Jary and Kissine examine the meaning of imperative sentences, taking the existing relevance-theoretic semantic analysis, in terms of the desirability and potentiality of the described state of affairs, as their point of departure. In their view, a complete account of the interpretation of imperatives has to explain how they can result in the addressee forming an intention to perform an action, and this requires the theory to make room for ‘action representations’ (in addition to factual representations, such as assumptions). They claim that the imperative form is uniquely specified to interface with such action representations.

In the last chapter, we looked at a range of uses of the imperative and suggested that, given an adequate characterisation of directive force, the imperative mood could safely be seen as having the communicative function of conveying that force. However, we argued that it did not necessarily follow from this that directive force is the encoded meaning of imperative morpho-syntax, but suggested, rather, that a suite-of-features approach was an attractive and viable alternative. In this chapter, we consider further data that might lend support to this view, in that they appear to show that the imperative is sometimes used to convey a conditional meaning, often without an associated directive force.

These data are important because if it can be shown that the imperative can receive a non-directive interpretation, then the case for viewing the imperative as encoding directive force is considerably weakened. Furthermore, these data might provide insights into the encoded meaning of the imperative that are occluded in directive uses. However, a number of authors have argued that, despite appearances to the contrary, some or all cases of what, following Davies (1986), we will call ‘imperative-like conditionals’ or ‘ILCs’ do not, in fact, contain imperatives. Rather, they argue, we are misled into seeing imperatives where there are none by the fact that English uses uninflected root forms for imperatives. Such a claim might be supported by showing that morpho-syntactic behaviour associated with less controversial imperatives is not exhibited by such cases. We therefore need to consider the data carefully before proceeding. At the same time, it will be useful to consider what Davies (1986) refers to as ‘imperative-like ultimatums’ or ‘ILUs’, which are similar, at least superficially, to ILCs in terms of structure and interpretation.

This book is about the semantics and pragmatics of imperatives. That is to say, it is about the interaction of linguistic meaning and contextual factors, including speaker intentions, in the interpretation of utterances of a particular type of linguistic form. To undertake to write such a book is to presuppose that the term ‘imperative’ picks out a distinct linguistic type with a meaning consistent across different instantiations, such that interesting things can be said about its contribution to utterance interpretation. In this chapter, we justify this assumption by showing that, cross-linguistically, the ‘imperative form’ can be identified by virtue of its function in communication. For now, we will just say that this function is to signal the performance of directive speech acts such as commands, orders, requests and pleas. However, as we will see in later chapters, the uses to which the type of linguistic form we are investigating can be put goes beyond this narrow range. Moreover, we will also see that identifying the form by virtue of this function in no way commits us, nor any other theorist, to claiming that this function is encoded by that form.

In any interesting sense of ‘encode’, if a form encodes a function, then it does more than merely indicate that that function is its most prototypical use. Rather, if a form encodes a function, then no literal and serious use of that form is possible without its performing the function at hand, so that comprehension of the form is nothing more than relating it to its typical function. This point is very important, as we will see again and again that one of the central issues about the imperative is whether or not every literal use of it necessarily corresponds to the performance of a directive speech act.

Introduction to Part II: from data to theory

In Part I, we have been concerned with defining imperatives and highlighting the semantic characteristics that they display. We have seen that imperatives can be defined in terms of their prototypical pragmatic function: that of performing directive speech acts, themselves best understood as utterances whose function is to provide the hearer with reason to act. We have noted that such a definition of the imperative mood has the attraction of being neither overly restrictive – it does not rule out the possibility that imperatives have non-directive uses – nor overly inclusive – it does rule in forms that are used to perform directives if this is not their prototypical function. We also argued that the imperative mood should be thought of as a sentential category that can be realised either morphologically, syntactically, or by a combination of both. On this conception of imperative mood, the absence of imperative verbal inflection is not sufficient reason to deny that a language has an imperative. It may be that, as appears to be the case with English, the imperative is realised by syntactic means alone. That said, however, it is important to distinguish the imperative from particles associated with directive force, such as please in English. Such forms can typically coerce a directive reading of a range of sentence types, regardless of their prototypical function. Furthermore, they are often restricted in regard to the type of directive in whose performance they can be employed. For example, please can be used in a request, but not in a command.

Modal verbs such as must and can have a number of possible uses. For example, must can express both obligation and certainty (relative to some body of information):

(1) Jack must go to jail.

(2) John must be a spy.

And can can express both permission and possibility.

(3) You can come in now.

(4) You can get the grades you want (if you study hard).

These are not the only uses that these verbs have. A large number of uses have been identified for each of them, and for the other modals (see, for instance, Papafragou (2000); Portner (2009)). The problem that this presents to the semanticist is how best to analyse these uses: are they evidence of ambiguity, or can a kernel of core meaning be identified that explains the range of possible uses each verb has? Kratzer (1977, 1991), in a move that has proved extremely influential in formal semantics, argues for the latter.

In an influential series of publications, the most important of which are collected in his Context and Content (1999), Stalnaker develops an account of assertion in terms of modifications to the common ground. The common ground is the background against which a conversation takes place. It is made up of the propositions that the participants either accept as true or agree to treat as true. (Since propositions can be thought of as sets of possible worlds, the background is thus a set of possible worlds; see Appendix A.) Asserting, according to Stalnaker, is an activity which aims at adding propositions to the common ground. When one asserts, one puts forward the content of one’s assertion for acceptance by the participants in the conversation. If the assertion is accepted, then the content of the assertion becomes part of the common ground.

In this framework, a proposition will be either true or false in any possible world. A proposition p thus picks out a set of possible worlds: all those worlds in which p is true. This is why one can say that propositions are sets of possible worlds (see Appendix A), but, for ease of exposition, we’ll treat a proposition and the worlds it picks out as separate entities. The common ground is made up of consistent propositions. This means that the common ground picks out only those possible worlds in which every one of the propositions accepted as common ground is true. Each of those worlds is a candidate for how the actual world is, as far as the participants are concerned. A felicitous assertion is true in some, but not all, of those possible worlds. Accepting an asserted proposition means that the common ground no longer picks out all those worlds in which the asserted proposition is false. An accepted assertion therefore reduces the number of worlds that are candidates for the actual world, and in doing so reduces uncertainty about how the world is.

The history of twentieth-century philosophy of language and linguistics was marked by the advent of truth-conditional semantics. Modelling the meaning of declarative sentences (and their utterances) in terms of their truth conditions has proved a powerful tool with which to predict their meaning on the basis of structural relations between their parts. Given the success of such methods – which continue to undergo constant sophistication and improvement – it is understandable that there is a great temptation to account for non-declarative sentences in the same way. And yet imperatives and other non-declaratives are a recalcitrant problem for truth-conditional accounts of linguistic meaning, for how can sentences that cannot be true or false be accommodated by such a theory? Those seeking to extend the truth-conditional-semantic programme to cover imperatives have taken two routes. Some have argued that imperatives are in fact declaratives with special features that makes truth judgements infelicitous: imperatives do have truth conditions, the argument goes, but – for various reasons – it is odd to describe them as true or false. Others have argued that the notion of truth should be seen as merely one way in which a sentence or utterance can be satisfied, or fulfilled. Hence, they go on to argue, we should analyse imperatives in terms of the type of fulfilment conditions that apply in their case: compliance conditions. As it turns out that compliance conditions are specified in terms of the truth conditions of a declarative sentence, then imperatives end up being analysed by reference to the truth conditions of the declarative sentence that specifies their compliance conditions. This type of approach is examined in the second section of this chapter. The first is devoted to theories that take the first route by seeking to provide a truth-conditional analysis of the imperative.

The ‘performative hypothesis’ and Davidson’s paratactic account

We will start the discussion of truth-conditional accounts of the imperative by what is known as the performative hypothesis. This view dates back to the 1970s, and, as far as we know, is unanimously rejected today. However, it is worth discussing, if only briefly, because the motivations to reduce non-declaratives to declaratives underlying it – and the problems these raise – extend to more contemporary theories, to be examined later in this section.

In several places of this book we use locutions like ‘a possible world where p’ is true or ‘being true in a possible world’. Furthermore, the notion of possible worlds is at the core of two theoretical tools that are used by some theorists to construct a semantics for imperative mood: Kratzer’s theory of modality and Stalnaker’s conception of context. These are summarised, respectively, in Appendix B and C. Here we will attempt to provide you with an intuitive understanding of the use of possible worlds in semantics. While we hope that it will be sufficient to grasp the matters addressed in the book, bear in might that by no means should this appendix be taken as a tutorial in intensional semantics (see, for instance, Chierchia and McConnell-Ginet (1990); Heim and Kratzer (1998)).

In truth-conditional semantics, the meaning of a sentence is given by the conditions under which it is true. Furthermore, this meaning is compositional, in the sense that the meaning of the whole sentence is a function of the meaning of its components. Take (1).

If language is a toolbox, then imperative sentences are what we reach for when we want to leave our hearer in no doubt that we want him to do something and what we want him to do. This is not to say that this is the only thing they are good for, nor that there is nothing else in the toolbox that will do the job, but simply to state that imperatives appear in some way designed for the job of getting the hearer to do something. We’ll have plenty to say in Chapter 1 about just what an imperative sentence is, but, for the moment, we can make do with a few examples:

(1) Stand by your man.

(2) Mind the gap.

(3) Love thy neighbour.

(4) Do not go gentle into that good night.

(5) Please try to be good.

(6) Somebody call a doctor!

(7) Do help yourself to more tea, vicar.

(8) Don’t anyone move.

What these examples show is that we take imperatives to be of a certain morpho-syntactic type. That a sentence can be used to get someone to do something does not suffice to make it an imperative. That is why we do not count (9) and (10) as imperatives. Again, we’ll have much more to say on this in Chapter 1, where we’ll also examine imperatives in languages other than English.

(9) All children must be accompanied by both parents.

(10) Can you pass the salt, please?

In the previous two chapters, we looked at two general approaches to the semantics of the imperative mood. The theories discussed in Chapter 4 either treat the imperative as encoding directive force or identify directive force with the imperative mood. Those discussed in Chapter 5, by contrast, attempt to explain the imperative by positing a relationship with the declarative mood. The theories discussed in this chapter take a different approach in that they neither posit a direct link between directive force and the imperative, nor seek to explain it in terms of another sentence type, such as the declarative. Rather, they treat the imperative as a distinct semantic type, its relationship with directive force constrained by what it encodes, but ultimately mediated by pragmatic considerations.

The theories in question can be grouped according to how integrated they are with a pragmatic theory. The first group have little to say about specific pragmatic mechanisms: they recognise that a pragmatic theory is needed to move from the semantics they propose for the imperative to directive force, but they make no claims about the details of the kind of theory required. Those in the second group, by contrast, offer accounts of the imperative that are elements of ambitious theories of linguistic communication. These aim to integrate imperatives within models of how linguistic communication is achieved. Both pragmatic theories we will discuss – Bach and Harnish (1979) and Wilson and Sperber (1988) – abide by the post-Gricean view that the identification of speaker meaning relies heavily on hearer inference. Consequently, they analyse mood in terms of the contribution it makes to the inferential side of utterance interpretation. However, let us begin by looking at the first group, i.e. those who make no specific claims about the pragmatic processes that, as they acknowledge, are required to complement the semantics they propose. In this group we place Huntley (1984), Davies (1986) and – somewhat more hesitantly – Portner (2009).

In Part I, we defined the imperative mood as a sentence type prototypically associated with the performance of directive speech acts. We also noted that characterising the imperative in this way does not amount to defining the imperative as encoding directive force. Nevertheless, it does not rule it out, either. In this chapter, we look closely at theories that do treat the imperative as encoding directive force. Put simply, according to these theories, the meaning of the imperative sentence just is directive force.

We will divide our survey of these theories into two main sections. The first will be devoted to models that, explicitly or implicitly, claim that the meaning of the imperative should be broken up into (at least) two components: a propositional content that determines which situation that imperative is ‘about’, and some other component, which encodes directive force. This way of thinking is perhaps the most traditional, and has its roots in Frege’s distinction between sense and force (see Recanati 2013). Another option is to reject this distinction, and to build a theory of imperatives as distinct semantic objects that consitute directive force. We will survey three such theories in the second part of this chapter.