English subjunctive studies: topics and issues
4.3 Mandative subjunctives: issues in methodology
4.3.2 Identification of mandative clauses
Studies involving mandative clauses are prompted by the regular appearance of the otherwise uncommon present subjunctive in such clauses, which are known to be associated with particular mandative21 lexical items, or ‘triggers’. As a result, most studies start with a list of such triggers and search for the different complements that follow them.
It is important to make clear that although the lists of triggers are made on the basis of their co-occurrence with subjunctives, the items involved should be seen as mandative triggers rather than simply subjunctive triggers; subjunctive clauses are just one of the types of mandative clause that they allow.
Visser, in the introduction to his long section on subjunctives, makes a similar point:
From the fact that these modally marked [i.e. subjunctive] forms occur after expressions of volition it should not be inferred that the latter are determinative in this respect. For one thing the units with the modally marked form . . . express the intended modality independently of these introductory expressions . . . The introductory expression does not ‘require’ a modally marked form of the verb, but is only added to indicate whether the content of the clause is a desire, an advice [sic], an exhortation, a command, a suggestion, a request, etc, and to make it clear who it is that utters the desire, command, suggestion, etc. For another, there is the fact that there are numerous examples in which expressions of volition are followed by object clauses with a modally non-marked [i.e. indicative] or zero form of the verb. (Visser 1963–73: 825)
In light of this, ‘trigger’ is perhaps not the best term to use, as it implies that something follows more or less automatically; nevertheless, it remains useful as a shorthand term.
Another approach to finding mandative clauses involves reading through whole texts, which is obviously more labour-intensive than searching electronically for trigger words and is impractical when working with anything but small corpora. Nevertheless, such an approach was adopted for some of her study by Övergaard (1995), who read through all of the texts she had collected for her diachronic study of the mandative subjunctive in the twentieth century. One interesting consequence of this is that she claimed to find a number of mandative subjunctives (and other variants) not governed by obvious mandative triggers, or as she puts it, ‘the matrix need not contain an explicitly mandative verb/noun for the utterance to have mandative meaning’ (1995: 82). Among the examples she gives is (96), her (159):
21 Sometimes characterised as ‘suasive’ (e.g. Quirk et al. 1985: 1182).
(96) I exploded and told J. that her single priority in my book is that O. be up and dressed and fed a decent breakfast to ensure her healthy circulation.
<A90: H1>
Here, the subjunctive is licensed by priority, which, while not being regularly associated with subjunctives, seems to me to contain enough deontic force for a mandative interpretation not to be inappropriate, particularly in the context. Perhaps the most useful point to be drawn from Övergaard’s approach is the importance of the contribution of the context to the overall mandative meaning.22 Convincing examples of mandative subjunctives licensed by items with no trace of mandative meaning are not common. Huddleston & Pullum (2002: 1000) offer (97), apparently a real-world example:
(97) I would stress that people just be aware of the danger.
Yet, for me, be here feels more like an imperative than a true subjunctive, though much would depend on context and intonation, if, as seems likely, it is a spoken example.
When it comes to the practicalities of looking for mandative subjunctives in an electronic corpus, the trigger approach unsurprisingly remains the most popular one, not only because even in parsed corpora it is unlikely that all appropriate verb forms will have been consistently marked as subjunctives, but also because studies are concerned with other variants in the same environment, not just subjunctives (see Section 4.3.4). Practicality also tends to dictate which triggers are chosen. The list of 30 triggers, as shown in Table 4.1, in Johansson & Norheim’s (1988) ground-breaking study of the mandative
subjunctive in LOB and Brown has been relied on in numerous subsequent studies, principally to enable easy comparison. Yet there is no explanation of how these triggers were chosen in the first place, and it is not certain that they are necessarily the most appropriate triggers to use in a study aimed at providing an accurate picture of the situation in PDE.
22 The B-Brown corpus (AmE, 1931) contains an interesting example of this kind, where the trigger for the mandative subjunctive be is ‘a good thing’: ‘We think it a good thing, perhaps, that the correspondents responsible for influencing or inflaming public opinion here and abroad be given, if it be necessary, a sense of responsibility, although they would not be here if they had not already achieved that’ <B-Brown B23>. (Note also the presence of a conditional clause containing a present subjunctive.)
Table 4.1. Mandative triggers: Johansson & Norheim (1988).*
essential important necessary sufficient
Adjectives 2‡ anxious
* Based on Table 1 and the list of ‘corresponding nouns’ in Johansson & Norheim (1988: 28, 29).
‡ With Type 1 adjectives the mandative clause characteristically appears as subject or extraposed subject;
with Type 2, it is normally complement within the adjective phrase (see Huddleston & Pullum 2002: 999).
A larger list can be found in Huddleston & Pullum (2002: 999), as displayed in Table 4.2, though the authors make a point of calling it ‘a sample’ and stressing that ‘there can be no question of giving a definitive list of mandative items’ (2002: 999). Apart from the additional triggers this list contains, it is notable that three of the items in Johansson & Norheim’s list are not included – direct, wish and sufficient – presumably on the basis that there is not enough evidence of them being productive mandative items in PDE to warrant inclusion. Slightly oddly, the list does not include mandatory, even though it is one of the triggers that feature in the authors’ initial explanation of the mandative construction (2002: 995).
Table 4.2. Mandative triggers: Huddleston & Pullum (2002).*
Verbs / corresponding nouns
* Based on (14) and (15) in Huddleston & Pullum (2002: 999). Items marked by ‘†’ also readily allow non-mandative content-clause complements.
‡ With Type 1 adjectives the mandative clause characteristically appears as subject or extraposed subject;
with Type 2, it is normally complement within the adjective phrase (2002: 999).
The way in which Johansson & Norheim approached mandative nouns has also affected later studies. They listed ‘accompanying nouns’ for eight of their 17 verbs, and they simply combined the results for those nouns with the figures for the verbs (as several subsequent studies have also done), whereas treating them separately might have offered insights into differences between nominal and verbal predicates. For example, as Huddleston & Pullum point out, in PDE mandative clauses are commonly found after the noun wish, but ‘are hardly possible with the verb, which takes, rather, a modal preterite’
(2002: 999). Also, there are productive accompanying nouns for six of Johansson & Norheim’s verbs that were apparently not included in their study: advice, insistence, motion, order, stipulation and urging.
Other approaches have been taken by, for example, Crawford (2009: 261), who combed through numerous sources to come up with a list of 108 triggers, but then restricted his study to the 33 of those triggers that were found with at least one subjunctive in his corpora. This approach may give an accurate picture of mandative subjunctive use in those particular corpora, but it has its limitations if the broader aim is to look at overall complementation in mandative clauses.
If the intention is for the results of a corpus-based study to be comparable with studies involving different varieties or different (relatively recent) periods, it makes sense to use a fixed set of triggers. Yet it is perhaps unfortunate, though understandable, that the fairly restricted Johansson & Norheim list has ended up being used in so many studies.