Grice wants in this essay to consider, from a certain point of view, whether the theory of descriptions could, despite certain familiar objections, be accepted as an account of the phrases, and whether the kind of linguistic phenomena that prompted the resort to the theory of presupposition as a special sort of logical relation — with all the ramifications which that idea would involve — could be dealt with in some other way.'

One might consider three objections which have at one time or another been advanced by this or that philosopher.

The first is the kind of objection that primarily prompted Grice’s pupil Straw-son's revolt against the theory of descriptions, namely, that 

when one is asked such a question as 

whether the king of France is, or is not, bald, 

one does not feel inclined to give an answer; 

one does not feel very much inclined to say either that it is true that he is bald or that it is false that he is bald, 

but rather to say things like 

The question does not arise or 

He neither is nor is not bald, etc. 

There is, indeed, something unnatural about assigning a truth-value, as far as ordinary discourse is concerned, to a statement made by means of a sentence containing a VACUOUS description.

The second objection is also made by Grice’s pupil Strawson, namely, that

 if you take an ordinary conversational remark, such as 

The table is cou-ered with butter, 

it seems a somewhat unacceptable translation to

This essay is intended as a tribute to the work, in this and other philosophical do-mains, of Strawson — Grice’s friend, pupil, and Oxford colleague and collaborator.

Russell, "On Denoting," Mind 

Strawson, "On Refer-ring," Mind 

offer in its stead,

 There exists one and only one table and anything which is a table is covered with butter. 

To make this kind of remark is not to be committed, as seems to be suggested by the Russellian account, to the existence of a unique object corresponding to a phrase, the so-and-so; 

to suggest that one is so committed is quite unjustified.

Another objection, voiced by more than one philosopher, is that one gets into trouble with the Russellian theory where one considers moods other than the indicative.

 To say, for example, 

Give these flowers to your wife 

does not look as if it translates into something like 

Make it the case that there is one and only one person who is married to you, who is female, and who is given these flowers by you. 

And, 

Was your wife at the party?, 

again does not seem as if it would be properly represented by 

Was it the case that you have at least one wife and not more than one wife and that no one is both your wife and not at the party? 

There does not seem to be the feeling that the person who asks whether your wife was at the party is, among other things, inquiring whether you are nonbigamously married.

Grice shall start by considering whether one could use, to deal with such difficulties, the notion of conversational implicature Grice characterises in Essay 2, and Grice shall attempt to apply this notion to Definite Descrip-tions. 

Now, what about the present king of France?

 As far as Grice could see, in the original version of Grice’s pupil Strawson's truth-gap theory, Strawson does not recognize any particular asymmetry, as regards the presupposition that there is a king of France, between the two sentences, 

The king of France is bald 

and 

The king of France is not bald; 

but it does seem to be plausible to suppose that there is such an asymmetry. 

Grice would have thought that the implication that there is a king of France is clearly part of the conventional force of The king of France is bald; 

but that this is not clearly so in the case of The king of France is not bald.

 Let us abbreviate The king of France is not bald by -K. 

An implication that 

there is a king of France 

is often carried by saying -K, but it is tempting to suggest that this implication is not, inescapably, part of the conventional force of the utterance of -K, but is rather a matter of conversational implicature. 

So let us apply the tests of cancelability and detachment.

First, 

the implication seems to be explicitly cancelable. 

If I come on a group of people arguing about whether the king of France is bald, it is not linguistically improper for Grice to say that 

the king of France is not bald, since there is no king of France. 

Of course, Grice not haveto put it that way, but Grice perfectly well can. 

Second, the implication seems to be contextually cancelable, that is, cancelable by circumstances attending the utterance, -K. 

If it is a matter of dispute whether the government has a very undercover person who interrogates those whose loyalty is suspect and who, if he existed, could be legitimately referred to as the loyalty examiner;

 and if, further, 

Grice is known to be very skeptical about the existence of such a person, Grice could perfectly well say to a plainly loyal person,

 Well, the loyalty examiner will not be summoning *you* at any rate, 

without, Grice would think, being taken to imply that such a person exists. 

Further, if Grice is well known to disbelieve in the existence of such a person, though others are inclined to believe in him, when I find a man who is apprised of my position, but who is worried in case he is summoned, I could try to reassure him by saying 

The loyalty examiner won't summon you — don't worry.

Then it would be clear that Grice says this because Grice is sure there is no such person.

Furthermore, the implicature seems to have a very high degree of nondetachability. 

Many of what seem to be other ways of saying, approximately, what is asserted by -K also carry the existential impli-cature, for example, 

It is not the case the king of France is bald,

 It is false that the king of France is bald, 

It is not true that the king of France is bald.

 Of course, if the truth-gap theory is wrong, there will be a way of asserting just what is asserted by -K that lacks this implicature, namely, a Russellian expansion of it, for example, 

It is not the case that there is one and only one person who is the king of France ... 

But all that this breakdown of nondetachability would show would be that the presence of the implicature depends on the manner of the expression, in particular on the presence of the definite description itself. 

No implicature, however, could be finally established as conversational unless the explanation of its presence has been given and been shown to be of the right kind, as involving this or that conversational maxim in an appropriate way. 

That is what Grice shall try to deal with next.

Before we go further, it would be expedient to define the task somewhat more precisely. 

If we are looking for a possible formal counterpart for such a sentence as 

The king of France is bald,

 we have two candidates to consider: (a) 

(ux.Fx) Gx, 

in which Peano’s iota-operator is treated as being syntactically analogous to a quantifier; and (6) 

G (ex.Fx), 

in which Peano’s iota-operator is treated as a device for forming a term. 

If we select (a), when we introduce negation, we have two semantically distinguishable ways of doing so;

 ~ ((ux.Fx) Gx) 

and

 (x.Fx) ~ Gx. 

The second will, and the first will not, *entail* the existence of an x that is uniquely F. 

But if we select the latter, there is only one place (prefixing) for the introduction of negation: 

and in consequence

 ~G(x.Fx) 

will be an ambiguous structure — unless we introduce a disambiguating scope convention): 

on one reduction to primitive notation the existence of a unique F will be *entailed*, on the other it will not. 

Call these respectively the strong and the weak readings.

Now if there were a clear distinction in sense (in English) between, say, 

The king of France is not bald 

and 

It is not the case that the king of France is bald 

— if the former demanded the strong reading and the latter the weak one), it would be reasonable to correlate 

The king of France is bald 

with the formal structure that treats Peano’s iota-operator like a quantifier. 

But this does not seem to be the case; 

Grice sees no such clear semantic distinction. 

So it seems better to associate 

The king of France is bald 

with the formal structure that treats the iota-operator as a term-forming device. 

We are then committed to the structural ambiguity of the sentence 

The king of France is not bald.

The proposed task may now be defined as follows: 

On one reading 

The king of France is not bald 

entails 

the existence of a unique king of France, 

on the other it does not; 

but in fact, without waiting for disambiguation, people understand an utterance of 

The king of France is not bald 

as implying (in some fashion) the unique existence of a king of France. 

This is intelligible if, on one reading — the strong one — the unique existence of a king of France is entailed, on the other (the weak one), though not entailed, it is conversationally implicated.

What needs to be shown, then, is a route by which the weaker reading could come to implicate what it does not entail.

If one looks for some prima-facie plausibility for the idea of regarding the definite description as carrying an implicature of a nonconven-tional and conversational kind, where is one to find it? 

Well, one would have to select, 

first (and the case would have to be argued for), one or another of the different Russellian expansions as being that for which such an expression as 

The king of France is bald (or 

The king of France is not bald) 

is to be regarded as a definitional contraction.

And Grice thinks there will be some case for selecting one particular one, namely, the one that would run 

There is at least one king of France,

 there is not more than one king of France, and 

nothing which is the king of France is not bald.

 It seems to have a particular feature that recommends it and might fit in with some general principle of dis-course; 

namely, there are no conjunctions occurring in it within the scope of quantifiers. 

That is to say, it sets out separately three distinct clauses, and each one of these can be false while both of the others are true.

 Grice thinks this is perhaps appropriate because it may well have some connection with something that Grice is going to mention in a moment, namely, that this particular expansion is constructed in a way that makes it particularly suitable for denial on the part of somebody (an addressee) to whom it might be uttered.

 Grice would be inclined to suggest that we add to the maxims of Manner which Grice originally propounded some maxim which would be, as it should be, vague:

"Frame whatever you say in the form most suitable for any reply that would be regarded as appropriate"; 

or,

 "Facilitate in your form of expression the appropriate reply." 

It is very clear that one of the appropriate replies to something that you have asserted is the denial of what you say.

 If your assertions are complex and conjunctive, and you are asserting a number of things at the same time, it would be natural, on the assumption that any one of them might be chal-lengeable, to set them out separately and so make it easy for anyone who wanted to challenge them to do so. 

So, let us make the assumption that we accept some such a conversational maxim and also agree that denial is a natural and suitable form of response to an assertion. 

Let us also adopt the following abbreviational scheme: "A" is to represent 

There is at least one king of France, "

B" is to represent 

There is at most one king of France, 

"C" is to represent 

Whatever is king of France is bald,

"ABC" is to represent the conjunction of A, B, and C, which we are taking as the favored Russellian expansion of "D," which represents 

The king of France is bald.

Now we may hope to reach the conclusion that the production of this abbreviation (D) would violate our newly introduced conversational maxim of Manner unless one could assume that the utterer thinks he is within his rights, in that the utterer does not consider that a distinct denial of A or of B would be appropriate (that this was a response not to be looked for, in his view). 

As a start, given that some kind of denial has to be thought of as appropriate, as that is a natural response to any form of assertion, we might claim that one who employs the abbreviated form D ought either to be thinking it likely that, if there is to be a denial, it will be a wholesale denial, or else, to be thinking that, though the addressee may be going to reject one conjunct, one particular conjunct is, in some way, singled out as the one that is specially likely to be denied.

 It must, indeed, be the second possibility that is to be seriously considered, as the conjuncts cannot all be denied together consistently. 

If it is false that there exists at least one king of France, it is vacuously true that whatever is king of France is bald (that nothing is both king of France and not bald). 

So that leaves us with the demand to show that, in some way, one particular conjunct is singled out.

 Now this would be the case if it would be reasonable to suppose that the utterer thinks, and expects his addressee to think, that some subconjunction of A and B and C has what Grice might call NON-CONTROVERSIALZiTY or com-mon-ground status and, therefore, is NOT something that is likely to be challenged. 

One way in which this might happen would be if the utterer were to think or assume that it is common knowledge, and that people would regard it as common knowledge, that there is one and only one F. 

But that would be only one way in which it could arise.

For instance, it is quite natural to say to somebody, when we are discussing some concert, 

My aunt's cousin went to that concert, 

when we know perfectly well that the person we are talking to is very likely not even to know that we have an aunt, let alone know that our aunt has a cousin. 

So the supposition must be not that it is common knowledge but rather that it is noncontroversial — in the sense that it is something that we would expect the addressee to take from us — if he does not already know.

That is to say, I do not expect, when I tell someone that 

my aunt's cousin went to a concert, 

to be questioned whether I have an aunt and, if so, whether my aunt has a cousin. 

This is the sort of thing that Grice would expect his addressee to take from Grice, that is, to take Gricd’s word for. 

So, we have now got into the position that we might well be in the clear, as far as concerns representing the existential implication as a conversational implicature, if we could show that, in gen-eral, there should be a reasonable expectation, other things being equal, that, in the favored Russellian expansion of a definite descrip-tion, two of the clauses (in fact, the first and the second) would be matters that would have this common-ground status, and so not be controversial or likely to be open to challenge. 

We might, then, assume that, so far, it looks as if the addressee would be justified in concluding that two of the items must be given common-ground status, the only question is which two. 

Now the third clause (C) is general in form.

 And we can think of a general statement as being either something the establishment of which depends on the complete enumeration of a set of instances, or as something to be an inductive step.

Let us take the first possibility. 

Let us suppose it is enumerativelybased. 

That is, we are to think of 

Nothing that is F is not G 

as to be reached by finding the instances of F and seeing that none of them fail to be G. 

For it to be possible to establish this enumeratively if the whole sentence (D) is true, it must be the case that there exists just one F which is the basis of the enumeration. 

And so we have, in effect, a conjunctive statement that tells us there are a certain number of cases (just one) that would test a certain generalization, and then gives us the generalization.

 It would seem to be very peculiar to imagine that anybody could be in the situation in which he was prepared to speak of C, but not as being common ground, because he would have to be put in the position of saying something like,

 "I can accept that nothing is F but not G, that 'what is F is G' is true, and is also to be established by complete enumeration.

 But what I am uncertain about is whether you are right about whether there are any instances of F, or, if so, how many."

 It is not necessary that it should be impossible for somebody to be in that position, but that it is certainly not to be expected; 

and what is to be generally assumed may depend, not on something being universally the case, but merely on its being ex-pectable.

Again, we can take the other possibility. 

Suppose we take the last clause, C, not as being an enumerative generalization, but as an open one.

 And there may be some cases in which that is how it is to be thought of.

 Even so, it is prima facie not to be expected that you would find somebody in the position of being prepared to concede the generalization but being concerned about whether and how often that generalization is instantiated. 

Again, I am not saying that that is not possible. 

But that would certainly be not the kind of situation that one would think of as being the natural one; 

and the implicature depends on what is to be expected, not on what is universally true.

If this line of argument (or something like it) goes through, we could perhaps explain why it is that somebody who says it is not the case that the present king of France is bald (someone who denies what is expressed by D) would also be implicating, though not explicitly stating, that there is a unique king of France. 

It is as if he is countering a remark which might be made to him, in which the speaker has indicated that he is expecting the challenge to come, if at all, in a particular direction, namely, to C; 

and he just says, "No, that is not so." 

The utterer denies that the present king of France is bald, and so naturally he will be taken to be going along with the expected restriction of comment that is implicitly carried by the presentation of the original statement in the abbreviated form (D), rather than in the full form in which each clause would have been set up for him to object to if he wants to. 

The position that I am outlining might be presented, in summary fashion, as follows:

An utterer S, who utters D (the affirmative form of a sentence having as subject a definite description) might expect an addressee to reflect (or intuit) as follows:

(By the conversational 'tailoring' principle) S has uttered D rather than its Russellian expansion; 

so there is one particular Rus-sellian conjunct that S expects me (if I reject anything) to reject, while accepting the other conjuncts.

That is, all but a particular one of the Russellian conjuncts are thought of by him as likely to possess common-ground status (to be treated as noncontroversial).

The first two conjuncts would, in most natural circumstances, be items which anyone would have to know or accept in order to have a good ground for accepting the third.

So the first two conjuncts are the ones to which the utterer attributes common-ground status."

An utterer  S', who utters D (the negation of D) might expect a hearer H' to reflect (or intuit) as follows:

Speaker S' has uttered the negation of D; 

so he is speaking as if he were responding negatively to S (above), that is, to one who utters D.

So S' is fulfilling the expectations that S would have had about

H, that is, he is accepting the first two conjuncts and rejecting the third."

We may note, before moving on, that for a very large range of cases a different account of the existential implication carried by the negative form of a statement involving descriptions might be available.

Consider utterances of such a sentence as 

The book on the table is not open.

 As there are, obviously, many books on tables in the world, if we are to treat such a sentence as being of the form 

The F is not G 

and as being, on that account, ripe for Russellian expansion, we might do well to treat it as exemplifying the more specific form 

The F' which is d is not G, 

where "@" represents an epithet to be identified in a particular context of utterance ("@" being a sort of quasi-demonstrative). 

Standardly, to identify the reference of "p" for a par-ticular utterance of The book on the table is (not) open, a hearer would proceed via the identification of a particular book as being a good candidate for being the book meant, and would identify the reference of "p" by finding in the candidate a feature, for example, that of being in this room, which could be used to yield a composite epithet ("book on the table in this room"), which would in turn fill the bill of being an epithet which the speaker had in mind as being uniquely satisfied by the book selected as candidate.

 If the hearer fails to find a suitable reference for "@" in relation to the selected candi-date, he would, normally, seek another candidate.

 So determining the reference of "d" would standardly involve determining what feature the speaker might have in mind as being uniquely instantiated by an actual object, and this in turn would standardly involve satisfying oneself that some particular feature actually is uniquely satisfied by a particular actual object (e.g. a particular book). 

So utterances both of 

The book on the table is open 

and of 

The book on the table is not open 

would alike imply (in one way or another) the existence of a particular book on a table.

We might, indeed, if we regard this apparatus as reasonably well set up, try to use it to deal with the difficulty raised in the third main objection mentioned, namely, the difficulty about applying the Rus-sellian expansion to moods other than the indicative. 

First, Grice thinks a distinction is needed. 

Grice thinks that the objection, as Grice presented it, is put in a bad form.

 Grice thinks it is important to notice a distinction between what I might call causing something to be the case and what I might call ensuring that it is the case. 

If I tell somebody to cause it to be the case that a particular person has somewhere to live and enough to live on, it looks certainly as if I am thinking that he has to operate in order to promote both clauses; 

I mean, that he will have to find him somewhere to live and give him enough to live on. 

And it seems possible that he could hardly claim to have caused him to have somewhere to live and to have enough to live on unless he had done both of these things. 

But, if I merely tell somebody to ensure that the person has somewhere to live and enough to live on, I think he could also, afterward, claim that he had ensured this, even though, in fact, when he got onto the scene, he found that the man already had somewhere to live and that all he needed was something to live on.

All he has to do, so to speak, is to bring the state of affairs up to completion, if that is required. 

What exactly one is entitled to say if one finds not only that the man has somewhere to live, but also hasenough to live on, after one has been told to ensure that he has both of these things, perhaps is not quite so clear.

So to the question of the imperatives: 

part of the paradoxical character of the suggested Russellian account comes from the fact that I began with 

Make it the case that, 

which suggests that if I were to say 

Give these flowers to your wife,

 the expansion of that imperative must begin 

Make it the case that.. 

And then, when you put in the full Russellian expansion, it looks as if I am instructing you to make three changes, one corresponding to each clause.

 But if I put in 

Ensure that ….. 

instead, there would not be this implication. 

All that would be required is that you should bring the thing to completion, so to speak, insofar as there is a gap. 

And so it may well turn out, also (indeed, in some cases it turns out as a matter of logical fact or something like it), that you cannot do anything about some of the clauses.

You cannot now make it the case that you are now married to one and only one person. 

Either you are or you are not.

 That is outside your control; 

so, in many cases, the only clause left with respect to which you can act is the one covered by 

Make it the case that she has the flowers.

 But, of course, there will be some cases where this particular provision would not work.

 If I tell somebody who is NOT presently married and, as far as I know, has no immediate prospect of getting married, 

See that your next wife looks after you properly, 

Grice do not necessarily think that he is going to get married. 

Nor, Grice thinks, am I instructing him to get married. 

It would be possible, presumably, on a Russellian account, for him to take my instruction as telling him 

to select a wife, first, and 

then, second, to make sure that she looks after him properly. 

So we would need something to ensure that it was not taken this way.

At this point, if one supposed that it is being taken as an assumption by me, as common ground between us, as not to be questioned, that he will at some time or other have another wife (and the point is that when the time comes she should look after him properly), he will not take the imperative force, so to speak, as attaching to the selection of a particular wife. 

So, if it is a conversational implicatum that he will at some time or other have a wife, this will be excerpted from the instructions. 

Grice is inclined to think that this particular dodge works reasonably well for the range of cases considered; 

but I am not wholly happy about it, as it stands, because this general phenomenon of presupposition (or cases that look like presupposi-tion) is one that occurs in a large number of places. 

In recent years, linguists have made it increasingly difficult for philosophers to continue to keep their eyes glued to a handful of stock examples of (al-leged) presupposition, such as the king of France's baldness and the inquiry whether you have left off beating your wife.

There is, in fact, an enormous range of cases in which the questions about presuppositions arise, not least in connection with psychological verbs. 

One can distinguish, perhaps, a number of such cases in connection with psychological verbs. 

Let us take, first, think. 

If I say that somebody thinks (or believes) that such-and-such, there is no indication that what he thinks or believes is true. 

Supposing, however, I take the verb discover, and I say 

Somebody discovered that the roof was leaking. 

Here, it is not logically possible to discover that one's roof is leaking unless one's roof is leaking. 

On the other hand, I do not think (though, perhaps, this is doubtful) that so-and-so did not discover P also implies that P is true.

 I think I can say that some explorer went off to someplace expecting to discover that the natives were very interesting in certain respects, but he did not discover that because they were not. 

So here we have a case where there is a logical implication on the part of the affirmative, but not on the part of its denial. 

That looks like a case of entailment.

Then there is a third case, which perhaps is exemplified by the word know, in which to say that somebody did know that so-and-so was the case and to say that he did not know that so-and-so was the case both imply that it was the case. 

This is a specimen, I think, of the kind of verb that has been called factive. 

There is a distinction between this and a fourth case, because, though both the affirmation and denial of statements about particular people knowing that P carries with it a commitment to P, you can weaken the verb in such a way that this implication is lost. 

He knew that P and He didn't know that P both carry this implication, but He thought he knew that P does not. 

When I say He thought he knew that P, I am not committing myself to its being the case that P, but there are some verbs in which even the weakened forms also seem to carry this implication, partic-ularly, perhaps, a verb like regret (i.e., 

He thought he regretted his father's death, but it afterwards turned out that he didn't, as far as it makes sense, would, I think, still imply the committal to his father's death).

 I am not sure about the last distinction, and I think perhaps it does not matter very much. 

These are cases where there is some kind of a commitment on the part of a normal speaker, by using both the affirmative and the negative forms, to some common element's being true; 

and I do not see that it is going to be particularly easy to represent the implication in the case of regret as being one of a conversational kind. 

It does not look as attractive as the Russellian case. 

So, I would be interested in having recourse to a conventional device which would be a substitute in standard cases for an original conversational implicature. 

To this end I deploy a revised and slightly more complicated version of the square brackets device which was already introduced.

The new version may be used to reinterpret the original de-vice. 

The revised rules would read:

If expression A is of the denominated type T, then A[BIC is rewritten as BABC.

If expression A is not of type T, or is null, then A[BIC is rewritten as ABC.

In rewriting, nested brackets are eliminated, seriatim, from exterior to interior.

If no connective directly precedes a closing (right-hand) square bracket, "&" is supplied in rewriting, where needed, to preserve syntactical admissibility.

Any opening (left-hand) parentheses introduced in rewriting are closed terminally.

In preposing an expression containing a bound variable, the variable is changed.

Using the revised version, ~ ([PS]Q) will be rewritten as PS ~(PSQ), which is equivalent to PJ ~Q (the rewrite of ~([P]Q) on the original version). 

We could use the revised version to handle the alleged existential presuppositions of some and every, every F is G could be represented as ~((3x)(Fx &]~Gx), with the rewrite (y) (Fy 8~(3x)(Fx & ~ Gx)). 

We may also use the revised device in the formal representation of such a factive verb as regret.

Accordingly, 

x regrets (e.g., that Father is ill) 

is defined as:

[x knows* d &] x is anti x knows* d is defined as x thinks [ф]

So, x regrets d emerges as

[x thinks [Ф] & [x is anti $]

So, 

x does not regret d 

would be expressible as:

 ~([x thinks [ф] & ] x is anti $)

Replacing exterior square brackets, we get

x thinks [Ф] & ~ (x thinks [] 8 x is anti ф)

Replacing remaining square brackets, we get

ф 8 x thinks $ 8ф 8 ~ (x thinks $ 8 x is anti b)

Eliminating redundant occurrence of b, we get

ф 8 x thinks $ 8 ~ (x thinks $ 8 x is anti $) which is equivalent to

8 x thinks 8 ~ (x is anti )

We may, finally, consider the employment of the square brackets device to handle the possible difficulties for the Russellian account connected with the appearance of definite descriptions in sentences couched in a mood other than the indicative:

Arrest the intruder

could be thought of as representable (using "!" as an imperative operator) as

! ([(3,x) (x is an intruder) &] (Vy) (y is an intruder > you will arrest y))

Provided "!" is treated as belonging to the denominated type T,

(2) will be rewritten (on the original version of the square bracket device) as

a). (3,x) (x is an intruder) & ! (Vy) (y is an intruder > you will arrest y) and (on the revised version) as

b). (3,z) (z is an intruder) & ! ((3,x) (x is an intruder & (Vy) (y is an intruder > you will arrest y))

Since the first clause of 3(b) sTATes that there is just one intruder, the imperatival clause cannot be taken as enjoining that the addressee see to it that there be just one intruder.

In conclusion, Grice briefly summarizes the course of this essay, primarily in order to distinguish what I have been attempting to do from what I have not. I have endeavored to outline, without aligning myself with it, an exposition of the thesis that the existential presuppositions seemingly carried by definite descriptions can be represented within a Russellian semantics, with the aid of a standard attachment of conversational implicature; I paid attention both to the possibility that such implicata are cancelable and detachable and also to the availability of more than one method of deriving them from the operation of conversational principles. 

Promising though such an account may seem, I have suggested that it may run into trouble when it is observed that the range of cases of presupposition extends farbeyond the most notorious examples, and that perhaps not the whole of this range would prove amenable to the envisaged mode of treat-ment. 

At this point I took up the idea of a minimal strengthening of a Russellian pattern of analysis by the addition of a purely syntactical scope device, which could at the same time be regarded as a conventional regimentation of a particular kind of nonconventional impli-cature. 

I have not, in this essay, given any consideration at all to what might well turn out to be the best treatment of definite descriptions, namely to the idea that they are, in the first instance at least, to be regarded as being, semantically, a special subclass of referential expressions. 

H. P. Grice.