The subjunctive conundrum in English 1

The subjunctive conundrum in English 1 Bas Aarts Department of English Language and Literature, UCL There have been various approaches in the literature to the questions of whether the English subjunctive is a separate mood (conceived of as an inflectional form), and whether subjunctive verbs and/or clauses should be regarded as finite or non-finite. In this article I discuss these issues, and I will argue that the properties of the English subjunctive can be accounted for by recognising a subjunctive clause type, characterised by a number of distinctive syntactic properties. Keywords: subjunctive, mood, inflection, clause type 1. Introduction The view espoused in Palmer (1987: 46) that the notion of a subjunctive mood is a simple transfer from Latin and has no place in English grammar is generally accepted in most modern descriptive frameworks. But the consequences of accepting such a view have not been sufficiently appreciated in the literature. In this article I will discuss a number of approaches to the English subjunctive, and I will argue that none of them deals adequately with the fallout of denying the existence of an inflectional subjunctive in English. I will propose that English subjunctive clauses can be described by making reference to the notion of Subsective Gradience (as proposed in Aarts 2007), and that the grammar of English should recognise a subjunctive clause type, along with declaratives, interrogatives, imperatives and exclamatives. 1 This article is based on a plenary lecture I delivered at the First Triennial Conference of the International Society for the Linguistics of English (ISLE) in Freiburg, Germany. I am grateful to two anonymous referees, Ekkehard König, Bettelou Los, Ingo Plag, Geoffrey Pullum, and Tim Waller for comments. I am also grateful to Teresa Fanego for valuable suggestions. Folia Linguistica 46/1 (2012), 1 20. issn 0165 4004, e-issn 1614 7308 Mouton de Gruyter Societas Linguistica Europaea doi 10.1515/flin.2012.1

2 Bas Aarts 2. Some background I will start with a brief reminder of the history of the subjunctive in English. In Old English (OE) there were two forms for the singular and plural subjunctive, as in the paradigm for the class I weak verb trymman, shown in Table 1. Table 1. The paradigm for OE trymman (from Hogg 2002: 42 3) Indicative Subjunctive Present Past Present Past 1 Sg trymme trymede Sg trymme trymede 2 Sg trymest trymedest Plural trymmen trymedon 3 Sg trymeð trymede Plural trymmað trymedon Note that in OE there was a present and past subjunctive. Overall the number of inflections for the subjunctive was limited, even at that time, with only singular and plural forms, but no person distinctions. The situation was slightly more complicated for the verb bēon ( to be ) whose present tense paradigm had alternative forms based on a b-root and s-root, as in Table 2, and different forms for the past tense, based on a w-root. As noted by Elizabeth Traugott (1992: 240), the subjunctive had a wider range of uses in OE than in present-day English. It was used, for example, in reported speech. As is well-known, the subjunctive inflection was later lost, and modal verbs or the indicative took over (Moessner 2002, 2006, 2007, Auer 2006). Most modern grammarians agree that there is no justification for recognising an inflectional subjunctive in English (Quirk et al. 1985: 97, Palmer Table 2. The paradigms for the verb be in Old English (from Lass 2006: 58) Present Past Indicative Subjunctive Indicative Subjunctive s-root b-root s-root b-root w-root only 1 Sg eom bēo sīe bēo wæs wær-e 2 Sg eart bi-st sīe bēo w ǣ r - e w ǣ r - e 3 Sg is bi-þ sīe bēo wæs w ǣ r - e Plural sindon/sint/ bēo-þ sīe-n bēo-n w ǣ - r o n wǣr-en ear-on

The subjunctive conundrum in English 3 1987: 46, Haegeman and Guéron 1999: 324 6, Huddleston and Pullum et al. 2002: 89 90, Anderson 2001, 2007: 12). The present subjunctive is held to be identical to the base form of regular verbs, while the past subjunctive is identical to the past tense, with be a special case. By recognising that there is no inflectional subjunctive mood in English, syncretism in the verbal paradigms of English is avoided, which is a welcome result. However, there is much less agreement as to whether subjunctive verbs are finite or nonfinite. And there is also no agreement as to whether the containing clause is finite or non-finite. For PDE there have been claims by many scholars, that the subjunctive is dying out. Famously, Fowler listed the facts shown below regarding the subjunctive in his Dictionary of Modern English Usage. (These are repeated in the second edition; 1965: 595 8.) that it is moribund except in a few easily specified uses; that owing to the capricious influence of the much analysed classical moods upon the less studied native, it probably never would have been possible to draw up a satisfactory table of the English subjunctive uses; that assuredly no one will ever find it possible or worth while to do so now that the subjunctive is dying; and that subjunctives met with today, outside the few truly living uses, are either deliberate revivals, especially by poets, for legitimate enough archaic effect, or antiquated survivals giving pretentious flavour to their context, or new arrivals possible only in an age to which the grammar of the subjunctive is not natural but artificial. (Fowler 1965: 595) Peters (2004: 520) claimed that Fowler s views may have led to a decline of the use of the subjunctive, given Fowler s negative attitude to it. It is not clear whether his influence would have been that great, but certainly he was not the only person to hold negative views about the subjunctive. Frank Palmer was also uncompromising when he wrote that the notion of a subjunctive mood is a simple transfer from Latin and has no place in English grammar (1987: 46). However, in the 1990s attitudes toward the subjunctive changed. Robert Burchfield in his 1998 revision of Fowler s book changed the entry on the subjunctive, and claimed the opposite of what Fowler claimed, namely that the use of the subjunctive is increasing in BrE and other varieties. A number of studies have since shown that there has been a revival of the subjunctive in the twentieth century. Among them are Övergaard (1995), Hundt (1998), Serpollet (2001) and Leech et al. (2009).

4 Bas Aarts As for cross-varietal usage, it is well-known that there are important differences between British and American English. In the United States the subjunctive is very much alive, and much more frequent than in the UK. According to Algeo (2006: 264) what he calls the mandative indicative, as in I think it s crucial that he checks the facts, occurs with some frequency in the UK, but is very rare in the US. He notes that this construction is confusing for Americans because it is either interpreted as a statement of fact or unacceptable. If there is indeed an issue of mutual incomprehension, this makes the subjunctive a very interesting area of research into how varieties of English might be diverging from each other. 2 I will now discuss some modern approaches to the grammatical analysis of the subjunctive. 3. A gradience approach: Quirk et al. (1985) In Quirk et al. (1985: 149) the subjunctive is regarded as a marked mood. They write [t]erms for the two major categories of the present subjunctive are the mandative and the formulaic subjunctive... These are realized, like the imperative, by the base form of the verb (ibid.: 155). The base form can be finite or nonfinite (ibid.: 97). The subjunctive verb form is regarded as finite, despite the fact that, with the exception of be, there is no inflectional tense contrast and no person/number concord. Quirk et al. (1985: 150) propose what they call a scale of finiteness, as shown in Table 3, applying the criteria shown below the table. Quirk et al. (1985: 150) remark that [t]he five criteria listed above lead to inconsistency, in that subjunctive and imperative verb phrases are according to criteria (a) and (e) finite, whereas according to criterion (c) they are non-finite. (ibid. 1985: 150). Criterion (d) is not mentioned in this quotation, though the table clearly shows that according to this criterion subjunctives are regarded as nonfinite. While the table captures the hybrid nature of subjunctive clauses extremely well, there are some modifications we could implement. In my view criterion (a) should be marked? for the subjunctive, given that locutions like God save the Queen are fossilised, and not productive. With regard 2 Note also that in Britain the construction with should is common, as in I think it s crucial that he should check the facts, whereas this is unusual in the US. See below.

The subjunctive conundrum in English 5 Table 3. Quirk et al. s scale of finiteness (1985: 150) (a) (b) (c) (d) (e) indicative + + + + + subjunctive +? + Finite imperative +? + infinitive Nonfinite (a) Finite verb phrases can occur as the verb phrase of independent clauses. (b) Finite verb phrases have tense contrast. (c) There is person concord and number concord between the subject of a clause and the finite verb phrase. (d) Finite verb phrases contain, as their first or only word, a finite verb form which may be either an operator or a simple present or past form. Do-support is used in forming (for example) negative and interrogative constructions [when no auxiliary verb is present]. (An anonymous referee notes that do-support occurs only if no auxiliary verb is present.) (e) Finite verb phrases have mood, which indicates the factual, nonfactual, or counterfactual status of the predication. to criterion (b), as is well known, the past subjunctive is syncretic with the past tense for most verbs, and cannot be distinguished from it. However, given that be is an exception, criterion (b) should perhaps also be marked?. Criterion (e), by stating that [f]inite verb phrases have mood, which indicates the factual, nonfactual, or counterfactual status of the predication confuses the syntax of subjunctive verb phrases with their semantics, and for that reason we could drop it altogether. As we will see in a moment, however, reasons do exist for regarding subjunctive clauses as finite. The scale of finiteness implies the existence of gradience as defined by Aarts (2007) in allowing for verb phrases to be finite to a greater or lesser degree. This is an attractive idea, which we also find in Palmer s work: the subjunctive, if it does not have full tense marking..., is less finite than the declarative (Palmer 1986: 162), and in the work of Givón (1990, 1993, 2001), discussed critically in Bisang (2007: 117f.). See also Koptjevskaja- Tamm (2009: 230). Despite the perceived gradience Quirk et al. opt for a sharp distinction between finite and nonfinite verb forms, as Table 3 shows: the indicative, imperative and subjunctive are finite verb forms, whereas the infinitive is nonfinite. By percolation verb phrases and clauses that contain a finite verb are also finite. Quirk et al. thus follow the tradition at least in part:

6 Bas Aarts while recognising that there is no distinctive inflectional subjunctive, the verb form that realises the subjunctive is nevertheless finite. But claiming that subjunctive verbs are finite, while at the same time maintaining that there is no inflectional subjunctive in English is problematic. One of the consequences of this stance is that there is no way to distinguish subjunctive clauses from indicative clauses. We cannot say that a subjunctive clause is a clause that contains a subjunctive verb form, because there are no subjunctive verb forms. Both subjunctive and indicative clauses in this view are finite, headed by a finite verb. 4. Two generative approaches Working in the generative tradition Radford (1988: 291) also treats subjunctive clauses as finite on two grounds, as the quotation below makes clear. quotethe claim that subjunctive Clauses are finite can be defended on both universalist and particularist grounds. On universalist grounds, we can argue that in languages which have a richer inflectional system than English, subjunctive Clauses do indeed turn out to be inflected... The second (obviously more compelling) reason for treating subjunctive Clauses in English as finite is that they share certain morphosyntactic properties in common with indicative Clauses which differentiate them from nonfinite Clauses. For example, neither indicative nor subjunctive Clauses can be subjectless, whereas nonfinite Clauses can indeed be subjectless.... Secondly, subjunctive Clauses pattern like indicative Clauses (and unlike non-finite Clauses) with respect to the case-marking of any overt Subject which the clause contains. (Radford 1988: 291; emphasis in original) I think we can safely disregard the universalist reason, which Radford himself indicates is unpersuasive. It has long been accepted that the grammar of English should not be modelled on that of another language, such as Latin, as has been done in the past, and reviving this bad practice on grounds of universality would be unfortunate. We are left with Radford s morphosyntactic reasons for analysing subjunctive clauses as finite. These are more persuasive, at least as far as the observation regarding an obligatory subject is concerned. The case-marking argument appears to be circular, at least within the GB-model that Radford espouses. Here a subject receives nominative Case from a finite inflection node ( INFL or I ). Subjunctive clauses have nominative subjects because they have a finite

The subjunctive conundrum in English 7 inflection node. Why do they have a finite inflection node? Because they have nominative subjects. I will return to the morphosyntactic criteria that Radford discusses later. An alternative view is put forward in later books by Radford (Radford 1997: 161, 2009: 107 9) and also in Haegeman and Guéron (1999: 324 6), amongst others. 3 These authors treat clauses that contain a subjunctive verb as finite, despite the subjunctive verb itself being realised as a nonfinite verb. How is this implemented? They propose that a sentence like (1) contains a covert modal auxiliary, indicated by the boldface [M] in (2). In this way the subordinate clause can be finite, while the verb be is nonfinite. To support their analysis the authors point to the analogy between (2) and (3), and between (4) and (5). (1) It is important that he be invited. (2) It is important that he [M] be invited. (3) It is important that he should be invited. (4) It is important that he [M] not receive the information. (5) It is important that he should not receive the information. Interestingly, the idea of an ellipted modal is based on ideas which Visser traces back to the eighteenth century. In the eighteenth century the notion that there was an ellipsis of auxiliaries seems to have prevailed: 1761 Joseph Priestley (The Rudiments of English Grammar p. 84), discussing the utterance We shall overtake him though he run, writes: Almost all the irregularities in the construction of any language arise from the ellipsis of some words which were originally inserted in the sentence, and made it regular: let us endeavour to explain this manner of speaking, by tracing out the original ellipsis: may we not suppose that the word run in this sentence is the radical form (which answers to the infinitive mood in other languages) requiring regularly to be preceded by another verb expressing doubt or uncertainty, and the entire sentence to be We shall overtake him though he should run?. (Visser 1966: 788) Radford mentions a number of further publications which posit a null modal should, rather than a null M (ibid. 108 9, 136). Positing a covert modal auxiliary is problematic, because again we have circularity: why is the subordinate clause finite? Because it contains a non-overt modal. Why does it contain a non-overt modal? Because we want the clause to be finite. Haegeman and Guéron (1999: 325) say that [i]f there is a non-overt modal 3 See Radford (2009: 136) for references.

8 Bas Aarts in the sentence, then the properties of subjunctive clauses follow. It is difficult to see how the properties of subjunctive clauses can follow from something that is not overtly present. 5. A constructional approach to the subjunctive I turn now to the analysis adopted in Huddleston and Pullum et al. (2002). Before discussing their approach to the subjunctive a few comments on their treatment of the present tense are necessary. These authors claim, in common with other grammarians, that it makes sense to say that English has a present tense, even for the verb forms that are realised by the base form of the verb (which Huddleston and Pullum et al. 2002 call the plain present tense ). This is because there exists a third person singular inflection for regular verbs, and because the paradigm for be shows a fuller set of forms in the present tense. Thus, in they work hard we can call work a plain present tense inflectional form. However, Huddleston and Pullum et al. (2002) argue that this line of reasoning cannot apply to the subjunctive: because English has no unique inflectional subjunctive forms at all, it does not make sense to say that in I demand that he go the verb go is a present subjunctive inflectional form. It follows that it also does not make sense to call it a finite base form, as Quirk et al. (1985) do. Instead, for Huddleston and Pullum et al. (2002: 83) so-called subjunctive verbs are not marked for being finite or nonfinite in form. They occur in what they call the plain form. Note that for these authors the plain form is distinct from the plain present tense form. The former can occur in finite and non-finite clauses, as we will see presently, whereas the latter always occur in finite clauses. In Huddleston and Pullum et al. (2002) the notion of finiteness is defined as follows. The general term finite is related to its everyday sense of limited. More specifically, a finite verb is characteristically limited with respect to person and number. In its traditional application to English, for example, takes is finite in that it is limited to occurrence with a 3rd person singular subject. Being limited is thus a matter of being marked for the relevant categories. (Huddleston and Pullum et al. 2002: 89) In English for many grammarians being limited with respect to person and number entails being tensed (Trask 1993: 103 4, Leech 2006: 41). By

The subjunctive conundrum in English 9 percolation, a verb phrase or clause is finite by virtue of its head carrying tense. Tensed verbs, with the exception of the modals, can head an independent predication. For Huddleston and Pullum verb forms are primary or secondary. They write: Clauses whose verb is a primary form are finite, those whose verb is a past participle or gerund-participle [= present participle] are nonfinite, but those with a plain form verb can be either, depending on the construction (Huddleston and Pullum et al. 2002: 88). Clauses containing a plain present tense form are finite and contain a finite plain present tense form, as we have seen. Subjunctive clauses are examples of finite clauses headed by a verb in the plain form. All this means that for Huddleston and Pullum a clause that is marked as finite is not necessarily tensed. What s more, a clause that is tensed is not necessarily finite. An example of a tensed nonfinite clause is given in the italicised portion of (6). Huddleston and Pullum regard the perfect construction as a secondary tense, and for this reason the italicised clause is tensed, but nonfinite. (6) He was pleased to have finished his exam. For Huddleston and Pullum, finiteness is a syntactic category of the clause (ibid.: 88 9), not an inflectional category. 4 As we have seen, they deny the existence of subjunctive verb forms in English (2002: 89 90). Instead, for them the subjunctive is a construction (ibid.: 77, 89, 993f.): The term subjunctive is generally applied to an inflectional category of the verb but,..., we are here reinterpreting it as the name of a syntactic construction a clause that is finite but tenseless, containing the plain form of the verb (Huddleston and Pullum et al. 2002: 993, italics mine). Viewed in this way the italicised portion of (7) is an example of a non-tensed finite subjunctive clause with a verb in the plain form. (7) It is essential that he possess the quality of the unsung hero: a hero who is, if anything, embarrassed if the singing starts. [The Times, 22 May 2007] As we have just seen, for Huddleston and Pullum et al. primary forms are inflected, i.e. they are defined as being limited with respect to person and number ( the primary forms are the tensed ones ; ibid.: 88). But they also say that [t]he prototypical finite clause contains a tensed verb (or irrealis were), but there are grounds for extending the boundaries of the category to include the imperative and subjunctive constructions too. (Huddleston 4 On the notion of finiteness, see also Plag (1993: 102f.) and the articles in Nikolaeva (2007).

10 Bas Aarts and Pullum et al.: 89). If imperative and subjunctive clauses are less prototypically finite because they do not contain a finite verb this can only mean that they are limited with respect to person and number to a lesser degree. And this is indeed the case according to these authors because subjunctive verbs lack a subjunctive inflection. But if they are not inflected, this entails that defining the notion finite subjunctive construction becomes an issue. Huddleston and Pullum do not offer an explicit solution to this problem, though they do offer an implicit solution by appealing to a notion of what we might call syntactic resemblance. They say that subjunctive clauses are finite because they are like finite clauses in three ways (Huddleston and Pullum et al. 2002: 90): they have obligatory subjects; they take the same subordinators as tensed clauses, and they alternate with tensed clauses In sum, for Huddleston and Pullum there are no subjunctive verb forms, only finite subjunctive constructions, headed by verbs which are neither finite nor nonfinite. The main problem with the Huddleston and Pullum account is that questions arise with respect to the exact nature of the subjunctive construction that these authors invoke. They leave the concept undefined, because they do not make clear how they understand the concept of construction. As a theoretical notion, constructions do not play a defined role in their grammar. 6. Anderson (2001 and 2007) In two recent articles from 2001 and 2007 John Anderson defines the notion of finiteness as the capacity to license an independent predication (Anderson 2007: 1; see also Anderson 2001: 160). He calls this syntactic finiteness, as distinct from morphosyntactic finiteness (2001: 160). For Anderson the capacity to license an independent predication seems to be a necessary and sufficient condition for finiteness. He concludes that for this reason the English subjunctive is clearly syntactically nonfinite (2001: 161), because subjunctive clauses cannot stand on their own. In Anderson s account, the verb form left in (8) is also nonfinite, which seems surprising initially, but follows from his reasoning; after all, the clause in

The subjunctive conundrum in English 11 which this non-temporal use of the verb leave occurs cannot stand on its own. In Anderson s words: The preterite form occurs under rection by the would of the following clause (2001: 162). (8) If I left on Tuesday, you would regret it. Using the same reasoning, the verb should in (9) is nonfinite for Anderson. And in (10) smokes is also regarded as nonfinite, because the relative clause cannot stand on its own. (9) They demand(ed) that I/Babsie should leave. (10) This is the pipe which Bill smokes in bed. As for a sentence like (11), here the that-clause is finite, despite the fact that it cannot stand on its own, and this is because, the subordinator is, in Anderson s words, extra-predicational, so not pertinent to questions of finiteness (2007: 9). (11) We said (that) Bill smokes a pipe in bed. In (10) the relative pronoun which is clearly not extra-predicational : it functions as the direct object of the verb smoke. Anderson says that the English subjunctive shows morphological reduction; if we recognize a continuum here, the subjunctive is relatively less morphosyntactically finite than the indicative, but more so than the infinitive, in so far as it distinguishes... preterite vs. nonpreterite (though not necessarily as a reflection of temporal distinctions) and it licenses unmarked subjects. (ibid.) This is reminiscent of the model in Quirk et al. (1985) shown in Table 3 and of Huddleston and Pullum et al. s (2002) account. Despite mentioning a continuum, it does not seem to play a role in Anderson s analysis. Characterising finiteness only with reference to the capacity to license an independent predication is perhaps too restricted a definition of the notion of finiteness. Summarising the analyses we have looked at so far, Quirk et al. opt to analyse subjunctive clauses as finite, headed by a finite verb. Radford and Haegeman and Guéron regard the subjunctive verb as nonfinite, and appeal to a non-overt modal auxiliary verb to be able to analyse the subjunctive clause as finite. This is similar to Huddleston and Pullum et al. s analysis, except that for them the containing clause is a subjunctive construction, which is regarded as finite by appealing to an unexpressed notion of syntactic resemblance. For Anderson subjunctive verbs and their clausal

12 Bas Aarts projections are nonfinite. All these approaches are in some way or other unsatisfactory, as we have seen. I will turn now to my own approach to English subjunctives. 7. Subjunctive as clause type I would like to suggest a novel approach to subjunctives in which the notion of a subjunctive construction is accepted, and analysed as peripherally finite in terms of the morphosyntactic features that it displays. More specifically, I propose an approach to the subjunctive in terms of what I have elsewhere called Subsective Gradience (SG; Aarts 2007). This obtains when elements that belong to a particular grammatical form class or construction display the properties of that class or construction to a greater or lesser extent. With regard to the subjunctive construction we can now say that on the scale of finiteness subjunctive clauses are peripherally finite if we regard the finite prototype as being characterised by the properties listed below Table 4, which is a recast and extended version of Quirk et al. s Table 3. Using these criteria we posit that prototypical finite clauses conform to all the criteria, whereas subjunctive clauses fully conform to only three criteria, and partially to two. We have a case here of Subsective Constructional Gradience (Aarts 2007). At this point the reader may be wondering why subjunctive clauses instantiate Subsective Gradience (SG), rather than Intersective Gradience (IG). This is because IG obtains when a formative or construction can be characterised by properties from two sets of morphosyntactic properties that pertain to those elements or constructions. An example of Constructional Intersective Gradience is the coordination-subordination gradient. In the case under discussion we have only one set of properties that apply, which makes this an example of SG. In this connection, I would like to propose here that the grammar of English recognise a subjunctive clause type, along with declaratives, interrogatives, imperatives and exclamatives. I conceive of clause types as a system of grammar, described in detail in Huddleston (1984, 1994), and consolidated in Huddleston and Pullum et al. (2002). In this framework the clause types are the analytic reflexes of the synthetic moods of English. Thus declarative or indicative mood is implemented in a declarative clause type, and imperative mood is implemented in an imperative clause type.

The subjunctive conundrum in English 13 Table 4. Subsective Gradience in subjunctive clauses (a) (b) (c) (d) (e) (f) (g) ± ± + + + (a) Finite clauses can license an independent predication. > The subjunctive mostly cannot, but formulaic subjunctives are an exception. (b) Finite clauses have tense contrast. > There are no present/past subjunctive verb forms for most verbs. The verb be (which Huddleston and Pullum 2002 call irrealis were) is an exception. (c) There is person concord and number concord between the subject of a clause and the finite verb phrase. > This does not apply to the subjunctive. (d) Finite clauses contain a verb form which may be either an operator or a simple present or past form. Where no auxiliary verb is present do-support is used in forming (for example) negative and interrogative constructions. > This does not apply to the subjunctive. (e) Finite clauses have an obligatory subject, in the nominative case where appropriate. > This applies to the subjunctive. (f) Finite clauses make use of particular subordinators, typically that. > This too applies to the subjunctive. (g) Finite clauses can alternate with subjunctive clauses after appropriate triggers, e.g. verbs such as demand, insist, require, adjectives such as desirable, necessary, imperative, etc. > This applies to the subjunctive. a, b a An anonymous referee asks a number of questions about the proposed properties, for example How do we know all relevant features have been counted?, How do we know the same weight can be attributed to all features? and Why are functional/semantic criteria not considered? These are issues which I address in Aarts (2007: 225 8). It is important to bear in mind that what I propose in Aarts (2007) is an idealised model of linguistic reality. This is a reasonable strategy in an attempt to get to grips with the often rather messy facts. b It might be objected that the observation that subjunctive clauses alternate with tensed clauses is perhaps not a fully convincing reason for saying that they are finite. We might observe, for example, that the fact that Jim to be a fool alternates with that Jim is a fool after e.g. believe does not make the former any more finite. However, the alternation between finite clauses and subjunctive clauses after certain triggers is a systematic one, whereas the alternation between finite clauses and infinitive clauses is not. I m grateful to an anonymous referee for pointing this out to me. (The clause type system also allows for interrogative and exclamative clauses, though there were of course never interrogative or exclamative moods in English.) Each clause type can be characterised syntactically, and has an associated predominant use. Thus, declaratives are typically used to make statements which the speaker believes to be true, interrogatives are typically used to ask questions, etc. As is well known, indirect speech acts

14 Bas Aarts of the can you pass the salt-type are very common. In this particular case an interrogative structure is used to issue a directive. In order to talk of a subjunctive clause type, a case has to be made for the subjunctive clause as a structure with a distinctive morphosyntactic make-up. This can be done by enumerating the properties of subjunctive clauses, as follows. Subjunctive clauses... (i) do not take do-support when negated; instead the verb is preceded by not; (ii) are subordinate (with a few formulaic exceptions), i.e. cannot occur on their own; (iii) have verbs which occur in the plain form (thus the third person singular present tense is not marked by -s); (iv) do not show a tense contrast (except for be). The first and second points on this list distinguish the subjunctive clause type from the declarative clause type. The first point is illustrated by (12), as compared with the declarative clause in (13). (12) I suggest that he not go. (*I suggest that he don t go.) > subjunctive (13) I suggest that he does not go. > indicative We see that in negated subjunctive clauses the negative marker not is placed before the verb. Visser (1966: 847) speculates that this is a recent syntactic pattern. By contrast, declarative clauses feature do-support. The second point distinguishes the subjunctive clause type from the declarative, interrogative and exclamative clause types which can either occur as main clauses or as subordinate clauses. It also distinguishes the subjunctive clause type from the imperative, which can only occur in main clauses. If we recognise a subjunctive clause type we obtain a neat parallel with the declarative, interrogative, imperative and exclamative clause types, each of which also have a distinctive syntactic makeup. In English we have no interrogative, imperative and exclamative moods, and arguably no indicative mood, conceived of as verb inflections: instead we have declarative, interrogative, imperative and exclamative clause types. In the same way, there is no subjunctive mood, again conceived of as a verb form, but it is nevertheless possible to recognise a subjunctive clause type, as we have seen. The resulting clause type system of English would look like Table 5.

The subjunctive conundrum in English 15 Table 5. The clause type system in English Mood Indicative Clause type Declarative Interrogative Imperative Exclamative Subjunctive Typical use Make statement Ask a question Issue a directive Utter an Issue a directive (mandative which the exclamation speaker believes clauses) to be true If we regard mood in English as being essentially analytic, as suggested above, then we can eliminate the top row of this table (cf. Huddleston and Pullum et al. 2002: 172, who make no use of the label indicative ). See also Palmer (1986: 23f.). The label use in this table refers to the typical illocutionary force associated with the clause type immediately above it. From the point of view of usage, typical subjunctive clauses are like imperative clauses in having a directive force. Thus in example in (14) below (a mandative subjunctive) the speakers wished something to be brought about, namely that certain items are supplied to the union. (14) For both safety and security purposes I require [that a list of all locks and keyholders, together with a full set of keys be supplied to the union]. 5 The present proposal bears similarities to an analysis put forward in Huddleston (1984: 360), where he proposes a jussive 6 clause type, which comprises two subtypes shown in Table 6: imperative and non-imperative. Within the non-imperative type we have the Heaven forbid-type in main clauses, and the type exemplified by the subordinate clause in It s essential that they be present. Huddleston reserves the label subjunctive for verbal inflections in his 1984 book. A problem with Huddleston s (1984) proposal is that the jussive clause type seems to be too commodious, and appears justified mainly, if not exclusively, on the basis of the semantic resemblance between imperative and subjunctive constructions. The model in Table 6 is not adopted in Huddleston and Pullum (2002). In that publication the Heaven forbid-type of the non-imperative jussive is classed as a minor clause type, more specifically an optative (Huddleston and Pullum et al. 2002: 944). For the subordinate type in It s essential that 5 British component of the International Corpus of English (ICE-GB); W1B-028 094. 6 This notion goes back to Lyons (1977: 745), for whom the jussive is a clause type associated with the imperative.

16 Bas Aarts Table 6. The jussive clause type (from Huddleston 1984: 360) Imperative Non-imperative let gr -Subject Be careful Main Heaven forbid +Subject You be careful Subordinate It s essential that they be present. +let gr Let s go to the beach If that is what the premier intends, let him say so. they be present the label subjunctive construction is used, as we have seen. In the proposal I presented in Table 6 the subordinate subjunctive is classed as one of the major clause types. The advantage of this analysis is that there now obtains a parallelism with the other clause types, which gives meaningful content to Huddleston and Pullum et al. s notion of a subjunctive construction. One issue remains. What do we do with what Quirk et al. call putative should? This is the type of construction we have in (9) and (15). (15) It is crucial that she should leave the premises on time. Table 7. Properties for clauses containing putative should (a) (b) (c) (d) (e) (f) (g) + + + + + (a) Finite clauses can license an independent predication. > Clauses with putative should cannot. (b) Finite clauses have tense contrast. > Should is morphologically, if not semantically, a past tense. (c) There is person concord and number concord between the subject of a clause and the finite verb phrase. > This does not apply to putative should-clauses. (d) Finite clauses contain a verb form which may be either an operator or a simple present or past form. > This applies to clauses containing putative should. (Do-support is irrelevant here.) (e) Finite clauses have an obligatory subject, in the nominative case where appropriate. > This applies to clauses containing putative should. (f) Finite clauses make use of particular subordinators, typically that. > This applies to clauses containing putative should. (g) Finite clauses can alternate with subjunctive clauses after appropriate triggers, e.g. verbs such as demand, insist, require, adjectives such as desirable, necessary, imperative, etc. > This applies to putative should-clauses.

The subjunctive conundrum in English 17 Is this should finite or non-finite? Huddleston and Pullum et al. note that that this construction involves a use of a modal auxiliary that cannot be identified with one of the uses characteristic of main clauses (ibid.) Main clause should (as in you should leave) typically carries the deontic meaning of obligation. We have already seen that for Anderson should in (9) and (15) is nonfinite by virtue of the fact that the containing that-clause cannot stand on its own, precisely because of the specialised meaning of should. For Huddleston and Pullum et al. the that-clauses in (9) and (15) are not ordinary declaratives, but what they call specialised modal constructions (2002: 994). They would presumably be finite in their account. However, as the Quirk et al.-style grid in Table 7 shows, putative should-clauses are not prototypically finite, since only five out of the seven properties are marked with a plus symbol. If that-clauses containing putative should are indeed finite, this means that we must allow our grammar to recognise finite clauses that cannot stand on their own. To conclude, I summarise the different approaches to the subjunctive discussed in this article in Table 8. On the grounds that subjunctive clauses display a particular syntactic configuration, the analysis proposed here differs from the others in suggesting that the grammar of English recognise a subjunctive clause type. Table 8. The (non)finiteness and clause type of subjunctives Finite verb Nonfinite verb Finite clause/ construction Nonfinite clause Clause type Quirk et al. (1985) + + Decl. Radford (1988) + + Decl. Radford (1997)/Haegeman and Guéron (1999) + + Decl. Huddleston and Pullum et al. (2002) n/a n/a + Decl. Anderson (2001/2007) + Decl. Aarts (2011) n/a n/a + Subj. Not in bibl (or is it this article?)

18 Bas Aarts References No reference in the text Aarts, Bas. 2007. Syntactic gradience: The nature of grammatical indeterminacy. Oxford: Oxford University Press. Algeo, John. 2006. British or American English? Cambridge: Cambridge University Press. Anderson, John M. 2001. Modals, subjunctives, and (non-)finiteness. English Language and Linguistics 5: 159 66. Anderson, John M. 2007. Finiteness, mood and morphosyntax. Journal of Linguistics 43: 1 32. Auer, Anita. 2006. Precept and practice: the influence of prescriptivism on the English subjunctive. In Christiane Dalton-Puffer, Dieter Kastovsky, Nikolaus Ritt & Herbert Schendl, eds. Syntax, style and grammatical norms: English from 1500 2000. Frankfurt: Peter Lang, 33 53. Bisang, Walter. 2007. Categories that make finiteness: Discreteness from a functional perspective and some of its repercussions. In Irina Nikolaeva, ed. Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press. 115 37. Burchfield, Robert. 1998. The new Fowler s modern English usage. Revised 3rd edn. Oxford: Oxford University Press. Fowler, H. W. 1965. A dictionary of modern English usage. Revised by E. Gowers, 2nd edn. Oxford: Clarendon Press. Givón, Talmy. 1990. Syntax: A functional-typological introduction. Amsterdam: John Benjamins. Givón, Talmy. 1993. English grammar: A function-based introduction. Amsterdam: John Benjamins. Givón, Talmy. 2001. Syntax. Amsterdam: John Benjamins. Haegeman, Liliane & Jacqueline Guéron. 1999. English grammar: A generative perspective. Oxford: Blackwell Publishers. Hogg, Richard M. 1992. Phonology and morphology. In Richard M. Hogg, ed. The Cambridge history of the English language. Vol. I: The beginnings. Cambridge: Cambridge University Press. 67 167. Hogg, Richard M. 2002 An introduction to Old English. Edinburgh: Edinburgh University Press. Huddleston, Rodney. 1984. Introduction to the grammar of English. Cambridge: Cambridge University Press. Huddleston, Rodney. 1994. The contrast between interrogatives and questions. Journal of Linguistics 30: 411 39. Huddleston, Rodney & Geofffrey Pullum et al. 2002. The Cambridge grammar of English. Cambridge: Cambridge University Press. Hundt, Marianne. 1998. It is important that this study (should) be based on the

The subjunctive conundrum in English 19 analysis of parallel corpora: On the use of mandative subjunctive in four major varieties of English. In Hans Lindquist et al. eds. The major varieties of English. Papers from MAVEN 97. Växjö: Växjö University, 159 75. Koptjevskaja-Tamm, Maria. 2009. Review of Irina Nikolaeva, ed. 2007. Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press. Folia Linguistica 43: 213 49. Lass, Roger. 2006. Phonology and morphology. In Richard Hogg & David Denison, eds. A history of the English language. Cambridge: Cambridge University Press, 43 108. Leech, Geoffrey. 2006. A glossary of English grammar. Edinburgh: Edinburgh University Press. Leech Geoffrey, Marianne Hundt, Christian Mair & Nicholas Smith. 2009. Change in Contemporary English: A grammatical study. Cambridge: Cambridge University Press. Lyons, John. 1977. Semantics, vol. 2. Cambridge: Cambridge University Press. Moessner, Lilo. 2002 Who used the subjunctive in the 17th century? In Sybil Scholz, Monika Klages, Evelyn Hantson & Ute Römer, eds. Language: Context and cognition. Papers in Honour of Wolf-Dietrich Bald s 60th Birthday. Munich: Langenscheidt-Longman, 227 35. Moessner, Lilo. 2006. The subjunctive in Early Modern English adverbial clauses. In Christian Mair & Reinhard Heuberger, eds. Corpora and the history of English. Papers dedicated to Manfred Markus on the occasion of his sixty-fifth birthday. Heidelberg: Winter, 249 63. Moessner, Lilo. 2007. The mandative subjunctive in Middle English. In Gabriella Mazzon, ed. Studies in Middle English forms and meanings. Frankfurt/Main: Peter Lang, 209 26. Nikolaeva, Irina, ed. 2007. Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press. Övergaard, Gerd. 1995. The mandative subjunctive in American and British English in the 20th century. Acta Universitatis Upsaliensis. Studia Anglistica Upsaliensia 94. Stockholm: Almqvist and Wiksell International. Palmer, Frank. 1986. Mood and modality. Cambridge: Cambridge University Press. Palmer, Frank. 1987. The English verb. 2nd edn. London: Longman. Peters, Pam. 2004. The Cambridge guide to English usage. Cambridge: Cambridge University Press. Plag, Ingo. 1993. Sentential complementation in Sranan: On the formation of an English-based creole language. Tübingen: Max Niemeyer. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A comprehensive grammar of the English language. London: Longman. Radford, Andrew. 1988. Transformational grammar: A first course. Cambridge: Cambridge University Press.

20 Bas Aarts Radford, Andrew. 1997. Syntactic theory and the structure of English: A minimalist approach. Cambridge: Cambridge University Press. Radford, Andrew. 2009. Analysing English sentences: A minimalist approach. Cambridge: Cambridge University Press. Serpollet, Noëlle. 2001. The mandative subjunctive in British English seems to be alive and kicking... Is this due to the influence of American English? In P. Rayson, A. Wilson, T. McEnery, A. Hardy & S. Khoha Proceedings of the Corpus Linguistics 2001 Conference. Lancaster University: UCREL Technical Papers 13: 531 42. Trask, R. L. 1993. A dictionary of grammatical terms in linguistics. London: Routledge. Traugott, Elizabeth Closs. 1992. Syntax. In Richard M. Hogg, ed. The Cambridge history of the English language. Vol. I: The beginnings. Cambridge: Cambridge University Press, 168 289. Visser, F. Th. 1966. An historical syntax of the English language. Leiden: Brill. Author s address Bas Aarts Department of English Language and Literature UCL Gower Street London WC1E 6BT UK b.aarts@ucl.ac.uk Received: 30 April 2010 Revision invited: 8 June 2010 Revised version received: 21 July 2011 Accepted: 2 September 2011