What Does the Law Say?Semantics and Pragmatics in Statutory Language

(forthcoming in ANALISI E DIRITTO)

Andrei Marmor

USC Legal Studies Research Paper No. 07-9

LEGAL STUDIES RESEARCH PAPER SERIES

University of Southern California Law SchoolLos Angeles, CA 90089-0071

This paper can be downloaded without charge from the Social Science Research Networkelectronic library at http://ssrn.com/abstract=1009622

Draft, August 2007, forthcoming in Analisi e diritto

What Does The Law Say?

Semantics and Pragmatics in Statutory Language

Andrei Marmor

Most of the time it is perfectly clear what the law says. We simply understand the

meaning of the relevant legal instruction and we know how to apply it to particular cases

or, at least, we know how it applies to this case. To be sure, even the clearest legal

instruction does not necessarily settle the particular legal case at hand. There may be

other relevant laws that conflict with this one, and then judicial interpretation may be

required to determine which one prevails. Or, sometimes the law may be crystal clear, but

it yields very problematic, perhaps even absurd, results in the particular case at hand. And

then, perhaps, judges may need to modify the law or set it aside. In other words, a clearly

understood law does not always determine the legal result. Barring such potential

conflicts, however, it is often the case that what the law says is just clear enough to

determine what needs to be done. This is basically a matter of linguistic communication,

or so I will presume here. Legal prescriptions are acts of communication. As in other

instances of linguistic communication, more often that not we simply understand what

the speaker has communicated or, at least, we understand enough in the relevant context.

My purpose in this essay is to explore what this understanding consists of; in particular, I

want to distinguish the semantic and pragmatic aspects of the language of the law,

analyzing the various components that determine what the law says in a particular

2

context.1 When we analyze these various linguistic components, we will be able to see

what determines whether the law is clear enough or not, and how. For the sake of

simplicity, I will confine my discussion to statutory law. Other forms of legal prescription

might differ from statutory regulation in several respects, but I will not go into this here. I

will begin with some general comments about semantics and pragmatics, laying out some

of the tools and distinctions that I would like to use here. I will then suggest that the

pragmatic aspects of understanding statutory law differ in important ways from the

pragmatics of ordinary conversations. I will focus on some of these differences which

make the understanding legal language somewhat problematic, and I will point toward

some solutions.

1. The Conceptual Framework.

Understanding an instance of linguistic communication is a complex matter.

Different components of language and context contribute to our ability to understand

what someone has said or asserted. How to draw the line between these various

components is a contentious issue. Nevertheless, I believe that we can draw some basic

distinctions. In particular, we need to distinguish between –

(i). literal meaning of words and sentences (basically determined by the rules of

language)

(ii) the semantic content of an expression/sentence relative to a context of utterance,

(iii) the assertive content of an utterance in the context of speech, namely, the

proposition(s) asserted/stated by the speaker in the specific context and, finally,

1 I have come to realize that in my Interpretation and Legal Theory (2nd ed, 2001, Hart Publishing) I have not paid enough attention to the pragmatic aspects of understanding an act of communication. This essay is a belated attempt to remedy, at least to begin to remedy, this serious neglect.

3

(iv) further communicative content that the speaker is committed to by uttering the

expression in the specific context (which may include, for example, conversational

implicatures, presuppositions, and perhaps other content that is obviously and

transparently entailed by the speech in its context of utterance).

This list, which I will explain shortly, does not exhaust the main components of

linguistic communication. First, we must add the various syntactical aspects of language,

in particular, the rules of compositionality that determine how expressions can be

combined with other expressions to engender particular meanings. Second, there is a

speech-act aspect of language, concerning ways in which we do things with linguistic

expressions, that is, beyond the action of making a statement or expressing a proposition.

Although these are very important aspects of language that contribute to our ability to

understand a speech situation, I will not be concerned with them. I think that they are not

particularly relevant to the issues I discuss here.

A brief clarification of this conceptual framework should suffice. Let us begin

with the distinction between the literal meaning of words and the semantic content of

their use in specific speech-contexts. The semantic content of an expression relative to a

context is basically a combination of what is contributed to the content of communication

by the literal meaning of the words/sentences uttered and objective features of the

utterance, such as who is speaking, time, place, and other relevant contextual facts.

Probably the least controversial example of the distinction can be given by considering

pure indexicals (e.g. I, now, today, next week). If the speaker says ‘I will go to the

cinema today’, what has been said is partly determined by the literal meaning of the

words used, and partly by the objective referent of ‘I’ and ‘today’. The question of

4

whether demonstratives (like he, you, that, this etc.) work in the same way is rather

controversial, and I will not go into this here. 2 Intuitively, however, the idea is that the

meaning of a sentence in a context of utterance is often a combination of the literal

meaning of the words used and some objective features of the utterance that together

determine what the sentence says in the relevant context.

In addition to the semantic content of the utterance, further contextual background

may be required in order to understand what the speaker has asserted. In other words, the

assertive content of an expression – the content of the proposition that the speaker

asserted -- may depend on specific contextual background, knowledge of which is shared

by speaker and hearer. Consider, for example, possessive ascriptions: Suppose a speaker

says “I have just finished reading Hilary’s book.” In order to know what has been

asserted here, one would need to know whether “Hilary’s book” refers to a book that

Hilary wrote, or a book that belongs to her, or maybe a book about her. Under normal

circumstances, the propositional content of such expressions is specified by the context of

the utterance, knowledge of which is shared by speaker and hearer. Without this

contextual background it is impossible to determine what has been asserted.

Here is another example. Consider these two cases of numerical expressions.

A. Do you have two children?

B: Yes.

A: Do you have two dollars?

B: Yes.

Now we can clearly see the difference: in the first case, it is pretty clear that B

says that he has two and only two children; if it turns out that B has 4 children, we’d be

2 D Kaplan, ‘Demonstratives’ in Almog et al eds, Themes from Kaplan, Oxford 1989 and cf Soames, ‘Direct Referece, Propositional Attitudes, and Semantic Content’, Phil Topics, vol 15, 1987, 47.

5

surprised by his answer to A’s question. In the second case, the opposite is true; We

would normally assume that B has at least two dollars, but possibly more. Presumably,

what accounts for the difference is the pragmatic aspect of such conversational situations.

It has to do with some general contextual background that is shared by speaker and

hearer.3

Finally, it is widely agreed that the content of an assertion can go substantially

beyond what has been explicitly said or asserted. Let’s call it the implied content of the

utterance. Roughly, then, the implied content of S in context C, can be defined as the

content that the speaker, in the specific context of C, is committed to by uttering S, and

that the hearers are expected to know that the speaker is committed to, and the speaker

can be expected to know this. Thus, in addition to what has been explicitly said by an

utterance it can be said to contain such as what is conversationally implicated by it, what

is presupposed by it, and perhaps other content that is obviously and transparently

entailed by the utterance in its context of utterance. Note the qualification of “obviously

and transparently” – surely not everything that is actually entailed by saying that P (in a

context C) forms part of the implied content of an utterance. Speakers cannot be expected

to know everything that is logically or otherwise entailed by the content of what they

say.4

The paradigmatic example of implied content is the concept of implicature,

introduced by the philosopher Paul Grice.5 Let me briefly review some of Grices’ main

ideas about implicatures. His main insight, I take it, is that our ability to understand

3 See Soames, ‘Drawing the Line Between Meaning and Implicature – and Relating both to Assertion’, forthcoming in Nous. (Available on Soames’ website: http://www-rcf.usc.edu/~soames/) 4 I will say more on the significance of this towards the end of the paper. 5 Grice, P. (1989), Studies In The Way of Words, Harvard University Press.

6

content of expressions beyond their semantic/assertive6 content is due to a combination of

two kinds of factors: general norms of conversation that apply to the relevant speech

situation, and specific contextual knowledge that is shared by speaker and hearer in the

circumstances of the utterance. In normal conversational situations, when the main

purpose of speech is the cooperative exchange of information, there are certain general

maxims that apply. Grice helpfully listed and classified these maxims of ordinary

conversation, and they are basically as follows:

a. maxims of quantity – make your conversational contribution as informative as required,

viz., don’t say too little and don’t say too much.

b. maxims of quality – don’t say what you believe to be false, and don’t say something if

you do not have adequate evidence for it.

c. maxim of relevance – make your contribution relevant to the conversation.

d. maxims of manner – avoid obscurity, ambiguity, be brief and orderly.7

As noted, these maxims apply to ordinary conversations. In other speech

situations, some of these maxims may not apply, and others might be followed instead.

Now, the maxims are norms that directly instantiate the specific functions or purposes of

communicative interactions and facilitate those functions. Some specific conventional

settings, however, may determine what kind of maxims are relevant and should be

followed. It is part of the conventions of theater, for instance, that some of the regular

conversational maxims are suspended, and this is something that follows from the

6 The question of beyond what, exactly, conversational implicatures operate is somewhat controversial. Grice typically speaks about the distinction between what is said and what is implicated; presumably, by ‘what is said’, Grice includes assertive (and not just semantic) content. Scott Soames, however, argues that a great deal of assertive content is also partly determined by pragmatic features of conversation, including implicatures. See, S Soames, ‘Drawing the line…’ 7 Grice, Studies in the Ways of Words, 28

7

conventions constituting theater. As we shall see in a moment, the legal case is rather

challenging in this respect. I’ll get to it a minute.

The next step is introduced by the notion of implicatures. A certain content is

implicated by a speaker if it is not part of what the speaker said (viz, it is not part of the

semantic/assertive content of the sentence uttered in the context of its utterance), but

nevertheless implicated by what the speaker said in the specific speech situation, given

the conversational maxims that apply. In other words, a speaker S conversationally

implicates q by saying p in context C, iff

a. S observes the relevant conversational maxims in C,

b. the assumption that S meant (or intended that) q is required in order to make sense of

S’s utterance of p in context C, given the conversational maxims that apply,

c. S believes/assumes that his/her hearers can recognize condition b, and can recognize

that S knows that.8

To mention one familiar example, consider the following situation: A, standing

near his immobilized car that ran out of gas, asks for the help a local person, B, passing

by. Knowing these facts, B says “there is a gas station in the next village”. Now, B has

not actually said that the gas station is open and would have gas to sell. But given the

maxims of conversation (e.g. be relevant, don’t say something you believe to be false,

etc.), it would be natural to assume that these were implicated by what B has said.9 It is

content that B is committed to, given the situation and the conversational maxims that

8 This last condition of transparency is actually rather problematic and controversial. Grice himself was aware of a serious problem here considering the implicatures involved in using disjunction. See Soames, ‘Drawing the line….’ 9 Grice, p 32.

8

apply. As Grice himself emphasized, there are two main features essentially associated

with conversational implicatures:

1. Implicatures are always cancelable. The speaker in our example could have added,

“but I’m not sure that the gas station is open” – in which case, the implicature would be

explicitly canceled. Generally speaking, cancelability is an essential feature of

conversational implicatures.

2. Conversational implicatures are entirely context specific; they are not conventionally

determined by the rules of language. There is always some derivation, as a Grice called it,

that leads us to construe the content of an implicature; some story has to be told to make

it explicit.

So, all this was in a way of background. The main point of this theoretical

framework was to show that there is a great deal of pragmatic enrichment (a term coined

by Scott Soames) one needs to take into account in explicating the communicative

content of ordinary speech. What a speaker actually says in a context of speech often

depends on a pragmatic context, knowledge of which is presumed to be shared by

speaker and hearer. As we shall see later, however, the case of conversational implicature

is only one standard example; other forms of implied content are different in some crucial

respects.

2. The Pragmatics of Legislative Language.

When we focus on the legal context, some of these aspects of communication turn

out to be particularly problematic. In order to see some of these problems, we need a

9

clearer picture of what is required for pragmatic enrichment. Generally speaking,

pragmatic enrichment requires (at least) three conditions:

1. A speaker who has certain communication intentions.

2. A conversational context that, at least to some extent, is common knowledge,

that is, shared by speaker and hearer.

3. Some Gricean conversational maxims that apply to the relevant speech

situation.

In the legal case, I want to argue, all these conditions are rather problematic Let me

explain, taking these three conditions in turn.

A. Communication Intentions.

The problems about the first condition are well known and I will not dwell on them here.

The speaker in the legal case is often a parliament, or some other legislative agency. As

we all know, attributing communication intentions to a collective agency raises some

conceptual problems that have to be resolved. Since I have elaborated on this

elsewhere10, I will not repeat the argument here. Suffice it to say that I don’t believe that

the conceptual problems of attributing intentions to a collective body are insurmountable.

A group of people can share certain intentions such that, under certain circumstances, the

intention is rightly attributable to the group, as such. In any case, I will put aside these

general conceptual issues here. Some of the specific problems about determining the

relevant kind of intentions that can ground implicatures I will explore shortly.

B. Conversational Context.

The context of legislation is, of course, often much more opaque than a regular

conversational situation. To begin with, there are problems of insufficient information.

Judges and litigants are not parties to the legislative conversation, so to speak, and they

10 See my Interpretation and Legal Theory 2nd ed. chapter 8.

10

have to rely on secondary sources to gather the relevant information. How reliable those

sources are varies greatly between different legal cultures and practices, and it varies

greatly with particular circumstances. Note, however, that part of what makes context

often clearer than one would have assumed is the fact that the law typically speaks to a

legal community, not to lay people. Legislatures in a modern society mostly address their

laws to legal experts, such as judges, lawyers, administrative agencies, etc. There is

typically a legal culture that would share a great deal of information and contextual

knowledge that is much greater than the relative contextual knowledge of the population

at large. This often makes context clearer than one would have assumed.

Nevertheless, even when information is fully available, the conversational context

may still be problematic. This is mostly due to the fact that there are no clear criteria

about relevance of contextual background. In ordinary conversations, when the purpose

of speech is the cooperative exchange of information, the context is relatively closed, as it

were, with boundaries determined by the conversational situation. The conversational

situation of an act of legislation is often much more open-ended, without clear

boundaries. We know where the story ends, that is, with the act voted on, but we often

don’t quite know where it begins and how wide the scope of the background should be

taken to be. And, to complicate matters further, it may not be entirely clear who the

participants to this conversation are. Should we include only legislators? That seems too

restrictive; an act of legislation is often the result of a complex interaction between

legislative bodies and other political agents, such as administrative agencies, lobbying

groups, and various other constituencies that try to influence the emerging legislation.

Once again, the main problem here is not the complexity of the situation but the fact that

11

there are no clear criteria of relevance. True, some legal cultures may have certain rules

or conventions that partly determine criteria of relevance. A legal system may have a rule

or convention determining, for example, that lobbying should be excluded from

consideration and should not, as matter of law, be part of the conversational background

of understanding legislation. But usually, even in legal cultures that have formed some

rules about such issues, a great many of them remain undecided.

For now, I am just pointing out the difficulties. How exactly these difficulties

affect the scope of what understanding a legislative directive consists in, I will explore

later. Let us now turn to the third, and main, difficulty.

C. Conversational Maxims.

It is here, I think, where most of the problems reside. The Gricean maxims of

conversational implicatures are the norms that apply to an ordinary conversation, where

the purpose of the participants is the cooperative exchange of information. But the legal

case is quite different. The enactment of a law is not a cooperative exchange of

information. Therefore, we should not be surprised if some of the Gricean maxims may

not apply to the context of legislation and, more problematically, it is often not clear

which norms do apply. The main reason for the difference resides in the fact that

legislation often involves strategic behavior. Let me mention some familiar examples:

The most familiar aspect of legislation is that it is almost always a result of a

compromise. Compromise often consists in what I would like to call tacitly

12

acknowledged incomplete decisions – that is, decisions that deliberately leave certain

issues undecided.11 This is closely tied to the problem of collective agency:

A says P intending to implicate x

B says P intending to implicate not- x

A and B act collectively, intending their collective speech in saying P to remain

undecided about x.

Now, the problem is that the underlined ‘intending’ is often not so clear; we may

have cases of conflicting intentions, hopes, expectations, etc. (viz, both A and B

intending – or hoping, or expecting -- their intentions to prevail). One dimension of this

problem is that we may have different and conflicting conversational maxims here: we

may have a norm that assumes something like ‘follow the intentions/implications of the

median legislator’ or, quite the opposite, ‘follow the intention/implications of those who

initiated the bill’. Let me explain what I mean here.

Suppose that there is a bill imposing some new restrictions on carbon emissions

on motor vehicles. Assume that there are 100 member of parliament. Typically, the bill

would be initiated by a group of legislators who are particularly concerned with the

environment and the dangers of global warming, etc. Assume that this group, call it the

‘greens’, consists of 20 legislators. In order to pass the new bill, the greens need the

support of at least 31 other legislators (call them the median legislators), many of whom

may not feel so strongly about environmental protection. In fact, most of them may be

rather reluctant to impose further burdens on the automobile industry. Thus, in order to

gain the majority needed to pass the bill, the greens have to act strategically. They have to

convince the median legislators to support the bill. Typically, they can do that in two

ways (and usually, in a combination of both): either by giving them something in return 11 Three is nothing new in this idea, it has been noted by numerous writers.

13

(e.g. their vote on a different bill that the others care more about), or by modifying the

suggested bill to accommodate the concerns of the less enthusiastic supporters. In this

latter case, we will normally see some modifications to the initially proposed bill,

changes in its formulation that result from negotiation and compromise. It is in the very

nature of such compromises, however, that parties can reach an agreement because they

can employ the device of tacitly acknowledged incomplete decisions. That is, the

opposing parties will deliberately leave some formulations of the bill open to conflicting

interpretations, hoping that the interpretation they favor will eventually prevail.

Clearly, the problem in such cases is that there are several conversational maxims

that can apply. Different content might be implicated by different groups of legislators,

and it is far from clear which one ought to prevail. If you follow the intentions of the

initiators of the bill, the greens in our case, you would give effect to intentions of a

minority; after all, we assumed that the greens do not have enough votes to pass the bill.

If, on the other hand, you follow the intentions of the median legislators, you give effect

to the intentions of those who did not really care much about the objectives of the bill and

often know very little about it (and note that they also form a minority).12 In Gricean

terms, the problem here is an inherent indeterminacy about what should count as the

relevant contribution to the conversation. And if you don’t know what counts as relevant,

there is no way to figure out the implied content of the speech.

To give another example of strategic behavior, consider cases in which the

legislature deliberately speaks in several voices, so to speak. There are legislative cases in

which the legislature intends to convey one message to the public at large, and a different

12 See, for example, Rodriguez, Daniel B. and Weingast, Barry R., "The Positive Political Theory of Legislative History: New Perspectives on the 1964 Civil Rights Act and its Interpretation" University of Pennsylvania Law Review, Vol. 151, 2003

14

one to agencies, or the courts, etc. Meir Dan-Cohen has explained this phenomenon, and

its rationale, in his detailed study of some criminal defenses.13 To mention just one

example of the rationale of such legislative double-talk, consider the defense of duress.

This is a very problematic defense: on the one hand, considerations of deterrence weigh

against recognizing such a defense; after all, we don’t want to encourage people to

succumb to threats and commit crimes out of fear or weakness of character. On the other

hand, considerations of human compassion call for a recognition of such a defense. It

would be very inconsiderate, almost inhuman, to punish people for things they have done

under enormous threat, especially when we know that we would have succumbed to the

threat ourselves in those same circumstances. Now, this is a serious conflict, but one that

allows a certain solution: the law could generate the impression that it does not recognize

duress as a defense, or that it would only grant it in extremely dire circumstances, but at

the same time, the law could instruct the courts to grant the defense when considerations

of fairness and compassion call for it. As Dan-Cohen demonstrates, this is more or less

what actually happens in common law. And it makes a lot of sense.

I hope you can see that the temptation to use this device might be great in many

legislative contexts. Legislators may wish to create the impression that they are doing one

thing -- e.g. seriously restricting campaign finance contributions -- while actually trying

to do the opposite -- allowing such contributions to flow freely but less transparently.

What we have in such cases, is almost like a conflicting implicature: looked at from one

angle, the legislature implicates one thing, looked at from a different angle, it implicates

the opposite.

13 See M Dan-Cohen, ‘Decision Rules and Conduct Rules: On Acoustic Separation in Criminal Law’, in his Harmful Thoughts, (Princeton 2002), 37

15

Furthermore, as the two examples mentioned just show, there is no general policy

that can apply across the board. In some cases, like the example of duress, the legislative

double-talk makes a lot of sense and is probably morally commendable. In other cases,

like the campaign finance example, the double-talk is morally questionable.

Let me pause to take some stock. I have tried to show that unlike regular

conversational contexts, where the parties to the conversation aim at a cooperative

exchange of information, a non-cooperative form of communication is present in the

legislative context. Legislation is often a strategic action that tries to overcome the lack of

initial cooperation between the relevant agents. This, I have tried to argue, makes it very

doubtful that we can get a clear sense of what would be the relevant Gricean maxims that

apply to such legislative speech situations.

In one crucial sense, however, the picture is not as muddy as it may seem. Over

time, the norms of statutory interpretation that are adopted by the courts partly determine

the conversational maxims of legislation. In following certain norms about the ways in

which courts will interpret statutory language, the courts basically create some kind of

Gricean maxims for the legislative context. For example, the extent to which courts are

willing to hear evidence about statutory history would partly determine the norms of

relevance about legislative implication. These norms would partly determine what counts

as a relevant contribution to the “conversation” between legislators and the courts, so to

speak. Thus, to some extent, and greatly depending on the interpretative culture of the

courts, some Gricean maxims do tend to be present even in the legislative context. But I

think we may safely assume that such maxims are much less determinate in content and

scope that those norms that apply to ordinary conversation. The general conclusion to

16

draw from this is that implicatures cannot be frequently relied upon in determining the

communicative content of legislative speech.

3. Implied Law?

As noted earlier, in ordinary conversational situations, the content of an assertion

can go substantially beyond what has been explicitly said or asserted. A great deal of

conversational content is often implied, not asserted. When I say, in a normal

conversational situation, that ‘I have two children’ -- I imply that I have two and only

two children, and not, say, three of four; but I only imply this, I haven’t quite said it. In

other words, in uttering these words in an ordinary context, I am committed to the content

that the number of my children is exactly two and not, as it were, at least two. Under

normal circumstances, such a commitment is one that the hearers are expected to

recognize, namely, they are expected to know that the speaker is committed to the

implied content and the speaker is expected to know this.

The question we must explore here, however, is whether such implied content of a

speech situation in the legislative case necessarily forms part of the law. In other words,

suppose that the legislature enacted a rule saying P in context C, and the context is such

that P would normally imply Q, that is, we would rightly infer that in saying P in context

C, the legislature is committed to Q. Is Q necessarily part of the law? Is it part of what the

law says? In order to answer this question, I think that we must distinguish between two

types of such communicative commitments: semantic and conversational. The thesis I

want to suggest here is that only semantic commitments would normally form part of the

law.

17

Before I explain the distinction, however, let me clarify one important assumption

that I have alluded to earlier: it is sometimes suggested by legal philosophers that

legislatures should be assumed to be committed to all the content that is logically or

otherwise entailed by what the legislature has explicitly communicated. There are two

possible grounds for holding such a thesis, but I think that both of them are wrong. First,

it can be seen as part of a theory of communication; it may be assumed that, generally, a

speaker is committed to everything that is actually entailed by what the speaker explicitly

said. As we noted earlier, this does not make sense. It makes no sense to maintain that we

are generally committed to have implied, in any sense whatsoever, everything that is

actually entailed by what we say on this or that occasion. In general, even if it is the case

that p entails q, in saying that “p” one has not necessarily implied that q. Surely, all those

who have used and expressed the axioms of arithmetic for centuries have not implied in

any sense whatsoever the truth of Godel’s theorems, but as we now know, Godel’s

theorems are entailed by it.14 Similarly, someone who says “cruel punishment ought to be

prohibited” is not necessarily committed to have implied that capital punishment ought to

be prohibited, even if it is true that the latter conclusion is entailed by adding the relevant

moral facts. Therefore, I assume that those who claim that the legislature should be taken

to be committed to everything that is entailed by what the legislature has enacted are

actually making a moral claim, not a linguistic one. The claim is that the legislature ought

to be regarded as committed to an overall coherence of its prescriptions and to everything

that somehow necessarily follows from it. Once again, this is normative thesis and should

be assessed as such. I have argued elsewhere that as a normative thesis, this idea fails,

14 I owe this example to Scott Soames.

18

and I will not repeat the argument here.15 From a linguistic perspective, however, I hope

it is clear enough that a speaker is not necessarily committed to everything that is actually

entailed by what she said. So let us return now to the question of what kind of

implications the legislature would be committed to.

In discussing Gricean implicatures, we have already seen one paradigmatic case

of conversational implication, that is, cases in which the speaker is committed to have

implied a certain content. It is crucial to note, however, that in such cases, the implication

is conversational, that is, entirely context dependent. As we noted earlier, it is a hallmark

of such implicatures that they can be explicitly cancelled by the speaker. There are,

however, other cases, in which the implication is not cancelable. Grice called them

‘conventional implicatures’, others refer to them as presuppositions.16 Either way, the

point is that there are cases in which the speaker is committed to a certain content simply

by virtue of the words she has uttered, regardless of the specific context of conversation.

In other words, certain implications are semantically encoded in the expression used.

Consider the following examples:

1. “X is A but B” – implicating that the conjunction of X being both A and B is

somehow surprising or particularly interesting etc.

2. “Even X can A” – implicating that (i) there are others, besides X, that can A,

and that (ii) amongst the relevant agents, X is least likely to A.

3. “X too is A” – implicating that there are others, besides X, who are A.

4. “X quit A-ing” – implicating that X had done A in the past with some

regularity.

15 See my ‘Should We Value Legislative Integrity?’, in R Bauman & T Kahana (eds.), The Least Examined Branch, The Role of Legislatures in the Constitutional State, Cambridge University Press, (2006), pp. 125-138. 16 See, for example: Karttunen & Peters, (1979), ‘Conventional Implicature’, Synthax and Semantics, Vol. 11, and cf. Soames, ‘Presupposition’ in Gabbay & Guenthner (eds), Handbook of Philosophical Logic, (1989), vol 4, 553.

19

5. “X insinuated to Y that A” – implicating that X did not explicitly say to Y that

A.

6. “In his speech X touched (remarked) on A” – implicating that A was not the

main subject of X’s speech.

All these cases exemplify one and the same phenomenon, namely, that a certain

content is implicated (or entailed) as an integral part of the literal meaning of the words

used. Furthermore, note that none of the two conditions Grice attaches to conversational

implicatures apply here: the implications in 1-6 are not cancelable by the speaker and,

more importantly, there is no need for any derivation, for any story to be told, as it were,

about how we got here. The implied content simply forms part of what the relevant words

literally mean; it is semantically encoded in the literal meaning of the relevant expression.

Note, however, that in all these cases, the implied content, though semantically

encoded, does not form part of what has been explicitly said. It is a commitment, and one

that necessarily follows from the words used, but not explicitly part of what has been

said. Suppose S utters the words “X is a politician but he is quite honest”. The use of

“but” clearly implies that S believes/assumes that politicians are not generally very

honest people. However, S did not quite say this. Now, it is very typical in such cases that

when we are confronted with an explicit denial of the implied content, we would feel a

certain unease. Suppose we confront S with a request for clarification: “Are you saying

that politicians are not usually honest people?”, and then, in response, S says: “Oh, no, I

did not say this.” Well, true enough, S did not say this, but we would also feel that S is

disingenuous in his reply. We cannot feel comfortable with it.

20

Now let us return to law. I want to suggest a positive and a negative claim. The

positive argument is that implied content of the semantic kind does form part of what the

law says. This should follow straightforwardly from the fact that such implied content, as

we have seen, is semantically encoded in the expression used. If by saying P, in

circumstances C, the legislature semantically implies that Q, then Q is part of what the

legislature has expressed. That is so, because this is a kind of commitment that follows

from the words used and thus not cancelable either. An attempt to cancel the commitment

would have been disingenuous or perplexing. On the other hand, when the implied

content is conversational, things become much more muddled. As we have noted in the

previous section, implicatures require conversational maxims to be inferable, and those

maxims in the legislative context are rather indeterminate and often lacking. To be sure, I

am not trying to suggest that conversational implicatures never form part of what the law

says; sometimes they may. It may be the case that a certain implicature is inferable in a

legislative context in a pretty straightforward way. But these cases, I submit, would be

rather exceptional and infrequent. Generally, due to the strategic nature of legislative

acts, we do not have conversational maxims that are sufficiently determinate to allow us

to infer that a certain implicature forms part of what the law says.

The distinction I suggest here, between semantically implied content – that

normally forms part of what the law says – and conversationally implied content – that

normally doesn’t, still leaves room for some intermediate or borderline cases. Those are

mostly cases in which a speech situation involves a presupposition. Some presuppositions

are difficult to classify, that is, difficult to determine whether they are akin to semantic

implication or just conversationally implied. Consider this conversation:

21

A: Can you come to my party tonight?

B: I’m afraid that I cannot, my cousin is in town tonight.

Now it seems pretty clear that B’s answer presupposes a certain content, namely, that if

B’s cousin is in town, B has to see his cousin or such. If it turns out that B has no

intention of seeing his cousin tonight, we would feel that his reply to A’s question was, at

best, irrelevant. Notice, however, that the implication here is not semantic. The implied

content is not semantically encoded in the expression used. In this sense, the implied

content looks more like an implicature. But it is not quite an implicature either, since it is

not cancelable. If B added to his reply something like “but, actually, I have no intention

of seeing my cousin”, or such, he would have expressed a very perplexing conjunction.

What we see here is that not every implication (or presupposition) can be

classified as either semantic or conversational. Some cases have features of both. I am not

quite sure how to classify such cases in the legal context. There is something to be said in

favor of the assumption that such presuppositions are part of what the law says,

particularly since they are not cancelable. On the other hand, since presuppositions of this

kind are not semantically encoded in the relevant expression, and take a certain pragmatic

inference to reach, it is not entirely clear to me that we are entitled to assume that they

necessarily form part of what the law says. Perhaps on linguistic grounds, such case are

not determinable.

Barring such intermediary cases, however, I tried to suggest here that we should

distinguish between two main types of implied content: conversational implicatures that

do not, normally, form part of what the law says, and semantically implied content that

normally does. This is a tentative suggestion, no doubt it requires further thought and

22

refinement. The pragmatic aspects of legal language are very complex, and certainly

much more work needs to be done in this area.17

Andrei Marmor Maurice Jones Jr. Professor of Law & Professor of Philosophy University of Southern California amarmor@law.usc.edu

17 I am grateful to Elizabeth Garrett for comments on a draft of this paper, and to the participants of the workshop at Bocconi University in June 2007, for the stimulating comments and discussions that have prompted the writing of this paper. A particular debt of gratitude I owe to my colleague Scott Soames for his invaluable insights and discussions of these issues during a seminar we taught together in the fall of 2006.