Grice finds it difficult to convey to Oxford just how happy Grice is, and how honoured Grice feels, in being invited to give these lectures.

Grice think of Oxford, his home, Town and Gown, as his spiritual and intellectual parents.

Grice is fashioned at Oxford.

And Grice finds it a moving experience to be, within these splendid and none too ancient walls, engaged in his occupation of rendering what is clear obscure.

Grice is, at the same time, delighted that the Old World should have called me in, or rather recalled me, to redress, for once, a balance.

Grice is finally, greatly heartened by my consciousness of the fact that that great English philosopher, under whose aegis I am now speaking, has extended to him his Lectureship as a gracious consolation for a record threefold denied to him of his Prize.

Grice prays that his present offerings may find greater favour in his sight than did those of long ago.

Oxford

The philosophical clarification of the concept of reason, or perhaps of the family of concepts which shelter under that title, is of interest to Grice, and to others, for more than one reason. 

The nature of reason is an interesting and important philosophical question in itself: 

reason is an important member of the class of ideas with which, as philosophers, we should be concerned. 

But, beyond that foundation of interest, there is the fact that more than one philosopher has held the view that vitally important philosophical consequences can be reached by derivation from the idea of a rational being. 

Aristotle, for example, thought that he could reach a characterization of the end for man via the following steps: 

the end for man is the fulfilment of man's function (ergon); 

the function of man is the optimal exercise of that capacity which distinguishes him from other kinds of creature; 

that capacity is reason or ration-ality;

 the optimal exercise of rationality is the contemplation of the truths of metaphysics; 

so that is the (primary) end for man.

And Kant considered that among the important dividends which could be derived from the idea of a rational being was the moral necessity of adherence to the Categorical Imperative.

Now Grice does not know whether or not any such grand conclusions can be derived from the concept of a rational being, though Grice must confess that Grice has a sneaking hope that they can, and a nagging desire to try to find out. 

Part of my trouble (which is not only mine) is the difficulty of discovering the rules of the game, of understanding what sort of procedure is to be counted as a derivation; 

another part of Grice’s trouble is being hopelessly unclear about the character of the starting point, about what the concept of a rational being is to be taken to be. 

So Grice’s primary role here will be that ofthe under-labourer, to engage in one or two enquiries which might help towards a clarification of the notion of reason or rationality; though Grice hopes that you will forgive him if, just occa-sionally, I make a speculative sally in the direction of the foothills of the mountain which Aristotle and Kant thought that they could climb, indeed had climbed.

Grice shall devote himself to the idea that if, as it seems not unreasonable to suppose, reason is, of its nature, the faculty which is manifested in reasoning, then it would be a good idea to investigate what reasoning is. 

next, Grice  pursues a parallel link between reason and reasons, with particular attention to the relation between practical and non-practical reasons:

No less intuitive than the idea of thinking of reason as the faculty which equips us to recognize and operate with reasons is the idea of thinking of it as the faculty which empowers us to engage in reasoning. 

Indeed, if reasoning should be characterizable as the occurrence or production of a chain of inferences, and if such chains consist in (sequentially) arriving at conclusions which are derivable from some initial set of premisses, and for the acceptance of which, therefore, these premisses are, or are thought to be, rea-sons, the connection between the two ideas is not accidental.

Let us, then, take as a first approximation to an account of reasoning the following: reasoning consists in the entertainment (and often acceptance) in thought or in speech of a set of initial ideas (propositions), together with a sequence of ideas each of which is derivable by an acceptable principle of inference from its predecessors in the set. 

Many would be inclined to subscribe to two further remarks about such an account: first, that the principles of inference which govern reasoning are non-empirical in character; and, second, that it is part of the business (possibly, even the prime business) of logical theorists to distinguish the various modes of inference (non-demonstrative as well as demonstrative) which enter into reasoning, and to systematize the principles of each such mode, thereby both explaining and perhaps (as theorists are wont to do) strengthening assent to (or even, in some instances,undermining assent to) the principles of inference which are intuitively found acceptable at a pre-theoretic stage, and so constitute the initial data for the theorist. 

I shall now enquire into the adequacy of this preliminary account.

One correction is plainly called for. 

Not all actual reasoning is good reasoning; some is bad, and some is downright appalling. 

But our preliminary account seems to leave no space for reasoning to go wrong except through the falsity of one or more of its premisses, or (perhaps) through the perverseness of the world in refusing to conform to the conclusion of an impeccable non-demonstrative inference. 

Obviously, our steps from premisses to conclusions are by no means always well conducted; indeed, in the golden age before logic became a (minor) branch of mathematics, logicians (or writers on logic) paid some attention to fallacies (an interest which could do with revival). 

We should not, then, commit ourselves to the idea that steps in actual reasoning are validly made, but only to the idea that they are thought by the reasoner to be validly made, or (perhaps) to the idea that they are either validly made or are thought to be validly made. 

A little further clarification, however, is needed, and there are one or two morals to be drawn. 

Jack says to Jill (whom he does not yet know very well), "Career women always smoke heavily. You smoke heavily, so you must be a career woman." 

Does fill reply, "You evidently accept the principle of inference: 'A's are always B, this is a B: therefore this is an A'; that is not an acceptable principle"?

 If she made this reply, one would suspect irony. Jack has perhaps argued as if he accepted such a prin-ciple, but I should resist the supposition that he actually did accept it, even temporarily or momentarily.

 People who reason badly may sometimes accept bad principles of inference; but normally what they do is better thought of as, in some way or other, misapplying good principles. But in what way?

 Suppose fill had replied, "You have evidently mistakenly supposed your inference to be of the valid form As are always B's, this is an A; so this is a B'; whereas its actual form was rather 'A's are always B's, this is a

B; so this is an A." Would this reply be an improvement on theone previously suggested for her? 

Not much, I think. 

People may indeed sometimes mistakenly suppose that their actual reasonings exemplify a particular valid principle when, in fact, they do not, and may in consequence reason badly; but I do not believe that this is the normal case, especially with reasonings as simple as the one under examination. Perhaps a better reply for fill might be:

"You seem to be confused, no doubt because that crack on, or in, your crown is still bothering you. I do not accuse you of thinking that the particular invalid form exhibited by your actual argument is a valid form, nor of thinking, wrongly, that your argument was in Barbara.

 I suggest, rather, that the difference between your argument form and Barbara did not on this occasion present itself to you, and that the similarity between your form and Barbara accounts for (as it were, causally) your supposition that the individual argument which you produced was a valid argument." The principal points which seem to me to be, so far, emerging are three in number. (1) 

Generally, though not invariably, bad argument does not have its own bad principles of inference, but rather arises from a misapplication of valid principles. (2) 

That our reasonings should exhibit valid principles is not merely something which usually obtains, nor merely something which both usually obtains and is believed by the reasoner to obtain, but is something which we, as reasoners, want to obtain. 

(It may even be something which should obtain, and something which we try to realize, but these contentions have not yet been argued.) (3) A belief or a desire that a particular piece of reasoning does or should conform to valid principles does not entail (though it does not exclude) a belief or a desire that it does or should conform to a particular valid prin-ciple. 

The operation of individual principles may be, and perhaps for a specifiable class of cases should be, quasi-causal. The meaning of "quasi-causal" of course awaits explication.

I hope that, as we go on, some or all of these points will be expanded and corroborated. Before embarking on the next sub-topic, however, I should like to suggest for consideration a hypothesis of a somewhat more general character. The discrimination and systematization of acceptable principles of inference provide for us a model, or rather, perhaps, an infinite set of models, by reference to which we may understand actual reasonings (an ideal construction to which actual reasonings approximate). Suchmodels or ideal constructions are models in three ways: (1) 

they are analytic models, since it is an adequate degree of approximation to them which confers on certain sequences of thought the title of Reasoning; (2) they are explanatory models, in that they provide (or play a central part in providing) accounts of how actual reasoning proceeds, and why it so proceeds; (3) 

they are normative models, in that they provide patterns of the ways in which actual reasoning should (ought to) proceed. If one were in a fanciful mood, one might be tempted to say that inferential validity is at once the formal cause, the efficient cause, and the final cause of reasoning.

My suspicion is that not only do these models possess this triple aspect, but it is no accident that they do; that there is a conceptual province for which these three aspects go together, and that this conceptual province is specially connected with living creatures.

Unless I am quite mistaken, a detailed and systematic development of this conjecture would be profoundly valuable.

A further batch of questions arises in connection with what 1 shall call 'incomplete reasoning, or 'not fully explicit' reasoning (though these labels may be mildly question-begging).

 Subject to correction from a maturer reflection, we may have on our hands two kinds of case: (a) that in which the incomplete reasoning may be converted into reasoning which conforms to canonical standards of respectability by the addition of further premisses which the reasoner has in mind, either (i) explicitly or (ii) subliminally; and (b) that in which there are no such further premisses which the reasoner has in mind, but the reasoner thinks that such premisses exist even though he does not know what they are. It may turn out that further refinement of this dichotomy is required. 

The rough distinction just drawn embodies two possible answers to the main problem raised by incomplete reasoning, namely, how canonical inference patterns are to be deployed to meet the obvious fact that most actual reasonings do not overtly conform to them, but are (for example) enthymematic.

Let us begin with a very simple example of the kind of argument which it is traditional to classify as an enthymeme, WhenJack sustains his head injury, Jill, putting lack's concerns before her own, says (or thinks), "He is (you are) an Englishman, so he (you) will be brave." 

There seem to be three ways of treating this example. (1) In line with the tradition of syllogistic logic, we might suppose that fill's reasoning involves, as a "suppressed pre-miss", the proposition that Englishmen are always brave, and that Jill has this proposition in mind, perhaps explicitly (if she is talk-ing) or subliminally (if she is thinking). 

Her full argument, which includes the suppressed premiss, is canonically valid; but only a truncated version of her actual reasoning is expressed (in speech or in thought). There are, of course, difficulties connected with the idea of subliminal thinking; but these are very general difficulties and so perhaps should not be pressed in this particular case. 

Aside from them, there is a special difficulty; someone who knows what Jill has said or thought might be inclined to comment, "That does not follow;" and he would be disposed towards this comment not because of the banal fact that the expression of fill's argument is not canonical in form, but because he regards it as being false that Englishmen are always brave.

 But on the present treatment this comment would be inappropriate, since the proposition in question is a premiss of her argument (her full argument), and the falsity of a premiss plainly does not entail that an inferentially licensed conclusion does not follow from its premisses. (2) We might regard "Englishmen are always brave" as expressing not a premiss of Jill's argument but (perhaps in a slightly misleading way) the specific principle or rule of inference of that argument. 

To be a little more precise, we can suppose that we have, for "informal' argu-ment, what might be called inference-schemata; expressed with a ruthless contempt for use and mention which will be characteristic of these lectures, one of them will run: "To infer from Fa to Ga if whatever satisfies 'F' also satisfies 'G'"; on the (contingent) assumption that whatever satisfies 'Englishman' also satisfies 'brave, we can derive the specific rule "To infer from a is an Englishman to a is (will be) brave." This option, if adopted, would allow (for informal argument) contingent (or contingently based) inference rules, though not necessarily contingently based inference-schemata.

(3) Possibly the most attractive idea is to suppose that we should consider ourselves faced not just with one argument or piece ofactual (fill's reasoning) and the other of which is non-actual or ideal (a reconstruction of fill's argument incorporating as a premiss the proposition which we are taking her to have had non-explicitly in mind): the former will be informal, the latter formal (and often canonical). 

Jill's actual argument will be (informally) valid just in case there is a legitimate reconstruction of it which is formally valid and which supplements the informal argument with premisses which are true (as well as being propositions which, in some sense, fill has in mind). 

This suggestion preserves the idea that the 'unexpressed' proposition is a suppressed premiss, and also the idea that inferential rules are non-contingent (since no special rules other than those of formal inference are invoked for informal inference); this suggestion would, I think, always be operable if the previous suggestion were operable (and vice versa).

I shall proceed as if I accepted this third option.

Another possible source of trouble is now visible. 

We have so far been assuming that the suppressed premiss in fill's argument is "Englishmen are always brave". But why should we make this assumption? The suppressed premiss could be "Englishmen are brave", "Englishmen are normally brave", "Englishmen are usually brave", "Englishmen are likely to be brave", and maybe there are further possibilities.

 Some or all of these other candidates, if chosen, would make the reconstructed argument non-demonstrative in char-acter, but apart from a possibly false premiss and false conclusion it could remain a respectable member of its kind (whatever that kind might be). These thoughts prompt the reflection that, in the contretemps envisaged in my previous section, Jack might have a killing answer to Jill's last reply: he might say, "You are evidently assuming that my argument from the premisses Career women always smoke heavily and you smoke heavily to the conclusion you are a career woman was a deductive argument. If your logical education had been a little less narrow, it would have occurred to you that, taken as a non-deductive argument, there is nothing wrong with it." 

But how is one to resolve the kind of indeterminacy which is now appearing, and making difficult the selection of a reconstructed argument? By asking the arguer? By invoking a Principle of Charity, to attribute to the arguer that reconstruction which is logically most satisfactory, or least unsatisfactory?

 If we choose the latter, what is the foundation of such a principle?Troubles of this kind become even more acute when we move to the consideration of less trivial examples; to examples, that is, in which the gap between explicit premisses and conclusion is, intuitively, far greater. Let me offer an extreme specimen. 

When 1 was an Oxford scholar at Corpus, there was a contemporary of mine whose name I cannot remember; it was the name of some English county, so let us call him 'Shropshire.

 His career at Oxford did not last very long; an unsurprising fact-given that, at an early philosophy tutorial, he claimed that the immortality of the soul is proved by the fact that, if you cut off a chicken's head, the chicken will run round the yard for a quarter of an hour before dropping.

Was that an instance of reasoning? Before you answer, let me produce another real life story for comparison with the one I have just given: about twenty years ago a distinguished logician, whose identity I will conceal under the name 'Botvinnik, published a proof, or the sketch of a proof, of an (alleged) theorem; his 'proof' was six pages long. Two Harvard graduate students (one now himself a distinguished logician) set themselves to expand Botvinnik's proof. They found the conclusion of it to be indeed a theorem; but their expansion was eighty-four pages in length.

Now, I have an 'expansion' of Shropshire's 'argument: It runs as follows:

If the soul is not dependent on the body, it is immortal.

If the soul is dependent on the body, it is dependent on that part of the body in which it is located.

If the soul is located in the body, it is located in the head.

If the chicken's soul were located in its head, the chicken's soul would be destroyed if the head were rendered inoperative by removal from the body.

The chicken runs round the yard after head-removal.

It could do this only if animated, and controlled by its soul.

So the chicken's soul is not located in, and not dependent on, the chicken's head.

So the chicken's soul is not dependent on the chicken's body.

So the chicken's soul is immortal.If the chicken's soul is immortal, a fortiori the human soul is immortal.

So the soul is immortal.

The question I now ask myself is this: why is it that I should be quite prepared to believe that the Harvard students ascribed their expansion of Botvinnik's proof, or at least some part of it, to Botvinnik (as what he had in mind), whereas I have no inclination at all to ascribe any part of my expansion to Shropshire?

Considerations which at once strike me as being likely to be relevant are:

that Botvinnik's proof without doubt contained more steps than Shropshire's claim;

that the expansion of Botvinnik's proof probably imported, as extra premisses, only propositions which are true, and indeed certain; whereas my expansion imports premisses which are false or dubious;

that Botvinnik was highly intelligent and an accomplished logician; whereas Shropshire was neither very intelligent nor very accomplished as a philosopher.

No doubt these considerations are relevant, though one wonders whether one would be much readier to accord Shropshire's production the title of"reasoning if it had contained some further striking deductions, such as that since the soul is immortal moral principles have absolute validity; and one might also ask whether the effect of (3) does not nullify that of (2), since, if Shropshire was stupid, why should not one ascribe to him a reconstructed argument containing plainly unacceptable premisses? But, mainly, I would like some further light on the following question: if such considerations as those which I have just mentioned are relevant, why are they relevant?

I should say a word about avowals. 

The following contention might be advanced. If you want to know whether someone R, who has produced what may be an incomplete piece of reasoning, has a particular completion in mind, the direct way to find out is to ask him. That would settle the matter. If, however, you are unable to ask him, then indirect methods will have to be used, which may well be indecisive. Indeterminacy springs merely from having to rely on indirect methods. I have two comments to make. First: itis far from clear to what extent avowals do settle the matter. 

Anyone who has taught philosophy is familiar with the situation in which, under pressure to expand an argument they have advanced, students, particularly beginners, make statements which, one is inclined to say, misrepresent their position. This phenomenon is perhaps accounted for by my much more important second point: that avowals in this kind of context generally do not have the character which one might without reflection suppose them to have; they are not so much reportive as constructive. If I ask someone if he thinks that so-and-so is a consequence of such-and-such, what 1 shall receive will be primarily a defence of this supposition, not a report on what, historically, he had in mind in making it. 

We are in general much more interested in whether an inferential step is a good one to make than we are in what a particular person had in mind at the actual moment at which he made the step. One might perhaps see an analogy between avowals in this area and the specification of plans. If someone has propounded a plan for achieving a certain objective, and I ask him what he proposes to do in such-and-such a contingency, I expect him to do the best he can to specify for me a way of meeting that contingency, rather than to give a historically correct account of what thoughts he had been entertaining. This feature of what I might call inferential avowals is one for which we shall have to account.

Let us take stock. 

The thesis which we proposed for examination has needed emendation twice, once in the face of the possibility of bad reasoning, and once to allow for informal and incomplete reasoning. The reformulation needed to accommodate the latter is proving difficult to reach. Let us take s and s' to be sequences consisting of a set of premisses and a conclusion (or, perhaps it would be better to say, a set of propositions and a further proposition), or a sequence (sorites) of such sequences.

(This is not fully accurate, but will serve.) Let us suppose that x has produced s (in speech or in thought). Let "formally cogent" mean "having true premisses, and being such that steps from premisses to conclusions are formally valid".

(1) We cannot define "s is a piece of reasoning by x" as "x thinks s to be formally cogent", because if s is an incomplete piece of reasonings is not, and could not reasonably be thought by x to be, formally cogent.

We cannot define "s is a piece of reasoning by x" as "(Bs) (s' is an expansion of s and s' is formally cogent)" because (a) it does not get in the idea that x thinks s' formally cogent and (b) it would exclude bad reasoning.

We cannot define "s is a piece of reasoning by x" as "x thinks that (As) (s' is an expansion of s and s' is formally cogent)", for this is too weak, and would allow as reasoning any case in which x believed (for whatever reason, or lack of reason) that an informal sequence had some formally cogent expansion or other.

(Compare perhaps Shropshire.)

(4) We cannot define "s is a piece of reasoning by x" as "(As') (s' is an expansion of s and x thinks s' to be formally cogent)* because of the indeterminacy which bedevils the search for such expansions; there may not be a definitely identifiable expansion which meets this condition.

So what are we to suggest on behalf of the provisional thesis?

I have so far been considering difficulties which may arise from the attempt to find, for all cases of actual reasoning, reconstructions of sequences of utterances or explicit thoughts which the reasoner might plausibly be supposed to think of as conforming to some set of canonical patterns of inference. I turn now to a different class of examples, with regard to which the problem is not that it is difficult to know how to connect them with canonical patterns, but rather that it is only too easy to make the connection.

Like some children (not many), they are too well behaved for their own good.

Suppose someone says to me, "John has arrived," and I reply,

"I conclude from that that John has arrived." Or he says, "John has arrived and Mary has also arrived," and I reply, "I conclude that Mary has arrived." Or he says,

"My wife is at home," and I

reply. "I reason from that that someone is at home." Is there not something very strange about the presence in my replies of the verbs

"conclude" and "reason"? It is true, of course, that if instead of my first reply 1 had said, "So John has arrived, has he?" the strangeness would have been removed; but here the word "so" serves not toindicate that an inference is being made, but rather as part of an idiomatic way of expressing surprise (one might have said, "Well, fancy that!"). 

Now having spent a sizeable part of my working life exploiting it, 1 am not unaware of the distinction between a statement's being false, and its being true but misleading or inappropriate or pointless, and on that account a statement which it would be improper, in one way or another, to make. But I don't find myself lured by the idea of using that distinction here.

Suppose, again, that I were to break off the chapter at this point, and switch suddenly to this argument: "I have two hands (here is one hand and here is another). If had three more hands, I would have five. If I were to have double that number I would have ten, and if four of them were removed six would remain. So I would have four more hands than 1 have now." Is one happy to describe this performance as reasoning? 

There is, however, little doubt that I have produced a canonically acceptable chain of statements.

Or suppose that, instead of writing in my customary free and casy style, I had framed my remarks (or at least the argumentative portions of my remarks) as a verbal realization, so to speak, of sequences of steps in strict conformity with the rules of a natural deduction system of first order predicate logic; 1 give, that is to say, an updated analogue of a medieval disputation. Would those brave souls who continued to read be likely to think of my performance as the production of reasoning, or would they rather think of it as a crazy formalization of reasoning conducted at some previous time?

The points suggested by this stream of rhetorical questions may be summarized as follows:

(1) Whether the samples presented fail to achieve the title of

"reasoning", or whether they achieve it by the skin of their teeth, perhaps does not very greatly matter; for whichever way it is, they seem to offend against something (different things in different cases, perhaps) very central to our conception of reasoning.

(2) Mechanical applications of ground rules of inference, either singly or in concatenation, are reluctantly (if at all) called reasoning.

Such applications may perhaps legitimately enter into (form individual steps in) authentic reasonings, but they are not themselves reasonings, nor is a string of them.

There is a demand that a reasoner should be, to a greater or lesser degree, the author of his reasonings, Parroted sequences are not reasonings when parroted, though the very same sequences might be reasoning if not parroted.

Some examples are deficient because they are aimless or point-less. Reasoning is characteristically addressed to problems: small prob-lems, large problems, problems within problems, clear problems, hazy problems, practical problems, intellectual problems; but problems.

A mere flow of ideas minimally qualifies as reasoning, even if it happens to be logically respectable. But if it is directed, or even monitored (with intervention should it go astray, not only into fallacy or mistake, but also into such things as irrelevance), that is another matter.

Finicky over-elaboration of intervening steps is frowned upon, and in extreme cases runs the risk of forfeiting the title of reasoning.

In speech such over-elaboration would offend against conversational maxims, against (presumably) some suitably formulated maxim of Quantity, In thought, it will be branded as pedantry or neurotic caution. At first sight, perhaps, one would have been inclined to say that greater rather than lesser explicitness is a merit; not that inexplicitness is bad, but that, other things being equal, the more explicitness the better. But now it looks as if proper explicitness is an Aristotelian mean, and it would be good some time to enquire what determines where that mean lies.

The burden of the foregoing observations seems to me to be that the provisional account of reasoning, which has been before us, leaves out something which is crucially important. What it leaves out is the conception of reasoning as an activity, as something with goals and purposes; it leaves out, in short, the connection of reasoning with the will. Moreover, once we avail ourselves of the great family of additional ideas which the importation of this conception would give us, we shall be able to deal with the quandary which 1 laid before you a few minutes ago. 

For we could say (for example) that x reasons (informally) from A to B just in case x thinks that A and intends that, in thinking B, he should be thinking something which would be the conclusion of a formally valid argument the premisses of which are a supplementation of A. This will differ from merely thinking that there exists some formallyvalid supplementation of a transition from A to B, which I felt inclined not to count as reasoning.

I have some hopes that this appeal to the purposiveness of authentic reasoning might be sufficient to dispose of the quandary on which I have directed it; but I am by no means entirely confident that this is the case, and so I offer a second possible method of handling the quandary, one to which I shall return in a later chapter when 1 shall attempt to place it in a larger context.' We have available to us (let us suppose) what I might call a 'hard way' of making inferential moves; we in fact employ this laborious, step-by-step procedure at least when we are in difficulties, when the course is not clear, when we have an awkward audience, and so forth. Inferential judgements, however, are normally desirable undertakings for us only because of their actual or hoped for des-tinations, and are therefore not desirable for their own sake (a respect in which, possibly, they may differ from inferential capacities).

Following the hard way consumes time and energy; these are in limited supply and it would, therefore, be desirable if occasions for employing the hard way were minimized. A substitute for the hard way, the quick way, which is made possible by habituation and intention, is available to us, and the capacity for it (which is sometimes called intelligence, and is known to be variable in degree*) is a desirable quality. The possibility of making a good inferential step (there being one to be made), together with such items as a particular inferer's reputation for inferential ability, may determine whether on a particular occasion we suppose a particular transition to be inferential (and so to be a case of reasoning) or not.

 On this account, it is not essential that there should be a single supplementation of an informal reasoning which is supposed to be what is overtly in the inferer's mind, though quite often there may be special reasons for supposing this to be the case. 

So

[' See also 'Reply to Richards', in Richard Grandy and Richard Warner (eds.)

Philosophical Grounds of Rationality: Intentions, Categories, Ends (Oxford: Oxford University Press, 1986), and The Conception of Value, ed. Judy Baker (Oxford: Oxford University Press, 1991).|

Editor's notes are enclosed in brackets to distinguish them from Grice's own notes.

1' See the discussion of 'flat' and variable rationality in the next section, where Grice discusses variation in degree. Grice interpolated this paragraph here in 1988, The surrounding text was complete in 1987-1Botvinnik is properly credited with a case of reasoning, while Shropshire is not.

What Can't Logic Catch?

I have, so far, been mainly (though not exclusively) considering problems connected with the task of relating the actual reasonings of ordinary people to patterns of complete argument some of which will be valid by canonical standards, and a systematiza-tion of which will be (hopefully) provided by formal logic. It is now time to ask whether there may not be some essential features of reasoning which logic cannot capture, not because logic is not yet sufficiently fully developed (because we need more logical sys-tems), but because those features are not of the right sort to be represented in a logical system at all.

Consider the following example: an unfortunate professor has undertaken, many months previously, to give ten formal lectures to a certain institution, and one month before the starting date the institution asks him for the titles of the individual lectures. He reasons as follows: "Oh, God: It's all a mess; I have piles of mater-ial, but none of it seems worth listening to, and it isn't in shape, and I am in a terrible muddle. And if I give them the ten titles I had in mind I'm not sure that they will fit what (if anything) finally emerges. It's always like this; I take something on long in advance, and then when the time comes I'm reduced to pulp. 

Why do 1 do it? Why don't I learn? I'd like to cancel the whole thing. However, I said I would do it, and if I cancel my name will be mud, so I must go through with it somehow. Anyway, I'm probably not in a fit state to make a proper judgement about the value of my material.

I can see they must have titles, and I can't think of anything better than the ones I had in mind, so I'll give them those and ask for latitude to depart from them if need be. And I'll get four lectures done by the starting date, and that will give me some leeway once they begin."

Such monologues as these are commonplaces in our lives; and not only monologues, since much reasoning is couched in dialogue Now I spoke of the poor professor as reasoning, and some might question the description (I am none too confident myself). But Ifind my reluctance to accept it dwindling when it occurs to me that, while my monologue certainly contained sentences which could not happily be prefixed by the phrase "he reasoned that..", it may well be that items which are not cases of reasoning that are elements in, and crucial elements in, specimens of reasoning; much as pausing to tie one's shoelace, or to look at the view, is not itself strictly an instance of walking, but may nevertheless be a very relevant element in something which is an instance of walking, namely, a walk.

 Should the description be allowed, then many actual reasonings contain such things as questions, the raising of real or imagined objections, suggestions of hypotheses, and even expressions of feeling. Are we confident that, in so far as such items are relevant to the character of reasoning, the logic of the future will be able, or even should try, to devise representation of them?

Or that, if the logician should not try, there is not a region which someone else should investigate systematically? For some of these features of actual reasoning will be extremely relevant to personal attributes which we take to be characteristic of good reasoners, like the ability to construct interesting hypotheses, to keep to the point, to point to analogies, to know when not to ratiocinate, and so on. And even if a formal depiction of these features were access-ible, it is not clear that it would be sufficient as a basis for the attribution of the corresponding qualities; in natural deduction systems there are devices which mark that an assumption is being made, so we can imagine a device which represents hypothesiza-tion; but detection of this mark might leave us in the dark about whether the represented hypothesization was a good one to make.

And to represent order is not to represent orderliness. We are perhaps encountering a question which is an analogue to a notorious problem in ethics, to reach a satisfactory account of the relative priority (one way or the other, or in different aspects both ways) of goodness in a man and the goodness of his acts. Indeed, the kind of features which I am attending to may belong to a department of ethics; that, indeed, was where Aristotle seemed to locate them (or something not too different from them), when he devoted part of Nicomachean Ethics VI to a discussion of intellectual excel-lences. But to relegate them to a kind of ethics would not itself solve problems; and perhaps I might conclude with a brief sketch of the way I feel inclined to regard the issues 1 am now raisingThe picture which I would like to suggest starts with the idea that, when we explore the nature of rationality, or ask what it is to have reason, we really have at least two concepts on our hands, which we should get our hands on. 

One is a 'flat, non-degree-bearing, non-variable concept; it is this concept which applies when it is said that man is a rational being. In respect of this concept no man is more or less rational than any other. The other is a variable, degree-bearing concept, and here some men are definitely more rational than others (though perhaps some philosophers would not agree: Descartes may have thought that not only was God not so niggardly as to leave it to Aristotle to make men rational, but he was not so unjust as to make one man more rational than another). The second concept is intimately connected with the first; and, further, differences in respect of the second concept are dit-ferences in value: it is better to be more rather than less rational.

Concepts linked in this kind of way are not uncommon; a man may know how to drive a car in the sense that he knows how to operate the controls in such a way as to get the car along the road, but he may know how to drive a car in the further and related sense that he sure knows how to drive a car, that he is a good driver. 

Now, if reasoning is thought of as the exercise of reason, and 'reason' (or having reason) is open to both a flat and a variable inter-pretation, we have at once a possible explanation of some of the phenomena which have emerged during this chapter. Hesitation, for example, over the application of the expression reasoning' to the aimless or trivial specimens which I produced for you might be accounted for by the fact that such specimens might be called

'reasoning' if reasoning is thought of as a manifestation of 'flat" reason, but not so called if reasoning is thought of as a manifestation of 'variable' reason; and specimens which enter into reasoning but are not instances of reasoning that might be specimens which are relevant to assessments in respect of variable reason without being instances of flat reason.

It might help to expand the characterization of these two concepts of rationality if I invoke a somewhat hackneyed item in philosophers' analogies, namely, chess, and consider a partial analogy between rationality and chess-capacity. Flat chess-capacity will consist in knowing and being able to apply the laws of chess (including, of course, the moves), much as flat rationality consistsin a capacity to apply rules of inference. 

Variable chess-capacity, expertise at chess, will be learned not from the rule-book, but by practice combined with, perhaps, instruction from other players and such works as Modern Chess Openings and Fine on the endgame. It essentially consists in being good at activities (chess games) conducted in accordance with the laws of chess, much as variable rationality consists in being (more, or less) good at activities conducted in accordance with the laws of reasoning, that is to say, at reasonings of one sort or another: 

For activities of a certain sort to be something one can be (more or less) good at, they must be directed towards goals; these goals may vary, from occasion to occasion, and they may be more or less specific and more or less ultimate; in chess we find such goals as victory, and also more local and intermediate goals, like strengthening one's position by establishing a knight on KBs; in reasoning we have as goals the solution of problems, which may be less specific and more ulti-mate, like understanding Kant's first version of the Categorical Imperative, or more specific and less ultimate, like understanding what a maxim is supposed to be (as a step towards the solution of the former problem). 

Both in chess-playing and in the exercise of rationality we find rough distinctions, in particular cases, between strategies and tactics; and the content of these, and the direction of the sequence of moves which they govern, will be dependent on the identity of the goals which are being pursued.

Not only will (what I might call) policies which are manifested in particular chess games and particular reasonings be more or less specific, and stand in relations of subordination to one another, but also excellences in the two domains may be subordinated to one another. Just as excellence at chess may be diversified, in that one player may be better at openings and less good at the endgame than another, or it may be that nobody handles a pair of bishops like Fischer, so does variable rationality preside (so to speak) over subordinate excellences. 

But now we must begin to take account of obvious differences between our analogues. 

Chess-playing is not about anything, whereas reasoning is always reasoning about some topic or question, which may fall within some special area like mathematics or theology; so not only are there specific rational excellences or qualities with respect to mathematics, which some may have in a greater degree than others, but one maybe good at mathematical reasoning while quite incompetent at theological reasoning.

 However, we seem to be willing to allow the attribution of general excellence, or quality, of reasoning as well as that of specialized quality; and general excellence seems also to be diversified. This licence we allow ourselves may be a consequence of the fact that general rational excellences (or some of them) seem to be special cases of capacities for achieving qualities which are desiderata for any practical undertaking, qualities like simplicity, economy, accuracy, inventiveness, and so on. 

Another derivative of the fact that reasoning is about things is that reasoning may be directed upon itself; one may not only use reasoning to plan a henhouse or a cathedral, but, if the first-level task is sufficiently formidable, one may plan how to plan a cathedral; one may even plan a philosophical methodology, though too few people do. A consequence of the last two observations, that general rational excellence is diversified and that reasoning may be turned on itself (together, perhaps, with some further premiss), is that we may be able to treat flat reason, not merely as something which is necessarily manifested in manifestations of variable reason, but as providing an inferential base for determining the nature of variable reason itself and, also, of its more specific subordinate competences (excellences). We might be able to argue, for instance, along the following lines.

Let us assume that it can be shown that truth for, better, some more general feature which will include truth as a special case, but will also apply to some of the objects of psychological attitudes other than belief, such as desire) is a value (in the sense of desideratum).

Inferential rules, which flat rationality is the capacity to apply, are not arbitrary, in that they pick out transitions of acceptance in which transmission of satisfactoriness (including where appropriate truth) is guaranteed or (in non-deductive cases) to be expected.

Since our actually making such transitions in a particular case is up to us (and subject to our particular needs and circum-stances), inferential rules can be seen as directives (the precise kind of which remains to be determined) observance, or non-violation, of which is a desideratum.

(d) Since reasoning is (and it may be of its essence that it is) sometimes addressed to problems, at least sometimes reasoning has a particular goal, which the reasoner aims at reaching.

Reasonings then are at least sometimes characterizable as successful or not successful.

Some particular qualities, exhibited in reasonings, can be characterized as certain or likely to lead to success (or, perhaps, failure) in pursuit of goals in reasoning, irrespective of the particular nature of those goals.

In so far as no sources of information (no premisses) other than the nature of the flat capacity of reason (and implicata thereof) have been used to reach such a characterization, at least some excellences are determinable as such by "the concept of a rational being"

Now, if such arguments are possible, their conclusions may not be of the grand sort which rationalists have hoped to draw from the idea of a rational being, but at least they would be substantive conclusions; and they might be of value towards the grander enterprises, both as providing possibly usable premisses and because the derivation of them would help to clarify the methodology of those enterprises.

These reflections might be reinforced by an approach from another quarter, namely, a consideration of the ordinary use of the words 'rational' and 'reasonable' (and their complementaries).

One might have expected that, when used as terms of praise, these words refer to very general qualities of reason (in which case one would be right), and indeed to the same quality (in which case one would be wrong). A few examples. It might be unreasonable of me to expect my wife to clean my football boots for me, but it would not be irrational. I may well have bought those boots at a very reasonable price, but it is not very clear how I might have bought them at a very rational price. 

To cheat someone in a business deal (as such) is neither unreasonable nor irrational; it is merely somewhat repulsive; to cheat a man when you knew you mightbe found out, and as a result lose a valuable client, 1 would regard as a better candidate for 'unreasonable' than for 'irrational'; to cheat him when you knew it was quite likely that you would be found out, and when, if you were, you would lose your job at a time when employment was very difficult to obtain, I would call 'irrational:

Yielding to a tempting invitation to go out drinking when I have already decided to spend the evening working on tomorrow's lecture, I would regard as (as such) neither unreasonable nor irrational, though it may be weak, and foolish. To yield to that temptation when I have not yet decided what to do, but know I really ought to get on with that work for tomorrow, might be unreasonable but would not be irrational. 

If I have bunglingly got my firm into a difficulty, and I go and confess the matter to my boss, he might be both reasonable about it and rational about it; he might be reasonable about it in that he was not too hard on me, and rational about it in that he coolly and in a reasoned way told me what was the best course to take.

Now I cannot give you a detailed solution to the problem of distinguishing between "rational" and "reasonable", not only because I do not have the time, but also because I am by no means sure what to say. But I do think that I know two keys to the solution of this problem. The first key is that "reasonable", unlike

"rational", is really a privative term; "unreasonable" is, as some were once wont to say, the "trouser-word" in this particular pair of complementaries. To be reasonable is to be (relatively) free from unreasonableness; and to be very reasonable is to be free from a high degree of unreasonableness which one might (or some might) have been expected to exemplify or display in the circum-stances. 

The second key is provided by Aristotle; in Nicomachean Ethics I, he remarks that both the ratiocinative and the non-ratiocinative (or desiderative) parts of the soul may have reason; the former intrinsically, as the source of rational principles or precepts, the latter extrinsically, as heeding or listening to those principles or precepts. My idea is to link the first of Aristotle's interpretations of "having reason" with the word 'rational', and the second with the word 'reasonable. In application to behaviour, to be rational is to possess (or, on a given occasion, to display) the capacity to reach principles or precepts relating to conduct; to be reasonable is (in general or on a particular occasion) to be freefrom interference, on the part of desire or impulse, in one's following such principles or precepts.

The present relevance of this discussion is that ordinary thought and speech embody the idea of reason as regulation, as something which should control desire or passion. So the phenomenon of incontinence should not be regarded, as it too often has been, as a stumbling block to some otherwise attractive theory of will and practical reason; if reason is regulation, then it must be possible for what is regulated to get out of hand, and this possibility should be provided for, in the theory, from the start.

Furthermore, if we incorporate, in our idea of reason, the idea of it as regulation of a sub-rational nature which may get out of hand, we have enlarged the "concept of a rational being" from which philosophically significant conclusions might be derivable, and so given such derivations a better chance of success. It is curious that both Aristotle and Kant, despite their sophistication with regard to the nature of practical reason, should have slipped up here in their different ways, that both should have succumbed to the fascination of the purely intellectual being. 

Both of them, it seems to me, at crucial moments thought of the rationality, the realization of which must be the supreme end for a rational being, as being the distinctive element in such a being, considered in isolation from other elements necessarily present in, but not necessarily peculiar to, such a being; so for Aristotle the primary end for man becomes approximation to the activity of a pure substance, and for Kant the practical law is something which has to be applicable to a being with a holy will, treated not just as a possibly useful fantasy, but as a being which might actually exist. 

This common ingredient in their practical philosophies seems to me to have been restrictive rather than beneficial. 

H. P. Grice.