In his ‘Method in philosophical philosophy: from the banal to the bizarre,’ Grice presents some of his ideas about how he wants to approach philosophical psychology.

Grice’s hope is, in effect, to sketch a whole system.

This is quite an undertaking, and Grice hopes that you will bear with him if in discharging it he occupies a little more of your time than is becoming in holders of his august office. 

While Grice is sure that you will be able to detect some affinities between his ideas and ideas to be found in philosophy, Grice proposes to leave such comparisons to you. 

Though at certain points Grice has had his eye on  discussions, the main influences on this part of his work have lain in the past, particularly in Aristotle, Hume, and Kant. 

Grice has the feeling that, among them, these philosophers have written a great deal of the story, though perhaps not always in the most legible of hands.

Grice begins by formulating, in outline, a sequence of particular problems which Grice thinks an adequate philosophical psychology must be able to lay to rest.

One concerns the real or apparent circularity with which one is liable to be faced if one attempts to provide an analysis of a central psychological concept by means of an explicit definition.

Suppose that, like some philosophers, we are attracted by the idea of giving a dispositional behaviouristic analysis of such a concept, and that we make a start on the concept of belief. 

As a first shot, we try the following:

x believes that p 

just in case 

x is disposed to act as if p is true.

In response to obvious queries about the meaning of the phrase, 'act as if p is true', we substitute:

x is disposed, whenever x wants (desires) some end E, to act in ways which will realise E given that p is true, rather than in ways which will realise E given that p is false. 

The precise form which such a definition might take is immaterial to Grice’s present purpose, provided that it has two features observable.

First,

that a further psychological concept, that of wanting, has been introduced in the definiens; and 

second,

that to meet another obvious response, the concept of belief itself has to be reintroduced in the definiens.

For the disposition associated with a belief that p surely should be specified, not as a disposition to act in ways which will in realise E given that p is true, but rather as a disposition to act in ways which x *believes* will — given that p is true — realise E. 

One who believes that p may often fail to act in ways that would realise E given that p is true, because he is quite unaware of the fact that such actions would realise E if p is true

Or he may quite often act in ways which would realise E only if p is false — because he mistakenly believes that such ways would realiss E if p is true.

If we turn to the concept of wanting, which the definition introduced into the definiens for belief, we seem to encounter a parallel situation. 

Suppose we start with:

x wants E 

just in case 

x is disposed to act in ways which will realise E, rather than in ways which will realise the negation of E. 

The same kind of objection, as that just raised in the case of belief, seems to compel the introduction of the concept of belief into the definiens of W, giving us:

x is disposed to act in ways which x *believes* will realise E rather than in ways which x believes will realise the negation of E. 

We now meet the further objection that one who wants E can be counted on to act in such ways as these — only provided that there is no E2 which he wants more than E1.

If there is such an E2, in any situation in which & conflicts with & he may be expected to act in ways he thinks will realise the negation of E. 

But the incorporation of any version of this proviso into the definiens for wanting will reintroduce the concept of wanting itself.

The situation, then, seems to be that if, along the envisaged behaviouristic lines, we attempt to provide an explicit definition for such a pair of concepts as those of belief and wanting, whichever member of the pair we start with we are driven into the very small circle of introducing into the definiens the very concept which is being defined, and also into the slightly larger circle of introducing into the definiens the other member of the pair. 

The idea, suggested to Grice by this difficulty, is that we might be well advised, as a first move, to abandon the idea of looking for an explicit definition of a central psychological concept, and look instead for an implicit definition, to be provided by some form of axiomatic treatment — leaving open the possibility that, as a second move, this kind of treatment might be made the foundation for a different sort of explicit definition. 

Such a procedure might well preserve an attractive feature of behaviouristic analyses, that of attempting to explain psychological concepts by relating them to appropriate forms of behaviour, while at the same time freeing us from the logical embarrassments into which such analyses seem to lead us.

We are now, however, faced with a further question.

If we are to think of a certain psychological concept as being implicitly defined by some set of laws, or quasi-laws, in which they figure, are we to attribute to such laws a contingent or a non-contingent status?

A look at some strong candidates for the position of being laws of the kind which we are seeking seems merely to reinforce the question. 

Consider the principle:

‘He who wills the end wills the means.’

— some form of which seems to me to be the idea behind both of the behaviouristic definitions just discussed.

Interpret it as saying that anyone who wills some end, and also believes that some action on his part is an indispensable means to that end, wills the action in question. 

One might be inclined to say that if anyone believes that a certain line of action is indispensable to the attainment of a certain end, and yet refuses to adopt that line of action, that would count decisively against the conceptual legitimacy of saying that it is his will to attain that end. 

Think Machiavelli!

To proceed in this way at least appears to involve treating the principle as a necessary truth.

On the other hand, one might also be inclined to regard the principle as offering a general account of a certain aspect of our psychological processes, as specifying a condition under which, when our will is fixed on a certain object, it will be the case that it is also fixed on a certain other object.

To take this view of the principle seems like regarding it as a psychological *law* — and, so. as contingent. 

Grice’s second problem  is, then, this: 

How — without a blanket rejection of Leibniz’s analytic/ synthetic distinction — are we to account for and, if possible, resolve the ambivalence concerning their status with which we seem to look upon this or that principle involving this or that psychological concept?

Cf. Conversation as rational cooperation.

To set this problem aside for a moment, the approach which Grice is interested in exploring is that of thinking of a central psychological concept as a theoretical concept.

It is a psycho-logical concepts (Italian, psychic) just because it is the primitive concept which belong to a certain kind of psychological theory, without also belonging to any presupposed theory — such as physiological theory.

And a psychological theory is a theory whose function is to provide, in a systematic way, an explanation of behaviour whichs differ from any explanation of behaviour which may be provided by, or may some day be provided by, any presupposed theory — such as physiological theory.

To explicate such a psychological concept is to characterize its role in the theory to which it primarily belongs, to specify, with this or that degree of detail, the laws — or quasi-laws — in which it figures, and the manner in which such a law or quasi-law, is linked to behaviour. 

Now to say this much is not to say anything very new; 

Grice is sure you have heard this sort of thing before. 

It is also not to say very much.

Allthat has so far been given is a sketch of a programme for the philosophical treatment of a psychological concept, not even the beginnings of a philosophical treatment itself.

What is needed is rather more attention to detail than is usually offered by advocates of this approach.

We need to pursue such questions as what the special features of a psychological theory are, why such a theory should be needed, what sort of laws or generalities it should contain, and precisely how such a law may be used to explicate a familiar psychological term. 

Grice shall be addressing hinself to some of these questions in the remainder of this address to you.

But, before any more is said, a further problem looms. 

Grice can almost hear a murmur to the effect that such an approach to philosophical psychology is doomed from the start. 

Do we *not* have privileged access to our own belief and desire? 

And, worse still, may it not be true that at least some of our avowals of our belief and our desire are incorrigible? 

How, then, are such considerations as these to be rendered consistent with an approach which seems to imply that the justifiability of attributing a belief and a desire to people — and maybe animals —, including ourselves, rests on the utility of the theory to which the concepts of belief and desire belong, in providing desiderated explanations of behaviour? 

This is Grice’s third problem.

Grice’s fourth problem, the Selection Problem, is connected with a different and less radical objection to the approach which Grice has just begun to sketch. 

Surely, it may be said, it cannot be right to suppose that a conjunction of all the laws of a psychological theory in which a psychological concept figures is what should be used to explicate that concept. 

Even if we are in a position to use any such law — which is not certain — we are not, and indeed never shall be, in a position to use all of them.

Moreover, this or that law may not be suitable. 

For all we know, some modification of one or other of the following laws' might be a psychological law, possibly an underived law:

Optimism Law: 

The more one wants p the more likely one is to believe p. 

Optimism/Pessimism Law: 

Given condition C1, the more one wants d the more likely one is to believe

d.

Given condition C2, the more one wants d, the less likely one is to believe d.

These do not seem the right kind of law to be used to explicate wanting or believing.

We need some selective principle.

What is this selective problem?

First, a preliminary observation.

If we are seeking to explicate a psychological concept by relating it to psychological theory, the theory which we invoke should be one which can be regarded as underlying our ordinary speech and thought about this or that psychological matter, and as such will have to be a part of folk-science.

There is no need to suppose that its structure would be of a sort which is appropriate for a professional theory, nor even to suppose that it would be a correct or acceptable theory; though Grice would hope that there would be a way of showing that, at least, some central parts of it would be interpretable within an acceptable professional theory — in such a way as to come out true in that theory, and that, indeed, this demand might constitute a constraint on the construction of a professional theory. 

This is perhaps a latter-day version of a defence of common sense in this area.

Let us unrealistically suppose, for the purposes of schematic discussion, that we have a psychological theory which, when set out in formal dress, contains a body of underived laws, the formulation of which incorporates two primitive predicate *constants*: J and V.

These two *constants* we want to correspond, respectively, to two psychological terms

"judging" — from Latin iudicatio — and "willing — from Latin vuolere. 

— which in their turn will serve as a regimented base for the explication of such familiar notions as believing and wanting. 

Let us think of I and V as correlated with, or ranging over, "instantiables" leaving open the question whether these instantiables are sets or properties, or both, or neither. 

Let us abbreviate the conjunction of these laws by "L", and let us ignore for the moment the obvious fact that any adequate theory along these lines would need to distinguish specific sub-instantiables under J and V, corresponding to the diversity of specific contents which judging and willing may take as modifications as "intentional objects.”

How are we to envisage the terms "judge" and "will" as being introduced? 

Two closely related alternative ways suggest themselves. 

First, what Grice shall call the way of Ramseified naming: 

“There is just one J, and just one V 

— such that L, 

and 

Let I be called "judging" and V be called “willing.”

On this alternative, the uniqueness claim is essential, since "judging" and "willing" are being assigned as names for this or that particular instantiable.

An alternative, which Grice may call the way of Ramsified *definition*, can dispense with the uniqueness claim. 

It will run: 

x judges just in case there is a J and there is a V such that L, and x instantiates J.

x wills just in case there is a J and there is a V such that L, and x instantiates V.

We now have available a possible explanation of the ambivalence we may feel with regard to the Leibnizian status of a psychological principle, which was the subject of a problem.

The difference between the two alternative procedures is relatively subtle, and if we do not distinguish them we may find ourselves under the pull of both. 

The first alternative renders a psychological principle contingent.

The second renders the same principle non-contingent.

Let me illustrate by using an even simpler and even less realistic example. 

Suppose that a psychological law tells us that 

anyone, in state P, hollers.

And that we seek to use this law to introduce the term "pain" — Latin: paena.

On the first alternative, we have: 

There is just one P, such that anyone, in state P, hollers.

Let us call P "pain". 

Since on this alternative we are naming the state P "pain", if the utilised law is contingent, so will be the principle 

Anyone who is in pain hollers.

If we use the alternative, we shall introduce "pain" as follows: 

x is in pain just in case 

there is a P such that:

anyone in state P hollers and 

x is in state P. 

On this alternative, it will be non-contingently true that anyone who is in pain hollers. 

For to be in pain *is* to be in some state which involves hollering.

While this suggestion may account for the ambivalence noted, it does not of course resolve it.

To resolve it, one would have to find a reason for preferring one of the alternatives to the other.

In a somewhat devious pursuit of such a reason, let us enquire further about the character of the postulated psychological instantiable.

Is it, or can it be, identifiable with a physical instantiable?

One possible position of a sort which has been found attractive by some slightly impetuous physicalists would be to adopt the first alternative — that of Ramsified naming — and to combine it with the thesis that 

the J-instantiable 

is to be identified with one physiological property and 

the V-instantiable 

with another such property. 

This position at least appears to have the feature, to which some might strongly object, of propounding a seemingly philosophical thesis which is at the mercy of a possible development in physiology.

One might add that the prospect that such a development would be favourable to the thesis does not seem to be all that bright. 

The adaptiveness of an organism may well be such as to make it very much in the cards that two soecimens of the same species may, under different environmental pressures, develop different sub-systems, even different sub-systems at different times, as the physiological underlay of the same set of a psychological instantiable, and that a given physiological property may be the correlate in one sub-system of a particular psychological instantiable, while in another it is correlated with a different psychological instantiable — or with none at all.

Such a possibility would mean that a physiological property which was, for a soecimen at a particular time, correlated with a particular psychological instantiable would be neither a necessary nor a sufficient condition for that instantiable.

It would at best be sufficient, and perhaps also necessary, for the instantiable within the particular kind of sub-system prevailing in the specimen at the time. 

An unqualified identification of the physical property and the psychological instantiable would now be excluded.

Faced with such a difficulty, a physicalist might resort to various manoeuvres. 

He might seek to identify a particular psychological instantiable with a disjunctive property, each disjoined constituent of which is a property consistingin having a certain physiological property in a certain kind of sub-system.

Or he might relativize the notion of identity, for the purpose of psycho-physical identification, to types of sub-systems.

Or he might abandon the pursuit of identification at the level of instantiables, and restrict himself to claiming an identities between individual psychological events or states of affairs 

e.g. 

Smith’s believing at t that p

and physiological events or states of affairs — e.g 

Smith’s brain being in such and such a state at t.

But none of these shifts has the intuitive appeal of their simplistic forebear. 

Moreover, each will be open to variants of a familiar kind of objection. 

For example, Smith’s judging at noon that they were out to get him might well be a case of judging something to be true on insufficient evidence.

But, to use Berkeley's phrase, it 'sounds rather harsh' to say that it is Smith’s  brain being in such and such a state at noon which is ultimately a case of judging something to be true on insufficient evidence.

For Grice’s part, Grice would hope that a much-needed general theory of categories would protect him against any thesis which would require him either to license such a locution as the last — or to resort to a never-ending stream of cries of 'Opaque context!' in order to block such a locution.

If the prime purpose of the notion of identity is, as Grice believes it to be, to license a predicate transfer, one begins to look a little silly if one first champions a particular kind of identification, and then constantly jibs at the predicate transfer which it seems to allow.

Such considerations as these can, Grice thinks, be deployed against the way of Ramsified naming even — when it is *unaccompanied* by a thesis about a psycho-physical identity.

However unlikely it may be that the future course of physiological research will favour any simple correlations between a psychological instantiabl and a physiological property, Grice does not see that he has any firm guarantee that it will not, that it will never be established that some particular kind of Smith’s brain state is associated with, say, Smith’s judging that snow is white. 

If so, if Grice adopts the first alternative with its uniqueness claim, Grice has no firm guarantee against having to identity Grice’s brain's being in some particular state with Grice’s judging that snow is white, and so being landed with the embarrassments Grice has just commented upon. 

Since Grice is inclined to think that he does in fact have such a guarantee although at the moment he cannot lay his hands on it, Grice is inclined to prefer the second alternative to the first.

As a pendant to the discussion just concluded, Grice expresses two prejudices. 

Any psycho-physical identification which is accepted will have to be accepted on the basis of some known, or assumed, psycho-physical co-relation.

it seems to Grice that in this area all that philosophical psychology really requires is the supposition that there *is* such a co-relation.

Whether the co-relation does or do not provide a legitimate foundation for identifications seems to Grice to be more a question in the theory of identity than in the philosophy of mind. 

Or philosophical psychology — since Ryle taught Grice to avoid ‘mental.’

Second, Grice is not greatly enamoured of some of the motivations which prompt the advocacy of a psycho-physical identification.

Grice has in mind a concern to exclude such a ‘queer' or "mysterious' entity, as a ‘soul,’ a a purely psychic event — such as ‘Someone is hearing a noise’ — a purely psychic property, and so forth. 

Grice’s taste is for keeping open house for all sorts and conditions of entities, just so long as, when they come in, they help with the house-work.

Provided that Grice can see an entity  at work, and provided that it is not detected in illicit logical behaviour — within which I do not include a certain degree of indeterminacy, not even of numerical indeterminacy —, Grice does not find such an entity as a psychic entity queer or mysterious at all. 

To fangle an ontological marxism, it works; therefore, it exists —

even though such an entity, perhaps an entity which comes on the recommendation of some form of transcendental argument, may qualify for the specially favoured status of an ens realissimum.

To exclude a honest working entity seems to Grice like metaphysical or ontological snobbery — a reluctance to be seen in the company of any but the best things.

Grice discusses two formal features which Grice thinks it might be desirable to attribute to some at least of this or that law or quasi-law of the theory to be used to explicate a psychological concept.

Aristotle distinguishes between things which are so of necessity and things which are so for the most part, and located in the second category things which are done.

More or less conformably with this position, Grice suggests that a law which determines a psychological concept is a ceteris paribus law — which resembles a probability generalisation in that it is defeasible, but differs from it in that a ceteris paribus law does not assign a weight.

Grice envisages a system which, without inconsistency, may contain a sequence consisting of a law of the form 

A's are Z, 

and a modifying law of the form 

A's which are B are Z' 

where Z' is incompatible with Z — and of the form 

A's which are B and C are Z, 

and so on. 

Analogous sequences can be constructed for a functional law; 

if A, B, and Z are determinables we may have, in the same system:

a thing which is A to degree a are Z to degree f' 

a thing which is A to degree a and B to degree B are Z to a degree which is the value of f (B, f' (a))

If we like, we can prefix "ceteris paribus" to each such generality, to ensure that 

"A's are Z" 

is NOT taken as synonymous with 

Every A is Z.

For any such system a restriction on Modus Ponens is required, since we must not be allowed to infer from 

A's are Z 

and 

x is both A and 

B to 

x is Z 

if A's which are B are Z' - where Z' is incompatible with Z — 

is a law. 

This restriction will be analogous to the Principle of Total Evidence required for a probabilistic system;

A first approximation might run as follows:

 in applying a law of the system to an individual case for the purpose of detachment, one must select the law with the most specific antecedent which is satisfied by the individual case. 

Results of treating a psychological law as ceteris paribus laws which will be attractive to Grice are 

first 

that 

It can no longer be claimed that we do not *know* any psychological law because we do not know all the restrictive conditions, and 

second, 

that 

a psychological theory may be included in a larger theory which modifies it.

Modification does not require emendation.

If a modifying sequence terminates, its constituent law can be converted into a universal law.

Otherwise, not.

If we hold a strong version of Determinism — roughly, that there is an acceptable theory in which every phenomenon is explained — we might expect the Last Trump to herald a Presidential Address in which the Final Theory is presented, entirely free from ceteris paribus laws.

But if we accept only a weaker version, with reversed quantifiers — for every phenomenon there is an acceptable theory in which it is explained — the programme for Judgement Day will lose one leading attraction.

Other items on the agenda, however, may prove sufficiently exciting to compensate for this loss.

Grice suspects that even the latitude given to our prospective psychological folk-theory by allowing it to contain or even consist of ceteris paribus laws will not make it folksy enough. 

There is, or was, in empirical psychology, a generality called "The Yerkes-Dodson Law"; based on experiments designed to test degrees of learning competence in rats set to run mazes under water, after varying periods of initial constraint, it states, in effect, that, with other factors constant, degrees of learning competence are co-related with degrees of emotional stress by a function whose values form a bell-curve.

 I have some reluctance towards calling this statement the expression of a law, on the grounds that, since it does not - and could not, given that quantitative expression of degrees of competence and of stress is not available — specify the function or functions in question, it states that there is a law of a certain sort rather than actually state a law. 

But whether or not the Yerkes-Dodson Law is properly so-called, it is certainly law-allusive, and the feature of being law-allusive is one which Grice would expect to find in a psychological law to be used to explicate a psychological concept.

Grice turns now to the task of outlining, in a constructive way, the kind of method by which, in accordance with his programme, a particular psychological concept, and a linguistic expression for it, might be introduced.

Grice takes into account the need to clarify the routes by which a psychological sub-instantiable comes to be expressed by the combination of a general psychological VERB and a complement which specifies content. 

Grice’s account will be only semi-realistic, since he considers the psychological explanation of only a very rudimentary sample of behaviour, and even with respect to that sample a proper explanation would involve apparatus which Grice omits - for example, apparatus to deal with conceptual representation and with the fact that a motivation towards a particular sort of behaviour may be frustrated. 

But since Grice is attempting only to illustrate a general method, not to make a substantial proposal, these over-simplifications should not matter. 

Grice constructs his account as if I espoused the first alternative: the way of Ramsified naming, rather than the second, since that considerably simplifies exposition.

A transition to the second alternative can, of course, be quite easily effected.

Let us suppose that a squarrel — a specimen something like a squirrel — has some nuts in front of it, and proceeds to gobble the nuts; and that we are interested in the *explanation* of this occurrence. 

Let us call the squarrel "Toby". 

Our ethological observations of Toby, and of other squarrels, tell us that Toby — and other squarrels — often gobble nuts in front of them, and also other things besides nuts, and, indeed, that they are particularly liable to do some gobbling after a relatively prolonged period in which they have done no gobbling at all. 

On the basis of these observations, and perhaps a range of other observations as well, we decide that a certain behaviour of squarrels — including Toby's gobbling nuts in front of him — is a suitable subject for psychological explanation; so we undertake to introduce some theoretical apparatus which can be used to erect an explanatory bridge between Toby's having nuts in front of him, and his gobbling them.

We make the following postulations (if we have not already made them long ago):

That the appropriate explanatory law will refer to three instantiables 

P, 

J, and 

V which (we stipulate) are to be labelled respectively 

“prehend", "join", and "will"

That for any type of creature T 

there is a class N of kinds of thing, 

which is a class of what (we stipulate) shall be called a  "necessity” for T", and that 

N (in N) is a necessity for T just in case N is *vital* for T

that is, just in case any member of T which suffers a sufficiently

prolonged non-intake of N will lose the capacity for all of those operations — including the intake of N —, the capacity for which is constitutive of membership of T 

[in a fuller account this condition would require further explication).

That 

if N is a necessity for T, ceteris paribus a moderately prolonged non-intake of N on the part of a member (x) of T will cause x to instantiate a particular sub-instantiable V; of "will", x's instantiation of which will (ceteris paribus) be quenched by a moderately prolonged intake of N; and 

We stipulate

V; shall be called "will for N" (or, if you like, "willing N").

That 

for any member x (e.g. Toby) of any creature-type T (e.g. squarrels) there is a class f of thing-types (which, we stipulate, is a class of what shall be called"thing-types familiar to x");

 a class R of relations (a class of what shall be called "relations familiar to x"), and 

a class  of action-types — a class of what shall be called "action-types in the repertoire of x"),

 which satisfy the following conditions:

that 

there are two ways (if you like, functions) w, and w, such that, ceteris paribus, if an instance of a thing-type F (familiar to x) is related to x by a relation R (familiar to x), this causes x to instantiate a sub-instantiable of prehending which corresponds in ways w, and we to F and R respectively; 

which sub-instantiable (we stipulate) shall be called "prehending an F as R" le.g.

"prehending nuts as in front"

that 

ceteris paribus if, for some N [e.g. 'squarrel-food'] which is a necessity for x's type, for some A which is in x's repertoire, and for some F and R familiar to x, x has sufficiently frequently, when instantiating will for N, performed A upon instances of F which have been related by R to x, and as a result has ceased to instantiate will for N: 

x instantiates a corresponding [corresponding to N, A, F, R] sub-instantiable of joining; which sub-instantiable shall be called "joining N with A and F and R".

We can now formulate a law in which we no longer need to restrict N, A, F, and R to members of classes connected in certain ways with a particular creature or type of creature, as follows:

Ceteris parious, 

a creature x which wills N, 

prehends an F as R, and 

joins N and A and F and R, 

performs A.

Finally, 

we can introduce "judging" by derivation from

joining"; 

we stipulate that 

if x joins N with A and F and R, we shall speak of x as judging A, upon-F-in-R, for N le.g. judging gobbling, upon nuts (in) in front, for squarrel-food).

AdWe now have the desired explanatory bridge.

Toby has nuts in front of him

Toby is short on squarrel-food (observed or

assumed); so

(iii) Toby wills squarrel-food [by postulate 3, connect-ing will with intake of N;(iv) 

Toby prehends nuts as in front [from (i) by postulate 4(a), if it is assumed that nuts and in front are familiar to Toby);v) 

Toby joins squarrel-food with gobbling and nuts and in front (Toby judges gobbling, on nuts in front, for squarrel-food) [by postulate 4(b), with the aid of prior observation]; so, by the PJV Law, from (iii), (iv), and (v),

Toby gobbles; 

and, since nuts are in front of him gobbles the nuts in front of him.

Grice ends the section by some general remarks about the procedure just sketched.

The strategy is relatively simple.

We invoke a certain caeteris-paribus law 

in order to 

introduce a psychological sub-instantiable

and their specification by reference to content: 

thus 

“willing N"

 is introduced as the specific form of 

Prichard’s willing 

which is dependent on the intake and deprivation of a necessity N, 

— "prehending F as R" as the variety of prehension which normally results from an F being R to one, and similarly for "joining". 

Then we use a caeteris-paribus law 

to eliminate reference to the psychological instantiable 

and to reach the behaviour which is to be explained.

 A more developed account would, no doubt, bring in an intermediate laws, relating simply to a psychological instantiable and not also to features of the common world.

The generalities used have at various points law-allusiveness.

Every reference to, for example, 'a sufficiently prolonged deprivation', 'a sufficiently frequent performance', or to 'correspondence in certain ways' in effect implies a demand on some more developed theory actually to produce a law which is here only asserted to be producible.

The psychological concept introduced have been defined only for a very narrow range of complements.

Thus, 

judging 

has so far been defined only for complements which correspond in a certain way with those involved in the expression of hypothetical judgements.

Judging is, one might say, so far only a species of "if-judging"; and

"will" 

has been defined, so far, only for complements specifying necessities. 

Attention would have to be given to the provision of procedures for extending the range of these concepts.

If the procedure is, in its general character, correct (and Grice has not even attempted to show that it is), the possibility of impact on some familiar philosophical issues is already discernible. 

To consider only "prehension", which is intended as a prototype for perception:

potch and cotch

"Prehending F as R" lacks the existential implication that there *is* an F which is R to the prehender.

This condition could be added, but it would be added. 

This might well please friends of the concept of a sense-datum, such as Paul.

The fact that reference to a physical situations has to be made in the introduction of expressions for specific prehensions might well be very *unwelcome* to a phenomenalist.

The fact that, in the introduction of expressions for specific prehensions, a demand is imposed on a further theory to define functions mapping such prehensions on to physical situations might well prove fatal to a sceptic about reality or the material world. 

How can such a sceptic, who is unsceptical about descriptions of sense-experience, combine the demand implicit in such a description with a refusal to assent to the existence of the physical situation which, it seems, the further theory would require in order to be in a position to meet the demand?

Grice has so far been occupying himself with questions about the structure of the kind of psychological theory which would fulfil the function of bestowing the breath of life on a psychological concept.

Grice moves now to asking by what methods one might hope to arrive at an acceptable theory of this sort.

One procedure, which Grice des not in the least despise, would be to take as a body of data our linguistic intuitions concerning what we would or would not be prepared to say when using a psychological term.

As a first stage, to look for a principle — (perhaps involving an artificially constructed concept — which would seem to generalise a feature of this or that section of our discourse, making adjustments when a provisionally accepted principle leads to a counter-intuitive or a paradoxical result; 

and, as a second stage, to attempt 

to systematise the principle which has emerged piecemeal in the first stage, paying attention to the adequacy of a theory so constructed to account for the data conformably with a general criterion for the assessment of a theory.

This would be to operate in two stages approximately corresponding to what the Greeks and Leibniz called an "analytic" (or "dialectical") procedure, and a “synthetic" procedure.

Grice is, for various reasons, not happy to confine myself to the dialectical method and the synthetic method.

In applying them, many alternative theoretical options may be available, upon which will often be difficult to hit.

 Grice would like, if it is possible, to find a procedure which would tell him sooner and louder if he is on the right track — a procedure, indeed, which might in the end tell Grice that a particular theory is the right one, by some test beyond that of the ease and elegance with which it accommodates the data, and saves the phenomena. 

 Grice suspects that a dividend of this sort is what Kant expects a transcendental argument to yield 

Grice would like a basis for the construction of some analogue of a transcendental argument.

 Grice is also influenced by a different consideration.

 Grice is much impressed by the fact that arrays of this or that psychological concept, of differing degrees of richness, are applicable to creatures of differing degrees of complexity, with Homo sapiens sapiens — so far — at the peak. 

So Grice would like a procedure which would do justice to this kind of continuity — or logically developing series — and would not leave Grice just pursuing a a separate psychological theory for each species. 

The method which Grice should like to apply is to construct (in imagination, of course), according to a certain principle of construction, a sequence of species, to serve as a model for a given species.. 

Grice’s species he calls a pirot — which, Russell and Carnap tell us, karulizes elatically.

The general idea is to develop, sequentially, the psychological theory for different sup-species of a pirot, and to compare what one thus generates with the psychological concept we apply to this or that suitably related species, and when inadequacies appear, to go back to the drawing-board to extend or emend the construction — which of course is unlikely ever to be more than partial.

The principle of pirot-construction may be thought of as embodied in what I shall call a the programme of a Genitor — the main aspects of which Grice shall now formulates.

We place ourselves in the position of The Genitor, who is engaged in designing this or that living thing — or rather, as Grice shall say, an operant.

 An operant may, for present purposes, be taken to be a thing for which there is a certain set of operations requiring expenditure of energy stored in the operant, a sufficient frequency of each operation in the set being necessary to maintain the operant in a condition to pertorm any in the set le. to avoid becoming an ex-operant). 

Specific differences within such sets will determine different types of operant. 

The genitor will at least have to design them with a view to their survival (continued operancy); 

For, on certain marginal assumptions which it will be reasonable for us to make but tedious for me to enumerate, if an operant (x) is not survival-oriented, there is no basis for supposing it to exist at all; since (by the assumption) x has to be produced in some way or other by other operants of the type, x will not exist unless pro-genitors are around to produce it; and if they are to have the staying power (and other endowments) required for x's production, x, being of the same type, must be given the same attributes. 

So in providing for the individual x, some provision for the continuation of the type is implicit. In order to achieve economy in assumptions, we shall suppose the genitor to be concerned only to optimize survival chances.

Since the Genitor is only a fiction, he is to be supposed only to design, not to create. 

In order that his designs should be useful to philosophical psychology, he must be supposed to pass his designs on to the engineer (who, being also a fiction, also does not have the power to create): 

The function of the engineer is to ensure that the designs of the genitor can be realized in an organism which behaves according to pre-psychological laws — e.g. those of physics. 

Since the genitor does not know about engineering, and since he does not want to produce futile designs, he had better keep a close eye on the actual world in order to stay within the bounds of the possible.

The mode of construction is to be thought of as being relative to some very generally framed "living-condition" concerning the relation of a pirot to its environment; 

the operations the capacity for which determines the type of the pirot are to be those which, given the posited condition, constitute the minimum which the pirot would require in order to optimize the chances of his remaining in a condition to perform just those operations. 

Some pirots (plant-like) will not require psychological apparatus in order that their operations should be explicably performed; others will. 

Within the latter class, an ascending order of psychological types would, I hope, be generated by increasing the degree of demandingness in the determiningcondition. 

I cannot specify, at present, the kind of generality which is to attach to such conditions; but I have in mind such a sequence as operants which do not need to move at all to absorb sources of energy, operants which only have to make posture changes, operants which, because the sources are not constantly abundant, have to locate those sources, and (probably a good deal later in the sequence) operants who are maximally equipped to cope with an indefinite variety of physiologically tolerable environments — i.e., perhaps, a 

rational pirot.

Further types (or sub-types) might be generated by varying the degree of effectiveness demanded with respect to a given living-condition.

Aristotle regarded types of soul — as I would suppose, of living thing — as forming a "logically developing series".

Grice interprets that idea as being the supposition that the psychological theory for a given type is an extension of, and includes, the psychological theory of its predecessor-type. 

The realization of this idea is at least made possible by the assumption that psychological laws may be of a ceteris paribus form, and so can be modified without emendation. 

If this aspect of the programme can be made good, we may hope to safeguard the unity of psychological concepts in their application to animals and to human beings. 

Though, as Witters notes, certain animals can only expect such items as food, while men can expect a drought next summer, we can (if we wish) regard the concept of expectation as being determined not by the laws relating to it which are found in a single psychological theory, but by the sequences of sets of laws relating to it which are found in an ascending succession of psychological theories.

Since the Genitor is only to install psychological apparatus in so far as it is required for the generation of operations which would promote survival in a posited living-condition, no psychological concept can be instantiated by a pirot without the supposition of behaviour which manifests it. 

An explanatory concept has no hold if there is nothing for it to explain. 

This is why 'inner states must have outward manifestations'.

 Finally, we may observe that we now have to hand a possible solution of the Selection Problem.

Only those laws which can be given a genitorial justification — which fall within 'Pure Psychology' — will be counted as helping to determine a psychological concept. 

It is just because one is dubious about providing such a justification for this or that envisaged "Optimism" law that one is reluctant to regard either of them as contributing to the definition of believing and wanting.

Kant thinks that, in relation to the philosophy of nature, an idea of pure reason could legitimately be given only a regulative employment.

Somewhat  similarly, as far as philosophical psychology is concerned, Grice thinks of the genitorial programme (at least in the form just characterised) as being primarily a heuristic device. 

Grice would, however, hope to be able to retain a less vivid reformulation of it, in which references to the genitor's purposes would be replaced by references to final causes, or (more positivistically) to survival-utility.

This reformulation I would hope to be able to use to cope with such matters as the Selection problem. 

But , also much as Kant thinks with respect to an idea of pure reason, when it comes to ethics, the programme of the Genitor, in its more colourful form, just may have a role which is *not* purely heuristic. 

The thought that, if one were genitor, one would install a certain feature such as RATIO, which characteristically leads to certain forms of behaviour might conceivably be a proper step in the *evaluation* of validation of a feature such as RATIO — and of the associated behaviour.

Grice is a little more explicit, and a great deal more speculative, about the possible relation to *practical* philosophy of his programme for philosophical psychology. 

Grice shall suppose that the programme of the Genitor has been realised to the point at which we have designed a class of pirot which, nearly following Locke (Essay, i. 27. 8), Grice might call a ‘very intelligent, rational’ pirot.

Specimens of this pirot will be capable of putting themselves in the position of the Genitor, of asking how, if they were constructing themselves with a view to their own survival, they would execute this task.

And, if we have done our work aright, the answer from the pirots will be the same as ours. 

In virtue of the rational capacities and dispositions which we have given the pirots, and which they would give themselves, each of them will have both the capacity and the desire to raise the further question 'Why go on surviving?

And — Grice hopes — they will be able to justify his continued existence by endorsing — in virtue of the aforementioned *rational* capacities and dispositions — a set of criteria for evaluating and ordering this or that end, and by applying these criteria, both to this or that end — such as hobbling — which he may already have, as an indirect aid to survival, and to this or that end which are yet to be selected.

Such an end, Grice may say, will not necessarily be restricted to a concern for himself. 

The justification of the pursuit of some system of this end and that end would, in its turn, provide a justification for his continued existence.

We might, indeed, envisage the contents of a highly general practical manual, which these pirots would be in a position to compile.

This manual, though perhaps short, might serve as a basis for more specialised manuals — such as a conversational manual — to be composed when the pirots have been diversified by the harsh realities of actual existence. 

The contents of the initial manual would have various kinds of generality which are connected with familiar discussions of universalizability.

The pirots have, so far, been endowed only with the characteristics which belong to the genitorially justified psychological theory.

So the manual will have to be formulated in terms of the concepts of that theory, together with the concepts involved in the very general description of living-conditions which have been used to set up that theory.

The manual will therefore have conceptual generality. 

There will be no way of singling out a special subclass of addressees.

So the injunctions of the manual will have to be addressed, indifferently, to any ‘very intelligent rational’ pirot.

And will thus have generality of form. 

And since the manual can be thought of as being composed by each of the so far indistinguishable pirots, no pirot would include in the manual injunctions prescribing a certain line of conduct in circumstances to which he is not likely to be subject.

Co-operation.

 Indeed could he do so, even if he would.

So the circumstances for which conduct is prescribed could be presumed to be such as to be satisfied, from time to time, by any addressee.

The manual, then, will have generality of application.

Such a manual might, perhaps, without ineptitude be called an IMMANUEL.

And the ‘very intelligent, rational’ pitots, each of whom both composes it and from time to time heeds it, might indeed be not just the Phola Dactylus, but Homo sapiens sapiens — ourselves — in our better moments, of course — when we transubstantiate ourselves as Persons.

Grice’s purpose is to give a little thought to the question 'What is the general principle exemplified, in the construction of the pirot — in progressing from one type of pirot to a higher type? 

What kinds of steps are being made?' 

The kinds of step with which Grice deals here are those which culminate in a licence to include, within the specitication of the content of the psychological state of certain pirots, a range of expressions which would be inappropriate with respect to lower pirots.

Such an expression include connectives, quantifiers, temporal modifiers, mood-indicators, modal operators, and, importantly, names and definitions of a psychological state like "judge" or will"; expressions the availability of which leads to the structural enrichment in the specification of content. 

In general, these steps will be ones by which items or ideas which have, initially, a legitimate place outside the scope of psychological instantiables — or, if you will, the expressions for Grice is disposed to regard as prototypical the sort of natural disposition which Hume attributes to us, and which is very important to Hume; namely, the tendency of the mind 'to spread itself upon objects', to project into the world items which, properly, or primitively, considered, are really features of our states of mind. 

Though there are other examples in Hume’s work, the most famous case in which he tries to make use of this tendency to deal with a philosophical issue is his treatment of necessity.

The idea of a necessary connection between a cause and its effect is traced, roughly, to an internal impression which attends the passage of the soul or mind from an idea or impression to an associated idea. 

Grice is not wholly happy about Hume's characterization of this kind of projection, but what I have to say will bear at least a generic resemblance to what may be found in Hume.

The following kinds of transition seem to me to be characteristic of internalization.

References to psychological states (-states) may occur in a variety of linguistic settings which are also appropriate to references to states which have no direct connection with psychology.

 Like volcanic eruptions, judgings or willings may be assigned both more precise and less precise temporal locations (e.g. last Thursday or next Tuesday, in the future or in the past); sometimes specifications of degree will be appropriate; judging (A] or willing [A] may be assigned a cause or an effect; and references to such y-states will be open to logical operations such as negation and disjunction. 

To suppose that a creature will, in the future, judge (A), or that he either judges [A] or judges (B), is obviously not to attribute to him a judging distinct from judging (Al, or trom judging[A] and from judging [B]; nevertheless, we can allow a transition from these attributions to the attribution ofwhich occur legitimately outside the scope of psychological verbs) come to have a legitimate place within the scope of such instantiables: steps by which (one might say) such items or ideas come to be internalizedI am disposed to regard as prototypical the sort of natural disposition which Hume attributes to us, and which is very important to him; namely, the tendency of the mind 'to spread itself upon objects', to project into the world items which, properly (or primitively) considered, are really features of our states of mind. 

Though there are other examples in his work, the most famous case in which he tries to make use of this tendency to deal with a philosophical issue is his treatment of necessity; the idea of necessary connection is traced (roughly) to an internal impression which attends the passage of the mind from an idea or impression to an associated idea. I am not wholly happy about Hume's characterization of this kind of projection, but what I have to say will bear at least a generic resemblance to what may be found in Hume.

The following kinds of transition seem to me to be characteristic of internalization. References to psychological states (-states) may occur in a variety of linguistic settings which are also appropriate to references to states which have no direct connection with psychology. Like volcanic eruptions, judgings or willings may be assigned both more precise and less precise temporal locations (e.g. last Thursday or next Tuesday, in the future or in the past); sometimes specifications of degree will be appropriate; judging [A] or willing [A] may be assigned a cause or an effect; and references to such y-states will be open to logical operations such as negation and disjunction. 

To suppose that a creature will, in the future, judge [A], or that he either judges [A] or judges (B], is obviously not to attribute to him a judging distinct from judging [A], or from judging[A] and from judging [B]; nevertheless, we can allow a transition from these attributions to the attribution ofjudgings which are distinct, viz. of "future-judging [A]"(expecting [A]) and of "or-judging [A, B]" Since these arenew y-states, they will be open to the standard range of linguistic settings for references to y-states; a creature may now (or at some time in the future) future-judge [A](or past-judge [A]). 

There will be both a semantic difference and a semantic connection (not necessarily the same for every case) between the terminology generated by this kind of transition and the original from which it is generated; while "x or-judges [A, B]" will not require the truth of "x judges (A] or x judges (B]", there will be a connection of meaning between the two forms of expres-sion; if a creature x or-judges its potential prey as being [in the tree, in the bush], we may expect x to wait in between the tree and the bush, constantly surveying both; or-judging [A, B] will, in short, be typically manifested in behaviour which is appropriate equally to the truth of A and to the truth of B, and which is preparatory for behaviour specially appropriate to the truth of one of the pair, to be evinced when it becomes true either that x judges [A] or that x judges [B].

This kind of transition, in which an "extrinsic" modifier attachable to y-expressions is transformed into (or replaced by) an "intrinsic" modifier (one which is specific-atory of a special (-state), I shall call first-stage internal-ization.

It should be apparent from the foregoing discussion that transitions from a lower type to a higher type of pirot will often involve the addition of diversifications of a y-state to be found undiversified in the lower type of pirot; we might proceed, for example, from pirots with a capacity for simple judging to pirots with a capacity for present-judging (the counterpart of the simple judging of their predecessors), and also for future-judging (primitive expecting) and for past-judging (primitive remembering).

It may be possible always to represent developmental steps in this way; but I am not yet sure of this. It may be thatsometimes we should attribute to the less-developed pirot two distinct states, say judging and willing; and only to a more developed pirot do we assign a more generic state, say accepting, of which, once it is introduced, we may represent judging and willing as specific forms (judicatively accepting and volitively accepting).

 But by whatever route we think of ourselves as arriving at it, the representation within a particular level of theory of two or more y-states as diversifications of a single generic y-state is legitimate, in my view, only if the theory includes laws relating to the generic -state which, so far from being trivially derivable from laws relating to the specific -states, can be used to derive in the theory, as special cases, some but not all generalities which hold for the specific y-states. Wemight, for example, justity the attribution to a pirot of the generic y-state of accepting, with judging and willing as diversifications, by pointing to the presence of such a law as the following: ceteris paribus, if x or-accepts (in mode m of acceptance) [A, B], and x negatively accepts, in mode m, [A], then x positively accepts, in mode m, [B]. 

From this law we could derive that, ceteris paribus, (a) if x or-judges [A, B] and negatively judges [A], then x positively judges [B]; and (b) if x or-wills [A, B] and negatively wills [A], then x positively wills (BJ. ( (a) and (b) are, of course, psychological counterparts of mood-variant versions of modus tollendo ponens.)

A further kind of transition is one which I shall label second-stage internalization. It involves the replacement of an "intrinsic" modifier, which signifies a differentiating feature of a specific y-state, by a corresponding operator within the content-specifications for a less specific v-state. 

If, for example, we have reached by first-stage internalization the -state of future-judging [A], we may proceed by second-stage internalization to the y-state of judging [in the future, A]: similarly, we may proceed from judging (judicatively accepting) [A] to accepting [it is the case that Al, from willing (volitively accepting) [A] toaccepting (let it be that Al, and from disjunctively judging (or-judging) [A, B] to judging [A or B). 

So long as the operator introduced into a content-specification has maximum scope in that specification, the result of second-stage internalization is simply to produce a notational variant:

x judges (A or B]" is semantically equivalent to "x disjunctively judges [A, B]", and if matters rested here, there would be no advance to a new level of theory. 

To reach a new level of theory, and so to reach y-states exemplifiable only by a higher type of pirot, provision must be made for y-states with contents in the specification of which the newly introduced operator is embedded within the scope of another operator, as in "x judges [if A or B, C]", or in "x accepts (if A, let it be that B]". 

An obvious consequence of this requirement will be that second-stage internalizations cannot be introduced singly; an embedded operator must occur within the scope of another embeddable operator.

 The unembedded occurrences of a new operator, for which translation back into the terminology of the previous level of theory is possible, secure continuity between the new level and the old; 

indeed, Grice suspects that it is a general condition on the development of one theory from another that there should be cases of "overlap" to ensure continuity, and cases of non-overlap to ensure that a new theory has really been developed.

Two observations remain to be made. First, if we enquire what theoretical purpose is likely to be served by endowing a type of pirot with the capacity for y-states the specification of which involves embedded operators, one important answer is likely to be that explanation of the ranges of behaviour to be assigned to that type of pirot requires a capacity for a relatively high grade of inference, in which the passage from premisses to conclusion is encompassed, so to speak, within a single thought; the pirot has to be capable, for example, of thinking [since, A, B, and C, then DJ. 

This capacity may well require otheraccepting (let it be that Al, and from disjunctively judging (or-judging) [A, B] to judging [A or B). So long as the operator introduced into a content-specification has maximum scope in that specification, the result of second-stage internalization is simply to produce a notational variant:

"x judges (A or B]" is semantically equivalent to "x disjunctively judges [A, B]", and if matters rested here, there would be no advance to a new level of theory. 

To reach a new level of theory, and so to reach y-states exemplifiable only by a higher type of pirot, provision must be made for y-states with contents in the specification of which the newly introduced operator is embedded within the scope of another operator, as in "x judges [if A or B, C]", or in "x accepts (if A, let it be that B]". 

An obvious consequence of this requirement will be that second-stage internalizations cannot be introduced singly; an embedded operator must occur within the scope of another embeddable operator. 

The unembedded occurrences of a new operator, for which translation back into the terminology of the previous level of theory is possible, secure continuity between the new level and the old; indeed, I suspect that it is a general condition on the development of one theory from another that there should be cases of "overlap" to ensure continuity, and cases of non-overlap to ensure that a new theory has really been developed.

Two observations remain to be made. 

First, if we enquire what theoretical purpose is likely to be served by endowing a type of pirot with the capacity for y-states the specification of which involves embedded operators, one important answer is likely to be that explanation of the ranges of behaviour to be assigned to that type of pirot requires a capacity for a relatively high grade of inference, in which the passage from premisses to conclusion is encompassed, so to speak, within a single thought; the pirot has to be capable, for example, of thinking [since, A, B, and C, then DJ. This capacity may well require other

M-states generated by internalization should conform to the restrictions propounded in Section VII; that is to say, they should not be assigned to a pirot without 

the assignment to the pirot of behaviour the presence of which they will be required to explain, and 

a genitorial justification for the presence of that behaviour in the pirotIf both generic and specific y-states are to be recognized by a theory, each such state should figure non-trivially in laws of that theory.

Second-stage internalization should be invoked only when first-stage internalization is inadequate for explanatory purposes.

Where possible, full internalization should be reached by a standard two-stage progression.

VII. Higher-Order Psychological States

In this section Grice focuses on just one of the modes of content-internalization, that in which psychological concepts (or other counterparts in psychological theory) are themselves internalized. 

I shall give three illustrations of the philosophical potentialities of this mode of internal-ization, including a suggestion for the solution of my third initial problem (Problem C), that of the accommodation, within my approach, of the phenomena of privileged access and incorrigibility.

As I have already remarked, it may be possible in a sufficiently enriched theory to define concepts which have to be treated as primitive in the theory's more impoverished predecessors. 

Given a type of pirot whose behaviour is sufficiently complex as to require a psychological theory in which psychological concepts, along with such other items as logical connectives and quantifiers, are internalized, it will I think be possible to define a variety of judging, which I shall call "judging*", in terms of willing. 

I doubt if one would wish judging* to replace the previously distinguished notion of judging; I suspect one would wish them to coexist; one would want one's relatively advanced pirots to be capable not only of the highly rational state of judging but also of the kind of judging exemplified by lower types, if only in order to attribute to such advanced pirots implicit or unconscious judgings. 

There may well be more than one option for the definition of judging*; since my purpose is to illustrate a method rather than to make a substantial proposal, I shall select a relatively simple way, which may not be the best.

The central ideal is that whereas the proper specification of a (particularized) disposition, say a man's disposition to entertain his brother if his brother comes to town, would be as a disposition to entertain his brother if he believes that his brother has come to town (ct. the circularities noted in Section I), no such emendation is required if we speak of a man's will to entertain his brother if his brother comes to town. 

We may expect the man to be disappointed if he discovers that his brother has actually come to town without his having had the chance to entertain him, whether or not he was at the time aware that his brother was in town. 

Of course, to put his will into effect on a particular occasion, he will have to judge that his brother is in town, but no reference to such a judgement would be appropriate in the specification of the content of his will.

Here, then, is the definition which I suggest. x judges* that p just in case x wills as follows: given any situation in which (i) x wills some end E, (ii) there are two non-empty classes (K, and K2) of action-types such that the performance by x of an action-type belonging to K, will realize E just in case p is true, and the performance by x of a member of K, will realize E just in case p is false, (iji) there is no third non-empty class (K3) of action-types, such that the performance by x of an action-type belonging to K3 will realize E whether p is true or p is false: then in such a situation x is to will that he perform some action-typebelonging to K,. 

Put more informally (and less accurately), x judges that p just in case x wills that, if he has to choose between a kind of action which will realize some end of his just in case p is true and a kind of action which will realize that end just in case p is false, he should will to adopt some action of the first kind.

We might be able to use the idea of higher-order states to account for a kind of indeterminacy which is frequently apparent in our desires. 

Consider a disgruntled employee who wants more money; and suppose that we embrace the idea (which may very well be misguided) of representing his desire in terms of the notion of wanting that p (with the aid of quantifiers). 

To represent him as wanting that he should have some increment or other, or that he should have some increment or other within certain limits, may not do justice to the facts; it might be to attribute to him too generic a desire, since he may not be thinking that any increment, or any increment within certain limits, would do. 

If, however, we shift from thinking in terms of internal quantification to thinking in terms of external quantification, then we run the risk of attributing to him too specific a desire; there may not be any specific increment which is the one he wants; he may just not have yet made up his mind. It is as if we should like to plant our (existential) quantifier in an intermediate position, in the middle of the word 'want'; and that, in a way, is just what we can do. 

We can suppose him (initially) to be wanting that there be some increment (some increment within certain limits) such that he wants that increment; and then, when under the influence of that want he has engaged in further reflection, or succeeded in eliciting an offer from his employer, he later reaches a position to which external quantification is appropriate; that is, there now is some particular increment such that he wants that increment.

Grice shall now try to bring the idea of higher-order states to bear on Problem C (how to accommodate privilegedaccess and, maybe, incorrigibility). 

Grice sets out in stages a possible solution along these lines, which will also illustrate the application of aspects of the genitorial programme

Stage 0. We start with pirots equipped to satisfy unnested judgings and willings (i.e. whose contents do not involve judging or willing).

Stage 1. It would be advantageous to pirots if they could have judgings and willings which relate to the judgings or willings of other pirots; for example, if pirots are sufficiently developed to be able to will their own behaviour in advance (form intentions for future action), it could be advantageous to one pirot to anticipate the behaviour of another by judging that the second pirot wills to do A (in the future). 

So we construct a higher type of pirot with this capacity, without however the capacity for reflexive states.

It would be advantageous to construct a yet higher type of pirot, with judgings and willings which relate to its own judgings or willings. 

Such pirots could be equipped to control or regulate their own judgings and willings; they will presumably be already constituted so as to conform to the law that ceteris paribus if they will that p and judge that not-p, then if they can, they make it the case that p. 

To give them some control over their judgings and willings, we need only extend the application of this law to their judgings and willings; we equip them so that ceteris paribus if they will that they do not will that p and judge that they do will that p, then (if they can) they make it the case that they do not will that p (and we somehow ensure that sometimes they can do this).

It may be that the installation of this kind of control would go hand in hand with the installation of the capacity for evaluation; but I need not concern myself with this now.]

We shall not want these pirots to depend, inreaching their second-order judgements about them-selves, on the observation of manifestational be-haviour; indeed, if self-control which involves suppressing the willing that p is what the genitor is aiming at,behaviour which manifests a pirot'sjudging that it wills that p may be part of what he hopes to prevent. So the genitor makes these pirots subject to the law that ceteris paribus if a pirot judges (wills) that p, then it judges that it judges (wills) thatp. 

To build in this feature is to build in privileged access to judgings and willings.

To minimize the waste of effort which would be involved in trying to suppress a willing which a pirot mistakenly judges itself to have, the genitor may also build in conformity to the converse law, that ceteris paribus 

if a pirot judges that it judges (wills) that P, it judges (wills) that p. 

Both of these laws however are only ceteris paribus laws.

And there will be room for counter-examples.

In self-deception, for example, either law may not hold.

We may get a judgement that one wills that p without the willing that p.

And we may get willing that p without judging that one wills that p, indeed, with judging that one does not will that p.

Grice abbreviates "x judges that x judges that p" by "x judges? that p", and "x judges that x judges that x judges that p" by "x judges" that p".

 Let us suppose that we make the not implausible assumption that there will be no way of finding non-linguistic manifestational behaviour which distinguishes judging? that p from judging that p. 

There will now be two options.

We may suppose that "judge? that p" is an inadmissible locution, which one has no basis for applying.

Or we may suppose that "x judges that p" and "x judges? that p" are manifestationally equivalent, just because there can be no distinguishing behavioural manifestation.

The second option is preferable, if we want to allow for the construction of a possibly later type, a *talking* pirot, which can EXPRESS that it judges? that p; and to maintain as a general, though probably derivative, law that ceteris paribus 

if x EXPRESSES that &, x judges that ф. 

The substitution of "x judges? that p" for " p" will force the admissibility of "x judges" that p".

 So we shall have to adopt as a law that x judges" that piff x judges? that p. 

Exactly parallel reasoning will force the adoption of the law that x judges that p if x judges? that p.

If we now define "x believes that p" as "x judges? that p", we get the result that 

a believes that p iff x believes that x believes that p. 

We get the result, that is to say, that 

— a belief is (in this sense) incorrigible, whereas a first-order judging is only a matter for privileged access.

Is this or that Psychological Concept Eliminable?

Grice has tried to shed a little light on the question 'How is a psychological concept related to psychological theory and to psychological explanation?"

Grice has not, so far, said anything about a question which has considerably vexed some of Grice’s friends as well as himself.

Someone —let us call him the eliminator — might say: 

‘You explicitly subscribe to the idea of thinking of apsychological concept as explicable in terms of its role in a psychological law, that is to say, in terms of their potentialities for the explanation of behaviour. 

You also subscribe to the idea that, where there is a psychological explanation of a particular behaviour, there is also a physiological explanation 

With respect to pirots, it is supposedly the business of the engineer to make a psychological state effective by ensuring that this condition holds. 

You must, however, admit that the explanation of behaviour, drawn from pirot-physiology, which is accessible to the engineer is  from a theoretical point of view, greatly superior to that accessible to the genitor; 

The former are, for example, drawn from a more comprehensive theory, and from one which yields  or when fully developed will yield — even in the area of behaviour, a more precise prediction.

So since the attribution of a psychological state owes any claim to truth it may have to its potentialities for the explanation of behaviour, and since these potentialities are inferior to the potentialities of this of that physiological state, Occam's Razor will dictate that attributions of psychological states, to actual creatures no less than to pirots, should be rejected en bloc as lacking truth —even though, from ignorance or to avoid excessive labour, we might in fact continue to utter them.

We have indeed come face to face with the last gasp of Primitive Animism, namely the attempted perpetuation of the myth that animals are animate.'

The eliminator should receive fuller attention than Grice can give him on this occasion, but I will provide an interim response to him, in the course of which Grice shall hope to redeem the promise made, to pursue a little further the idea that the common-sense theory which underlines this or that psychological concept can be shown to be at least in some respects true.

First, the eliminator's position should, in Grice’s view, be seen as a particular case of canonical scepticism, and should, therefore, be met in accordance with general principles for the treatment of sceptical problems. 

It is necessary, but it is also insufficient, to produce a cogent argument for the falsity of his conclusion; 

if we confine ourselves to this kind of argumentation we shall find ourselves confronted by the post-eliminator, who says to us 

“Certainly the steps in your refutation of the eliminator are validly made.”

“But so are the steps in his argument.”

A correct view of the matter is that from a set of principles in which a psychological concept occurs crucially, and through which that psychological concept is delineated, it follows both that the eliminator is right and that he is wrong.

So much the worse, then, for the set of principles and for the concepts;

they should be rejected as incoherent. 

A satisfactory rebuttal of the eliminator's position must, therefore, undermine his argument as well as overturn his conclusion.

In pursuit of the first of these objectives, Grice remarks that, to his mind, to explain a phenomenon, or a sequence of phenomena, P is, typically, to explain P qua exemplifying some particular feature or characteristic 

And so, one fact — or system — may explain P qua exemplifying characteristic C, while another fact — or system — explains P qua exemplifying a different charscteristic C2. 

In application to the present case, it is quite possible that one system, the physiology of the future, should explain the things we do considered as sequences of physical movements, while another system, psychological theory, explains the things we do considered as instances of sorts of behaviour.

We may, at this point, distinguish two possible principles concerning the acceptability of a system of explanation:

If a system S1 is theoretically more adequate, for the explanation of a class of phenomena P will respect to a class of characteristics C, than is system S2 for the explanation of & with respect to C, other things being equal, S1 is to be accepted and S2 is to be rejected.

Gricd has no quarrel with this principle.

If system S1 is theoretically more adequate, for the explanation of P with respect to C1 than is system S2 for the explanation of P with respect to C2, la different class of characteristics, other things being equal, S1 is to be accepted and S2 is to be rejected.

The latter principle seems to me to lack plausibility.

In general, if we want to be able to explain P qua C2, e.g. as behaviour, our interest in such explanation should not be abandoned merely because the kind of explanation available with respect to some other class of characteristics is theoretically superior to the kind available with respect to C2. 

The eliminator, however, will argue that, in the present case, what we have is really a special case of the application of Principle 1; 

for 

since any Ca in Cz which is exemplified on a given occasion has to be realised on that occasion in some , (some sequence of physical movements), for which there is a physiological explanation, the physiological explanation of the presence of C, on that occasion is ipso facto an explanation of the presence of C2, the behavioural feature, on that occasion, and so should be regarded as displacing any psychological explanation of the presence of Ca.

To maintain this position, the eliminator must be prepared to hold that if a behavioural feature is, on an occasion, realised in a given type of physical movement sequence,  that type of a movement sequence is, in itself, a sufficient condition for the presence of the behavioural feature.

Only so can the eliminator claim that the full power and prestige of physiological explanation is transmitted from a movement to behaviour, thereby ousting psychological explanation. 

But it is very dubious whether a particular a movement sequence is, in general, a sufficient condition for a behavioural feature, especially with respect to such a behavioural feature the discrimination of which is most important for the conduct of life.

 Whether a man with a club is merely advancing *towards* Grice or advancing *upon* Grice may well depend on a condition distinct from the character of his movements, indeed, it would be natural to suppose, upon a psychological condition.

 It might of course be claimed that any such psychological condition would, in principle, be re-expressible in psychological terms, but this claim cannot be made by one who seeks not the reduction of psychological concepts but their rejection.

If a psychological state is not to be reduced to a physiological state, the proceeds of such reduction cannot be used to bridge gaps between a movement and behaviour.

The truth seems to be that the language of behavioural description is part and parcel of the system to which psychological explanation belongs, and cannot be prised off therefrom. 

Nor can the eliminator choose to be hanged for a sheep rather than for a mere lamb, by opting to jettison behavioural description along with psychological concepts.

The eliminator himself will need behavioural terms like"describe" and "report" in order to formulate his non-recommendations or predictions.

And, perhaps more importantly, if there is, or is to be, strictly speaking no such thing as the discrimination of behaviour, there are, or are to be, no such things as ways of staying alive, either to describe or to adopt.

Furthermore, even if the eliminator were to succeed in bringing psychological concepts beneath the baleful glare of Principle 1, he would only have established that other things being equal psychological concepts and psychological explanation are to be rejected. 

But other things are manifestly not equal, in ways which he has not taken into account. 

It is one thing to suggest, as Grice suggests, that a proper philosophical understanding of ascriptions of psychological concepts is a matter of understanding the roles of such concepts in a psychological theory which explains behaviour.

It is quite *another* thing to suppose that creatures to whom such a theory applies will (or should) be interested in the ascription of psychological states only because of their interest in having a satisfactory system for the explanation of behaviour. 

The psychological theory which Grice envisages would be deficient as a theory to explain behaviour if it did not contain provision for interests in the ascription of psychological states otherwise than as tools for explaining and predicting behaviour, interests, for example, on the part of one creature to be able to ascribe these rather than those psychological states to another creature because of a concern for the other creature.

 Within such a theory it should be possible to derive strong motivations on the part of the creatures subject to the theory against the abandonment of the central concepts of the theory (and so of the theory itself), motivations which the creatures would (or should) regardas justified.

Grice illustrates with a little fable.

 The very eminent and very dedicated neurophysiologist speaks to his wife. 

"My (for at least a little while longer) dear," he says,

 "I have long thought of myself as an acute and well-informed interpreter of your actions and behaviour.”

“I think I have been able to identify nearly every thought that has made you smile, and nearly every desire that has moved you to act.”

“My researches, however, have made such progress that I shall no longer need to understand you in this way.”

“Instead I shall be in a position, with the aid of instruments which I shall attach to you, to assign to each bodily movement which you make in acting a specific antecedent condition in your cortex.”

“No longer shall I need to concern myself with your so-called thoughts and feelings.”

“In the meantime, perhaps you would have dinner with me tonight”

“I trust that you will not resist if I bring along some apparatus to help me to determine, as quickly as possible, the physiological idiosyncracies which obtain in your system.”

Grice has a feeling that Mrs. Churchland might refuse the proffered invitation.

Indeed, only from within the framework of such a theory, Grice thinks, can matters of evaluation, and so of the evaluation of modes of explanation, be raised at all. 

If Grice conjectures aright, the entrenched system contains the materials needed to justify its own entrenchment — whereas no rival system contains a basis for the justification of anything at all.

We must be ever watchful against the devil of scientism, who would lead us into myopic over-concentration on the nature and importance of knowledge, and of scientific knowledge in particular; the devil who is even so audacious as to tempt us to call in question the very system of ideas required to make intelligible the idea of calling in question anything at all; and who would even prompt us, in effect, to suggest that, since we do not really think but only think that we think, we had better change our minds without undue delay. 

H. P. Grice