only verbs that are attested with the subjunctive in the first place is a chicken-and-egg story that seems impossible to answer. Yet, the list provides a starting point for many diachronic investigations of mandative subjunctives and is thus usually taken to represent a “set of triggers”, as it were (e.g in Moessner 2010 or Fillbrandt 2006).

[7] However, it must be said that the overview in Table 3 only includes verbs and is thus somewhat restricted in its informative value. Nevertheless, it may serve as a point of reference for the potential instability of the triggering expressions throughout language history.


COERP = Corpus of English Religious Prose. 2008. Rütten, Tanja, Ingvilt Marcoe, Kirsten Gather & Thomas Kohnen.

DOEC = Dictionary of Old English Corpus; original release (1981) compiled by Angus Cameron, Ashley Crandell Amos, Sharon Butler, and Antonette diPaolo Healey (Toronto: DOE Project 1981); 2009 release compiled by Antonette diPaolo Healey, Joan Holland, Ian McDougall, and David McDougall, with TEI-P5 conformant-version by Xin Xiang (Toronto: DOE Project 2009).

HC = The Helsinki Corpus of English Texts. 1991. Compiled by Matti Rissanen (Project leader), Merja Kytö (Project secretary); Leena Kahlas-Tarkka, Matti Kilpiö (Old English); Saara Nevanlinna, Irma Taavitsainen (Middle English); Terttu Nevalainen, Helena Raumolin-Brunberg (Early Modern English). Department of Modern Languages, University of Helsinki.


Auer, Anita. 2006. “Precept and practice: The influence of prescriptivism on the English subjunctive”. Syntax, Style and Grammatical Norms: English from 1500-2000, ed. by Christiane Dalton-Puffer, Dieter Kastovsky, Nikolaus Ritt & Herbert Schendl, 33–53, Frankfurt/Bern: Peter Lang.

Auer, Anita. 2008. “Lest the situation deteriorates: A study of lest as a trigger of the inflectional subjunctive”. Standards and Norms in the English Language, ed. by Miram Locher & Jürg Strässler, 149–173. Berlin: de Gruyter.

Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan. 1999. The Longman Grammar of Spoken and Written English. Harlow: Pearson.

Crawford, William J. 2009. “The mandative subjunctive”. One Language, two Grammars? Differences between British and American English, ed. by Günter Rohdenburg & Julia Schlüter, 257–276. Cambridge: Cambridge University Press.

Donne, John. 1624. Devotions vpon Emergent Occasions. London.

Fillbrandt, Eva-Liisa. 2006. “The development of the mandative subjunctive in the Early Modern English period”. Trames 10 (60/55): 135–151.

Fowler, Henry. 1926 (1965). A Dictionary of Modern English Usage. 2nd, rev. edn. Oxford: Clarendon.

Grund, Peter & Terry Walker. 2006. “The subjunctive in adverbial clauses in Nineteenth-Century English”. Nineteenth-Century English: Stability and Change, ed. by Merja Kytö, Mats Rydén & Erik Smitterberg, 89–109. Cambridge: Cambridge University Press.

Harsh, Wayne. 1968. The Subjunctive in English. Tuscaloosa: University of Alabama Press.

Huddleston, Rodney & Geoffrey Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.

Hundt, Marianne. 1998. New Zealand English Grammar: Fact or Fiction? A Corpus-Based Study in Morphosyntactic Variation. Amsterdam: Benjamins.

Johansson, Stig & Else Helene Norheim. 1988. “The subjunctive in British and American English”. ICAME Journal 12: 27–36.

Kastronic, Laura & Shana Poplack. 2014. “The North-American English mandative subjunctive in the 21st century: Revival or remnant?” Penn Working Papers in Linguistics 20(2): 71–80.

Kohnen, Thomas. 2007. “Text types and the methodology of diachronic speech act analysis“. Methods in Historical Pragmatics, ed. by Susan M. Fitzmaurice & Irma Taavitsainen, 139–166. Berlin: de Gruyter.

Kohnen, Thomas. 2010. “Religious discourse”. Historical Pragmatics (Handbooks of Pragmatics 8), ed. by Andreas H. Jucker & Irma Taavitsainen, 523–547. Berlin: de Gruyter.

Labov, William. 1972. Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press.

Leech, Geoffrey, Marianne Hundt, Christian Mair & Nicholas Smith. 2009. Change in Contemporary English: A Grammatical Study. Cambridge: Cambridge University Press.

Los, Bettelou. 2007. “To as a connective in the history of English”. Connectives in the History of English, ed. by Ursula Lenker & Anneli Meurman-Solin, 31–60. Berlin: de Gruyter.

Moessner, Lilo. 2007. “The mandative subjunctive in Middle English”. Studies in Middle English Forms and Meanings, ed. by Gabriella Mazzon, 209–226. Frankfurt/Main: Peter Lang.

Moessner, Lilo. 2010. “Mandative constructions in Middle English”. ICAME Journal 34: 151–168.

Övergaard, Gerd. 1995. The Mandative Subjunctive in American and British English in the 20th Century. Stockholm: Almquist & Wiksell.

Peters, Pam. 1998. “The survival of the subjunctive: Evidence of its use in Australia and elsewhere”. English Worldwide 19(1): 87–103.

Quirk, Randolph, Geoffrey Leech, Jan Svartvik & Sidney Greenbaum. 1985. A Comprehensive Grammar of the English Language. London: Longman.

Rissanen, Matti. 1999. “Syntax”. The Cambridge History of the English Language, Vol. III, ed. by Roger Lass, 187–331. Cambridge: Cambridge University Press.

Rütten, Tanja. 2014. “Comparing apples and oranges: The study of diachronic change based on variant forms”. Anglistentag 2013 Konstanz: Proceedings, ed. by Silvia Mergenthal & Reingard Nischik, 373–385. Trier: Wissenschaftlicher Verlag Trier.

Rütten, Tanja, Ingvilt Marcoe, Kirsten Gather & Thomas Kohnen. 2008. The Corpus of English Religious Prose: Introduction.

Schneider, Edgar W. 2005. “The subjunctive in Philippine English”. Linguistics and Language Education in the Philippines and Beyond. A Festschrift in Honour of Ma. Lourdes S Bautista, ed. by Danilo T. Dayag & Stephen Quackenbush, 27–40. Manila: Linguistic Society of the Philippines.

Visser, F. Th. 1966. An Historical Syntax of the English Language. Part Two: Syntactical Units with one Verb. Leiden: E. J. Brill.