H. P. Grice: Conversational Implicata and Conversation As Rational Co-Operation

THn Causal Theory of Perception (CTP) has for some time
received comparatively little attention, mainly, I suspect, because
it has been generally assumed that the theory either asserts or
involves as a consequence the proposition that material objects
are unobservable, and that the unacceptability-, this -fo proposition is sufficient to dispose of the theory. I am inclined to
regard this attitude to the CTP as unfair or at least unduly
unsympathetic and I shall attempt to outline a thesis which
might not improperly be considered to be a version of the CTP,
and which is, if not true, at least not too obviously false.
What is to count as holding a causal theory of perception?
(1) I shall take it as being insufficient merely to believe that the
perception of a material object is always to be causally explained
by reference to conditions the specification of at least one of
which involves a mention of the object perceived; that, for
example, the perception is the terminus of a causal sequence
involving at an earlier stage some event or process in the history
of the perceived object. Such a belief does not seem to be
philosophical in character; its object has the appearance of
being a very general contingent proposition; though it is worth
remarking that if the version of the CTP with which I shall be
primarily concerned is correct, it (or something like it) will turn
out to be a necessary rather than a contingent truth. (2) It may
be held that the elucidation of the notion of perceiving a material
object will include spme reference to the r6le of the material
object perceived in the causal ancestry of the perception or of
the sense-impression or sense-datum involved in the perception).
This contention is central to what I regard as a standard version
of the CTP. (3) It might be held that it is the task of the
philosopher of perception not to elucidate or characterize the
H. P. GRICE
ordinary notion of perceiving a material object, but to provide a
rational reconstruction of it, to replace it by some concept more
appropriate to an ideal or scientific language: it might further
be suggested that such a redefinition might be formulated in
terms ofthe effect ofthe presence ofan object upon the observer's
sense-organ and nervous system or upon his behaviour or
" behaviour-tendencies " or in terms of both of these effects.
A view of this kind may perhaps deserve to be called a causal
theory of perception; but I shall not be concerned lvith theories
on these lines. (4) I shall distinguish from the adoption of a
CTP the attempt to provide for a wider or narrower range of
propositions ascribing properties to material objects a certain
sort of causal analysis: the kind of analysis which I have in
mind is that which, on one possible interpretation, Locke could
be taken as suggesting for ascriptions of, for example, colour and
temperature; he might be understood to be holding that such
propositions assert that an object would, in certain standard
conditions, cause an observer to have certain sorts of ideas or
sense-impressions.
In Professor Price's Perception,t there appears a preliminary
formulation of the CTP which would bring it under the second
of the headings distinguished in the previous paragraph. The
CTP is specified as maintaining (1) that in the case of all sensedata (not merely visual and tactual) " belonging to " simply
means being caused Dy, so that ' M is present to my senses ' will
be equivalent to ' M causes a sense-datum with which I am
acquainted '; (2) thaf perceptual consciousness is fundamentally
an inference frorn effect to causc. Since it is, I think, fair to says
that the expression " present to my senses "was introduced by
Price as a special term to distinguish one of the possible senses
of the verb " perceive ", the first clause of the quotation above
may be taken as propounding the thesis that " I am perceiving
M" (in one sense of that expression) is to be regarded as
equivalent to " I am having (or sensing) a sense-datum which is
caused by M." (The second clause I shall for the time being
ignore.) I shall proceed to consider at some length the feature
1 P. 66.
2 Cf. ibid., pp.2l-25.
122 THE CAUSAL THEORY OF PERCEPTION
which this version of the CTP shares with other non-causal
theories of perception, namely, the claim that perceiving a
material object involves having or sensing a sense-datum; for
unless this claim can be made out the special features of the
CTP become otiose.
2".
The primary difficulty facing the contention that perceiving
involves having or sensing a sense-datum is that of giving a
satisfactory explanation of the meaning of the technical term
' sense-datum '. One familiar method of attempting this task is
that of trying to prove, by means of some form of the Argument
from lllusion, the existence of objects of a special sort for which
the term 'sense-datum' is offered as a class-name. Another
method (that adapted in a famous passage by Moore) is that of
giving directions which are designed to enable one to pick out
items of the kind to which the term ' sense-datum' is to be
applicd. Thc gencral clraractcr of the objections to each of these
proccdr.rlcs is also fanriliar, ar.rcl Ishall, for present purposes,
assumc that ncithcr proccdurc is satisfactory. .,..
Various philosoplrcrs havc suggcstcd tlrat though attellpts to
indicate, or demonstrate thc cxistcncc of, spccial objccts to be
called sense-data have all failed, ncvertlrcless the expression
'sense-datum' can (and should) be introduced as a technical
term; its use would be explicitly defined by reference to such
supposedly standard locutions as " So-and-so looks Q (e.9.,
blue) to me ", " It looks (feels) to me as if there were a @ so-andSo ", " I seem to see something (D " and so on. Now as the
objection to such proposals which I have in mind is one which
might be described as an objection in principle, it is not to my
present purpose to consider how in detail such an explicit
definition-of the notion of a sense-datum might be formulated.
I should, however, remark that this programme may be by no
means so easy to carry through as the casual way in which it is
sometimes proposed might suggest; various expressions are
candidates for the key r61e in this enterprise e.g., " looks (" feels "
etc.), " seems ", " appears " and the more or less subtle differences
between them would have to be investigated; and furthermore
even if one has decided on a preferred candidate, not all of its
r23
H. P. GRICE
uses would be suitable; if for example we decide to employ the
expressions " looks " ete., are we to accept the legitimacy of the
sentence " It looks indigestible to me " as providing us with a
sense-datum sentence " I am having an indigestible visual sense_
datum " ?
A general objection to the suggested procedure might run as
follows: When someone makes such a remark as ..It looks red to
me " a certain implication is carried, an implication which is
disjunctive in form. It is implied either that the object referred
to is known or believed by the speaker not to be rcd,, or that it
has been denied by someone else to be red, or that the speaker is
doubtful whether it is red, or that someone else has expressed
doubt whether it is red, or that the situation is such that though
no doubt has actually been expressed and no denial has actuaiy
been made, some person or other might feel inclined towards
denial or doubt if he were to address himself to the question
whether the object is actually red. This may not be an absolutely
exact or complete characterizalion of the implication, but it is
perhaps good enough to be going on with. Let us refer to the
condition which is fulfilled when one or other of the limbs of
this disjunction is true tlie D-or-D condition (. doubt or denial '
condition). Now we may perhaps agree that there is liable to be
something odd or even absurd about employing an ., It looks
to me " locution when the appropriate D-or-D co'dition is fairlv
obviously not fulfilled; there would be something at least
primafacie odd about my saying " That looks red to me " (not as
a jollocutions in question (unless of course he were intending to
deceive his audience into thinking that the condition was fulfilled),
(b) that in cases where the D-or-D condition is unfulfilled the
utterance employing the " looks to me', locution, so far from
being uninterestingly true, is neither true nor false. Thus armed.
my objector now assails the latter-day sense-daturp theorist.
r24 TIIE CAUSAL TIIEORY OF PERCEPTION
Our every day life is populated with cases in which the sensible
characteristics of the things we encounter are not the subject of
any kind of doubt or controversy; consequently there will be
countless situations in which the ernployment of " looks to me "
idiom would bc out of order and neither true nor false. But the
sense-datum theorist wants his sense-datum statements to oc such
that some one or more of them is true whenever a perceptual
statement is true; for he wants to go on to give a general analysis
of perceptual statements in terms of the notion of sense-data.
But this goal must be unattainable if " looks to me " statements
(and so sense-datum statements) can be truly made only
in the /ess straightforward perceptual situations; and if the goal
is unattainable the CTP collapses.
It is of course possible to take a different view of the linguistic
phenomena outlined in my previous paragraph. One may
contend that if I were to say " it looks red to me " in a situation
in which the D-or-D condition is not fulfilled, what I say is
(subject to certain qualifications) true, not " neuter"; while
admitting that though truc it might be very misleading and that
its truth might bc vcry boring and its rnisleadingness very
important, one might still hold that its suggestio falsi is perfectly
compatible with its literal truth. Furthermore one might argue
that though perhaps someone who, without intent to deceive,
employed the " it looks to me " locution when he did not suppose
tliie D-or-D condition to be fulfilled would be guilty in some sense
of a misuse of language, he could be said not to be guilty of a
misuse of the particular locution in question; for, one might say,
the implication of the fulfllment of the D-or-D condition attaches
to such locutions not as a special feature of the meaning or use of
these expressions, but in virtue of a general feature or principle
of the use of language. The mistake of supposing the implication
to constitute a " part of the meaning 'r of " looks to me " is
somewhat similar to, though more insidious than, the mistake
which would be made if one supposed that the so-called implication that one believes it to be raining was " a part of the meaning "
of the expression " it is raining ". The short and literally
inaccurate reply to such a supposition might be that the sb-called
implication attaches because the expression is a propositional one,
not because it is the particular propositional expression which it
happens to be.
125
126 H. P. GRICE
Until fairly recently it seemed to me to be very difficult indeed
to find any arguments which seemed at all likely to settle the
issue between these two positions. One might, for example,
suggest that it is open to the champion of sense_data to lay down
that the sense-datum sentence " I have a pink sense-datum "
should express truth if and only if the facts are as they would
have to be for it to be true, if it were in order, to say .. Something
looks pink to me ", even though it may not actually be in ordei
to say this (because the D-or-D condition is unfulfilled). But
this attempt to by-pass the objector's position would be met
by the reply that it begs the question; for it assumes that there
is some way of specifying the facts in isolation from the implication standardly carried by such a specification; and this is
precisely what the objector is denying. As a result of frustrations
i of this kind, I was led to suspect that neither position should be
regarded as right or wrong, but that the linguistic phenomena
could be looked at in either way, though there might be reasons
for preferring to adopt one way of viewing them rather than the
other; that there might be no proofs or disproofs, but only
inducements. on this assumption I was inclined to rule against
my objector, partly because his opponent's position was more in
line with the kind of thing I was inclined to say about other
linguistic phenomena which are in some degree comparable, but
mainly because the objector's short way with sense-data is an
even shorter way with scepticism about the material world; and
I think a sceptic might complain that though his worries may well
prove dissoluble he ought at least to be able to state them; if
we do not allow him to state them we cannot remove the real
source of his discomfort. However, I am now inclined to think
that the issue is a decidable one, and that my objector's position
is wrong and that of his opponent right. I shall attempt to
develop a single argument (though no doubt there are otheis) to
support this claim, and as a preliminary I shall embark on a
discursus about certain aspects of the concept or concepts of
irnplication, using some more or less well-worn
(The obj e ctio n r e cons id e re d) Let us now revert to the main topic of this section of my paper. Let us call a statement of the type expressible by such a THB cAUSAL THEoRy oF pBRcEprroN 133 sentence as " it looks rcd to me " an l-statement. What are we to say of the relation between an ,L-statcment and the corresponding D-or-D condition, in terms o[ thc ideas introduced in the previous sub-section ? Or, rathcr, since this might be controversial, what would my objector think it correct to say on this subject. As I have represented his position, he is cxplicitly committed to holding that the fulfilment of the appropriate D-or-D condition is a necessary pre-condition of a Z-statement's bcing cithcr true or false. He is also more or less cxplicitly committcd to holding that the implication that the D-or-D condition is fulfllled is a matter of the meaning of the word " looks " (or of the phrase " looks to me "); that, for example, sorneone who failed to realise that there existed this implication would thereby show that he did not fully understand the meaning of the expression or phrase in question. It is conceivable that this last-mentioned thesis is independent of the rest of his position, that he could if necessary abandon it without destroying the rcrnainder of his position. I shall not, therefore, in what lbllows address myself directly to this point, though I have hopes that it may turn out to be solutum ambulando. Next, he would, I think, wish to say that the implication of the fulfllment of the D-or- D condition is neither detachable nor cancellable; but even il' hc slrould not wish to say this, he certainly must say it if his ob.jcction is to be of any importance. For if the implication is tlctachable or cancellable, all that the sense-datum theorist needs Io do is to flnd some form of words from which the implication is dotached or in which it is cancelled, and use this expression to tlclinc the 'notion of a sense-datum. It is not enough that ,r(r/,r(, ways of introducing sense-data should be vulnerable to his rrbjcction; it is essentialthat all should be vulnerable. Finally, it is not obvious that he is committed either to asserting or to tlcnying any of the possibilities as regards what may be spoken ol' ls being the vehicle of implication, so I shall not at moment l)ulsr.lo this matter, though I shall suggest later that he can only nurintain his position by giving what in fact is certainly a wrong nnswcr to this question. It is now time for the attack to begin. It seems to me that lfrt: contcntion that the fulfilment of the D-or-D condition is a rcr:cssary condition of the truth or falsity of an l-statement r34 H. P. GRICE cannot be upheld (at any rate in its natural interpretation). For an Z-statement can certainly be false, even if the D-or-D condition is unfulfilled. Suppose that I am comforted in normal daylight, by a perfectly normal pillar-box; suppose further that I am in the presence of a normal, unsceptical companion; both he and I know perfectly well that the pillar-box is red. However, unknown to him, I suffer chronically from Smith's Disease, attacks of which are not obvious to another party; these attacks involve, among other things perhaps, the peculiarity that at the time red things look some quite different colour to me. I know that I have this disease, and I am having (and know that I am having) an attack at the moment. In these circumstances I say, " That pillar-box looks red to rne ". I would suggest that here the D-or-D condition is not fulfilled; my companion would receive my remark with just that mixture of puzzlement and scorn which would please my objector; and yet when he learnt about my attack of Smith's Disease, he would certainly think that what I had said had been false. At this point it might perhaps be suggested that though I have succeeded in producing an example of an l-statement which would be false, I have not succeeded in producing an example of an Z-statement which is false when the D-or-D condition is unfulfilled; for in fact the D-or-D condition is fulfilled. For the speaker in my little story, it might be said, has some reason to doubt whether the pillar-box before him is red, and this is enough to ensure the fulfilment of the condition, even though the speaker also has information (e.g., that this is the pillar-box he has seen every day for years, and that it hasn't been repainted and so on) which enables him entirely to discount this prima facie reason for doubt. But this will not do at all. For what is this prima facre reason for doubting whether the pillar-box really is red ? If you like, it is that it looks blue to him. But this is an unnecessarily specific description of his reason; its looking blue to him only counts against its being really red because its looking blue is a way of failing to look red; there need be nothing specially important about its looking blue as distinct from looking any other colour, except red. So this rescueattempt seems to involve supposing that one way of fulfilling the precondition of an l-statement's having a truth value at all, THE CAUSAL THEORY OF PERCEPTION 135 consists in its having the truth-value F, or at least in some state of affairs which entails that it has the truth-value F. But surely, a statement should be false cannot be onc way of fulfilling a pre-condition of that statement's having a trutl-r-valuc; the mere fulfihnent of a pre-condition of a statement's having a trutl.r-value ought to leave it open (to be decided on other grounds) which truth-value it has. Let us assume that this rear-guard action has been disposed of. Then it is tempting to argue as follows: Since the objector can no longer maintain that fulfilment of the D-or-D condition is a pre-requisite of an l-statement's having a truth-value, he will have to admit that fulfilment is al most a partial truthcondition albeit of a special kind (i.e., is one of the things which have to be the case if the statement is to be true). It cannot be the only truth-condition, so there must be another truthcondition; indeed we can say what this is in the light of the preceding argument; it consists in the non-fulfi.lment of the statement's falsity-condition or falsity-conditions (which have just been shown to be independent of the D-or-D condition); to put it less opaquely, it consists in there being nothing to make the l-statement false. But now, it may be thought, all is plain sailing for the sense-datum theorist; he can simply lay down that a sense-datum sentence is to express a truth if and only if the second truth-condition of the corresponding l-statement is fulfilled, regardless of whether its flrst truth condition (the D-or-D condition) is fulfilled. It will be seen that the idea behind this argument is that, once the objector has been made to withdrap the contention that the fulfilment of the D-or-D condition is a condition of an Z-statement's having a truth-value, he can be forced to withdraw also the contention that the implication that the D-or-D condition is fulfilled is non-detachable; and this destroys his position. So far so good, perhaps, but unfortunately not yet good cnough. For the objector has a powerful-looking reply at his tlisposal. He may say: " Once again you are covertly begging thc question. You are assuming, quite without justification, that bccause one can, in some sense, distinguish the second truthcondition from the first, it is therefore the case that the implication ol' the fulfilment of the first (D-or-D) condition is detachable; 136 H. P. GRICE that is, that there must be a way of specifying the second condition 'which does not carry the implication that thb first condition is fulfilled. But your argument has certainly not proved this conclusion. Consider a simple parallel: it is perfectly obvious that objects which are not vermilion in colour may or may not be red; so being red is not a necessary falsity-condition of being vermilion. It is also true that being red is only a partial truthcondition of being vermilion if what this means is that to establish that something is red is not enough to establish that it is vermilion. But it does not follow (and indeed it is false) that there is any way of forrnulating a supplementary truth-condition for an object's being vermilion which would be free frorn the implication that the object in question is red. Tlis non sequitur is very much the same as the one of which you are guilty; the fulfilment of the D-or-D condition may pcrfcctly well be only a truth-condition of an l-statement, and only one of a pair of truth conditions at that, without its being the case that the implication of its fulfllment is detachable." He may also add the following point: " Though the contention that the fulfilment of the D-or-D condition is a pre-condition of the truth or falsity of the corresponding Zstatement cannot be upheld under the interpretation which you have given to it, it can be upheld if it is given another not unnatural interpretation. I cannot, in view of your counterexample, maintain that for an Z-statement to be true, or again for it to be false. the D-or-D condition must be fulfilled. But I can nraintain that the D-or-D condition's fulfilment is a condition of truth or falsity of an l-statement in the following sense, namely that if the D-or-D condition is fulfilled, then Z and F are the two possibilities between which, on other grounds, the decision lies (i.e., N is excluded): whereas if the D-or-D condition is nol fulfilled, then one has to decide not between these possibilities, but between the possibilities ff and F (i.e., I is excluded.) " This onslaught can I think be met, though at the cost of some modification to the line of argument against which it was directed. I think that the following reply can be made: " There is a crucial difference between the two cases which you treat as parallel. Let us endeavour to formulate a supplementary truth condition for the form of statement ' x is vermilion '; we might THE CAUSAL THEORY OF PERCEPTION suggest the condition that x has the feature which differentiates vermilion things from other red things. But to suppose that x satisfies this condition, but does not satisfy the first truthcondition, namely, that ;r should be red, would be to commit a logical absurdity; x cannot logically differ from red things which are not vermilionin just tl-re way in which vermilion things differ frorn red things which are not vermilion, without being red. Conscquently one cannot assert, in this case, that the second truth-condition is fulfilled without its being implied that the first is fulfllled, nor can one go on to cancel this implication. But in the case of an Z-statement there is no kind of logical implication between the second truth-condition and the first. For one thing, if there were such a logical connexion, there would also have to be such a logical connexion between the I-statement itself and the fulfilment of the D-or-D condition; and if this were so, the implication that the D-or-D condition is fulfrlled would have to be carried by what wss said or asserted by the utterance of an l-statement. But that this is not so can be seen from the unacceptability of such an hypothetical as ' If this pillar-box looks red to me, then I or someone else is, or miglrt be, inclined to deny that it is rcd or to doubt whether it is rcd.' For another thing, it is surely clear that if I were now to say ' Nothing is the case which would make it false for me to say that the palm of this hand looks pink to me, though I do not rncan to imply that I or anyone else is or might be inclined to dcny that, or doubt whether, it is pink' this would be a perfectly inteliigible remark even though it might be thought both wordy aud boring. Indeed I am prepared actually to say it. Conscquently, although you may be right in claiming that it has not bccrr shown that the implication of the futfrlment of the D-or-D condition is detachable (and indeed it may well be non-detachable), you must be wrong in thinking that this implication is not cancellable. Admittedly there is at least one case in which an irrrplication which is not logical in character is at least in a scnsc, non-cancellable; we found one in considering example (2) ' Slrc was poor but she was honest.' But if we look a little more closcly we can see that the reason why the implication here is, in l scnse, not cancellable is just that it rs detachable (by the use of ' ancl'). More fully the reason why it would be peculiar to say 137 138 H. P. GRICE 'She was poor but she was honest, though I do not mean to imply that there is any contrast' is that any one who said this would have first gone out of his way to find a form of words which introduced the implication, and then would have gone to some trouble to take it out again. Why didn't he just leave it out? The upshot is, that if you say that the implication of the fulfilment of the D-or-D condition is (a) not logical in character and (b) not detachable, then you must allow that it is cancellable. And this is all that the sense-datum theorist needs." If there is an answer to this argument, I do not at present know what it is. I will conclude by making three auxiliary points. (l) If I am right in thinking that my objector has gone astray, then I think I can suggest a possible explanation of his coming to make his mistake. His original resistance to attempts to distinguish between the facts stated by an Z-statement and the fulfilment of the D-or-D condition arose I think from a feeling that if the D-or-D condition were unfulfilled there would be no facts to state; and this feeling is I suspect the result of noticing the bafling character that the utterance of an Z-statement would have in certain circumstances. But precisely what circumstances ? I think the sort of imaginary example the objector has in mind may be the following: I and a companion are standing in front of a pillar-box in normal daylight. Each of us has every reason to suppose that the other is perfectly normal. In these circumstances he says out of the blue " This pillar-box looks red to rne " and (it is assumed) I am not allowed to take this as a joke. So I am bafled. I do not know what to make of his utterance. But surely the reason why I am baffied is that I cannot see what communication-function he intends his utterance to fulfil; it has the form of an utterance designed to impart information, but what information could he possibly imagine would be imparted to me which I do not already possess. So of course this utterance is baffiing. But what the objector may not have noticed is that if in these circumstances my companion had said not " This pillar box looks red to me " but " This pillar box is red ", his utterance would have been equally baffiing, if not more bafling. My point can be stated more generally. The I THE CAUSAL TTIEORY OF PERCEPTION objector wants to attribute to Z-statements certain special features (e.g., that of being neither T nor F in certain circumstances) which distinguish them from at lcast some other statements. lf so, he cannot derive support for his thesis from the fact that the utterance of an l-statement would be baffiing in certain circunrstances, when those circumstances are such that (mutatis mutandis) they would make any stqtement whatever baffing. He ought to take as his examples not Z-statements made about objects which both speaker and audience can see perfectly clearly, but "L-statements made about objects which the speaker can see but the audience cannot. But when the examples are thus changed, his case seems much less plausible. (2) If I am asked to indicate what it would be righttosayabout Z-statements and the implications involved in these utteranccs, I shall answer: very much the same sort of thing as I have earlier in this page suggested as regards disjunctive statements. I don't want to duplicate my earlier remarks, so I will deal with this very briefly. (i) The fulfilment of the relevant D-or-D condition is not a condition either of the truth or of the falsity of an Z-statement, though if this condition is not fulfilled the utterance of the l-statement may well be extremely misleading (in its implication). (ii) Like my examples (3) and (4) above, we may speak either of the speaker or of his saying what he did say as vehicles of the implication; the second of these possibilities is important in that, if I am right about it, it leads to point (iii). (iii) The implication is not detachable in my official sense. For if the implication can be regarded as being carried by his saying that (rather than something else), e.g., his mentioning this fact or putative fact rather than some other fact or putative fact, then it seems clear that any other way of stating the same fact or putative fact would involve the same implication as the original way of stating the fact in question. (iv) Comparably with examples (3) and (4), the implication is detachable in the further possible non-official sense which I referred to earlier in connexion with (4); there will be some conditions of utterance irr which the implication is no longer carried, e.g., if I am talking to my oculist about how things look to me. (v) The implication is cancellable (I need say no more about this). (vi) As in the case of example (4), the reason why the implication is standardly 139 140 H. P. GRICE carried is to be found in the operation of some such general principle as that giving preference to the making of a stronger rather than a weaker statement in the absence of a reason for not so doing. The implication therefore is not of a part of the meaning of the expression " looks to me ". There is however here an important difference between the case of Z-statements and that of disjunctives. A disjunctive is weaker than either of its disjuncts in a straightforward logical sense, namely, it is entailed by, but does not entail, each of its disjuncts. The statement " It looks red to me " is not, however, weaker than the statement " It is red" in just this sense; neither statement entails the other. I think that one has, neverthcless a strong inclination to regard the first ofthese statements as weaker than the second; but I shall not here attempt to determine in what sense of " wgaker " this may be true. (3) The issue with which I have been mainly concerned may be thought rather a fine point, but it is certainly not an isolated one. There are several philosophical theses or dicta which would I think need to be examined in order to see whether or not they are sufficiently parallel to the thesis which I have been discussing to be amenable to treatment of the same general kind. Examples which occur to me are the following: (1) You cannot see a knife as a knife, though you may see what is not a knife as a knife. (2) When Moore said he knew that tl-re objects before him were human hands, he was guilty of misusing the word "know". (3) For an occurrence to be properly said to have a cause, it must be something abnormal or unusual. (4) For an action to be properly described as one for which the agent is responsible, it must be thc sort of action for which people are condemned. (5) What is actual is not also possible. (6) What is known by me to be the case is not also believed by me to be the case. I have no doubt that there will be other candidates besides the six which I have mentioned. I must emphasize that I am not saying that all these examples are importantly similar to the thesis which I have been criticizing, only that, for all I know, they may be. To put the matter more generally, the position adopted by my objector seems to me to involve a type of manoeuvre which is characteristic of more than one contemporary mode of philosophizing. I am not condemning THE CAUSAL TIIEORY OF PERCEPTION I4I this kind of manoeuvre; I am merely suggesting that to embark on it without due caution is to risk collision with the facts. Before we rush ahead to exploit the linguistic nuances which we have detcctcd, we should make sure that we are reasonably clcar what sort of nuanccs they are. 5 I hopc that I may have succeeded in disposing of what I have found to be a frequently propounded objection to the idea of explaining thc notion of a sense-datum in terms of some member or members of the suggested family of locutions. Further detailed work would be needed to find the most suitable member of the family, and to select the appropriate range of uses of the favoured member when it is found; and, as I have indicated, neither of these tasks may be easy. I shall, for present purposes, assume that some range of uses of locutions of the form " It looks (feels, etc.) to X as if " has the best chance of being found suitable. I shall furthermore assume that the safest procedure for the Causal Theorist will be to restrict the actual occurrences of the term " sense-datum " to such classificatory labels as " sense-datum statement " or " sense-datum sentence "; to license the introduction of a " sense-datum terminology " to be used for the re-expression of sentences incorporating the preferred locutions seems to me both unnecessary and dangerous. I shall myself, on behalf of the CTP, often for brevity's sake talk of sense-data or sense-impressions; but I shall hope that a more rigorous, if more cumbrous, mode of expression will always be readily available. I hope that it will now be allowed that, interpreted on the lines which I have suggested, the thesis that perceiving involves having a sense-datum (involves its being the case that some sense-datum statement or other about the percipient is true) has at least a fair chance of proving acceptable. I turn now to the special features of the CTP. The flrst clause of the formulation quoted aboves from Price's Perception rnay be interpreted as representing it to be a necessary and sulficient condition of its being the case that X perceives M that X's sense-impression should be causally dependent on some tt P. 122 suora. 142 H. P. GRICE state of affairs involving M. Let us first enquire whether the suggested condition is necessary. Suppose that it looks to X as if there is a clock on the shelf; what more is required for it to be true to say that X sees a clock on the shelf ? There must, one might say, actually be a clock on the shelf which is in X's field of view, before X's eyes. But this does not seem to be enough. For it is logically conceivable that there should be some method by which an expert could make it look to X as if there were a clock on the shelf on occasions when the shelf was empty: there might be some apparatus by which X's cortex could be suitably stimulated, or some technique analogous to post-hypnotic suggestion. If such treatment were applied to X on an occasion when there actually was a clock on the shelf, and if X's impressions were found to continue unchanged when the clock was removed or its position altered, then I think we should be inclined to say that X did not see the clock which was before his eyes, just because we should regard the clock as playing no part in the origination of his impression. Or, to leave the realm of fantasy, it might be that it looked to me as if there were a certain sort of pillar in a certain direction at a certain distance, and there might actually be such a pillar in that place; but if, unknown to me, there were a mirror interposed between myself and the pillar, which reflected a numerically different though similar pillar, it would certainly be incorrect to say that I saw the flrst pillar, and correct to say that I saw the second; and it is extremely tempting to explain this linguistic fact by saying that the first pillar was, and the second was not, causally irrelevant to the way things looked to me. There seems then a good case for allowing that the suggested condition is necessary; but as it stands it can hardly be sufficient. For in any particular perceptual situation there will be objects other than that which would ordinarily be regarded as being perceived, of which some state or mode of functioning is causally relevant to the occurence of a particular sense-impression: this might be true of such objects as the percipient's eyes or the sun. So some restriction will have to be added to the analysis of perceiving which is under consideration. Pricea suggested that use should be made of a distinction between " standing " and a Op. cit., p.70, THE CAUSAL THEORY OF PERCEPTION "differential" conditions: as the state of the sun and of the percipient's eyes, for example, are standing conditions in that (roughly speaking) if they were suitably altered, all the visual impressions of the percipient would be in some respect different from what they would otherwise have been; whereas the state of the perceived object is a differential condition in that a change in it would afect only some of the percipient's visual impressions, perhaps only the particular impression the causal origin of which is in question. The suggestion then is that the CTP should hold that an object is perceived if and only if sorne condition involving it is a diferential condition of some sense-impression of the percipient. I doubt, however, whether the imposition of this restriction is adequate. Suppose that on a dark night I see, at one and the same time, a number of objects each of which is illuminated by a different torch; if one torch is tampered with, the effect on my visual impressions will be restricted, not general; the objects illuminated by the other torches will continue to look the same to me. Yet we do not want to be compelled to say that each torch is perceived in such a situation; concealed torches may illuminate. But this is the position into which the proposed revision of the CTP would force us. I am inclined to think that a more promising direction for the CTP to take is to formulate the required restriction in terms of the way in which a perceived object contributes towards the occurrence of the sense-impression. A conceivable course would be to introduce into the speciflcation of the restriction some part of the specialist's account, for example to make a lclerence to the transmission of light-waves to the retina; but thc objection to this procedure is obvious; if we are attempting to characterize the ordinary notion of perceiving, we should not cxplicitly introduce material of which someone who is perfectly capable of employing the ordinary notion might be ignorant. I suggest that the best procedure for the Causal Theorist is to indicate the mode of causal connexion by examples; to say that, lirl an object to be perceivedby X, it is suffrcient that it should be e:rusitlly involved in the generation of some sense-impression by .\' in thc kind of way in which, for example, when I look at my lr:rncl in a good light, my hand is causally responsible for its kroking at n1e as if there were a hand before me, or in which . . . 143 144 H. P. GRICE (and so on), whatever that kind of way may be; and to be enlightened on that question one must have recourse to the specialist. I see nothing absurd in the idea that a non-specialist concept should contain, so to speak, a blank space to be filled in by the specialist; that this is so, for example, in the case of the concept of seeing is perhaps indicated by the consideration that if we were in doubt about the correctness of speaking of a certain creature with peculiar sense-organs as seeing objects, we might well wish to hear from a specialist a comparative account of the human eye and the relevant sense-organs of the creature in question. We do not, of course, ordinarily need the specialist's contribution; for we may be in a position to say that the same kind of mechanism is involved in a plurality of cases without being in a position to say what that mechanisrrr is.5 At this point an objcction must be mentioned with which I shall deal only briefly, since it involves a manoeuvre of the same general kind as that which I discussed at length earlier in this paper. The CTP as I have so expounded it, it may be said, requires that it should be linguistically correct to speak of the causes of sense-impressions which are involved in perfectly normal perceptual situations. But this is a mistake; it is quite unnatural to talk about the cause, say, of its looking to X as if there were a cat before him unless the situation is or is thought to be in some way abnormal or delusive; this being so, when a cause can, without speaking unnaturally, be assigned to an impression, it will always be something other than the presence of the perceived object. There is no natural use for such a sentence as " The presence of a cat caused it to look to X as if there were a cat before him "; yet it is absolutely essential to the CTP that there should be. In reply to this objection I will make three points. (l) If we are to deal sympathetically with the CTP we must not restrict the Causal Theorist to the verb ' cause '; we must allow him to make use of other mernbers of the family of causal verbs or slt might be thought thar we need a further restriction, limiting the permissible degree of divergence between the way things appear to X and the way they actually are. But objects can be said to be seen even when they are looked at through rough thick glass or distorting spectacles, in spite of the fact that they may then be unrecognizable. THE CAUSAL THEORY OF PERCEPTION verb-phrases if he wishes. This family includes such expressions as " accounts for ", " explains ", " is part ofthe explanation of", " is partly responsible for ", and it seens quite possible that some alternative formulation of the theory would cscape this objection. (2) If I regard myself as being in a position to say "There is a cat", of " I see a eat",I naturally rcfrain from making the weaker statement " It looks to me as if there were a cat before me ", and so, a fortiori, I refrain from talking about the cause of its looking to me thus. But, if I was right earlier in this paper, to have made the weaker statement would have been to have said something linguistically correct and true, even if misleading; is there then any reason against supposing that it could have been linguistically correct and true, even if pointless or misleading, to have ascribed to a particular cause the state of affairs reported in the weaker statement? (3) X is standing in a street up which an elephant is approaching; he thinks his cyes must be deceiving him. Knowing this, I could quite naturally say to X, " The fact that it looks to you as if there is an elephant approaching is accounted for by the fact that an elephant is approaching, not by your having become dcranged." To say the same thing to one's neighbour at the circus would surely be to say something which is true, though it might be regarded as provocative. I have extracted from the first clause of the initial formulation ol' the CTP an outline of a causal analysis of perceiving which is, I hope, at least not obviously unacceptable. I have of course oonsidered the suggested analysis only in relation to seeing; a ntore careful discussion would have to pay attention to nonvisual perception; and even within the field of visual perception tlrc suggested analysis might well be unsuitable for some uses of llrr: word'see', which would require a stronger condition than llrat proposed by the theory. 6 Is the CTP, as so far expounded, open to the charge that it lt';'r'cscnts material objects as being in principle unobservable, and irr ctlrrscclucnce leads to scepticism about the material world? I Ir:rve sornc difliculty in understanding the precise nature of the 145 146 H. P. GRICE accusation, in that it is by no means obvious what, in this context, is meant by " unobservable ". (l) It would be not unnatural to take " unobservable " to mean " incapable of being perceived ". Now it may be the case that one could, without being guilty of inconsistency, combine the acceptance of the causal analysis of perceiving with the view that material objects cannot in principle be perceived, if one were prepared to maintain that it is in principle impossible for material objects to cause sense-imprcssions but that this irnpossibility has escaped the notice of common sense. This position, even if internally consistent, would seem to be open to grave objection. But even if the proposition that material objects cannot be perceived is consistent with the causal analysis of perceiving, it certainly does not appcar to be a consequence of the latter; and the exposition of the CTP has so far been confined to the propounding of a causal analysis of perceiving. (2) The critic might be equating " unobservable " with " not directly observable "; and to say that material objects are not directly observable might in turn be interpreted as saying that statements about material objects lack that immunity from factual mistake which is (or is supposed to be) possessed by at least some sense-datum statements. But if " unobservable " is thus interpreted, it seems to be true that material objects are unobservable, and the recognition of this truth could hardly be regarded as a rnatter for reproach. (3) " Observation " rnay be contrasted with " inference " as a source of knowledge and so the critic's claim may be that the CTP asserts or implies that the existence of particular material objects can only be a matter of inference. But in the first place, it is not established that the acceptance ofthe causal analysis ofperceiving commits one to the view that the existence of particular material objects is necessarily a matter of inference (though this view is explicitly asserted by the second clause of Price's initial formulation of the CTP); and secondly, many of the critics have been phenomenalists, who would themselves be prepared to allow that the existence of particular material objects is, in some sense, a matter of inference. And if the complaint is that the CTP does not represent the inference as being of the right kind, then it looks as if the critic might in effect be complaining that the TIIE CAUSAL THEORY OF PERCEPTION Causal Theorist is not a Phenomenalist. Apart from the fact that the criticism under discussion could now be made only by someone who not only accepted Phenomenalism but also regarded it as the only means of deliverance from scepticism, it is by no means clear that to accept a causal analysis of perceiving is to debar oneself from accepting Phenomenalisnr; tlrere seelns to be no patent absurdity in the idea that one could, as a first stage, offer a causal analysis of ' X perceives M', and then re-express the result in phenomenalist terms. If the CTp is to be (as it is often regarded as being) a rival to Phenomenalism, the opposition may well have to spring from the second clause of the initial formulation of the theory. There is a further possibility of interpretation, related to the previous one. If someone has seen a speck on the horizon which is in fact a battleship, we should in some contexts be willing to say that he has seen a battleship; but we should not, I think, be willing to say that he has observed a battleship unless l.re has recognized what he has seen as a battleship. The criticism lcvelled at the CTP may then be that it asserts or entails the irnpossibility in principle of knowing, or even of being reasonably assured, that one is perceiving a particular material object, even if one is in fact perceiving it. At this point wc must direct our attention to the second clause of the initial fonnulation of the CTP, which asserted that " perceptual consciousness is fundarncntally an inference from effect to cause ". I shall assume (l hope not unreasonably) that the essence ofthe view here being advanced is that anyone who claims to perceive a particular rttatcrial object M may legitimately be asked to justify his claim; attd that the only way to meet this demand, in the most fundatttcntal type of case, is to produce an acceptable argument to tfrc cffect that the existence of M is required, or is probably rcrprired, in order that the claimant's culretlt sense-impressions sltor.rld be adequately accounted for. A detailed exposition of thc CTP may supplement this clause by supplying general lttirrciplcs which, by assuring us of correspondences bctween ('iruscs and effects, are supposed to make possible the production ol' sirtisf actory arguments of the required kind. It is clcar that, if the Causal Theorist proceeds on the lines rvlrich I lravc just indicated, he cannot possibly be accusecl of t47 rc2 148 H. P. GRICE having asserted that material objects are unobservable in the sense under consideration; for he has gone to some trouble in an attempt to show how we may be reasonably assured of the existence of particular material objects. But it may be argued that (in which is perhaps a somewhat special sense of " consequence ") it is an unwanted consequence of the CTP that material objects are unobservable: for if we accept the contentions of the CTP (1) that perceiving is to be analysed in causal terms, (2) that knowledge about perceived objects depends on causal inference, and (3) that the required causal inferences will be unsound unless suitablc gcneral principles of correspondence can be provided, then wc shall havc to admit that knowledge about perceived objects is unobtainable: for the general principles offered, apart frorn beirrg dubious both in respect of truth and in respect of status, fail to yield the conclusions for which they are designed; and more successful substitutes are not available. If this is how the criticism of the CTP is to be understood, then I shall not challenge it, though I must confess to being in some doubt whether this is what actual critics have really meant. My comment on the criticism is now that it is unsympathetic in a way that is philosophically important. There seem to me to be two possible ways of looking at the CTP. One is to suppose an initial situation in which it is recognized that, while appearance is ultimately the only guide to reality, what appears to be the case cannot be assunred to correspond with what is the case. The problem is conceived to be that of exhibiting a lcgitimate method of arguing from appearance to reality. The CTP is theu regarded as a complex construction designed to solvc tlris problem; and if one part of the structure collapses, the renrainder ccases to be of much interest. The second way of looking at the CTP is to think of the causal analysis of perceiving as sonrething to be judged primarily on its intrinsic merits and not merely as a part of a solution to a prior epistemological problem, and to recognize that some version of it is quite likely to be correct; the remainder of the CTP is then regarded as consisting (l) of steps which appear to be forced upon one if one accepts the causal analysis of perceiving, and which lead to a sceptical difficulty, and (2) a not very successful attempt to meet this difficulty. This way of looking at the CTP THE CAUSAL THEORY OF PERCEPTION recognizes the possibility that we are confrontcd with a case in which the natural dialcctic elicits distressing conscquences (or rather apparent consequences) front truc propositious. To adopt the first attitude to tl-re exclr-rsion of thc sccond is both to put on one side what uray well bc an acccptablc bit of philosophical analysis and to neglect what might bo ar.r opportunity for deriving philosophical profit from the exposurc of opcrations of the natural dialectic. This, I suggest, is what the critics have tended to do; though, no doubt, they might plead historical justification, in that the first way of looking at the CTP may have been that of actual Causal Theorists. It remains for me to show that the CTP can be looked upon in the second way by exhibiting a line of argument, sceptical in clraracter, which incorporates appropriately the elements of the CTP. I offer the following example. In the fundamental type of case, a bona fide claim to perceive a particular material object M is based on sense-datum statements; it is only in virtue of the occurrence of certain sense-impressions that the claimant would rcgard himself as entitled to assert the existence of M. Since the causal analysis of perceiving is to be accepted, the claim to pcrceive -M involves the claim that the presence of M causally cxplains the occurrence of the appropriate sense-impressions. 'l'hc combination of these considerations yields the conclusion tlrat the claimant accepts the existence of M on the grounds that it is required for the causal explanation of certain senseirnpressions; that is, the existence of M is a matter of causal irrlbrence from the occurrence of the sense-impressions. Now a nrtlclcl case of causal inference would be an inference from smoke to fire; the acceptability of such an inference involves the possibility of establishing a correlation between occurrences ol'srnoke and occurrences of fire, and this is only possible because llrcrc is a way of establishing the occurrence of a fire otherwise llrrur by a causal inference. But there is supposed to be no way ol'cstablishing the existence of particular material objects except hy :r causal inference from sense-impressions; so such inferences t :r n not bc rationally justifled. The' specification of principles of tollcspondence is of course an attempt to avert this consequence Iry rc.jccting the smoke-fire model. [If this model is rejected, r('('()ut'sc ntay be had to an assimilation of material objects to I4y 150 H. P. GRICE such entities as electrons, the acceptability of which is regarded as being (roughly) a matter of their utility for the purposes of explanation and prediction; but this assimilation is repugnant for the reason that material objects, after having been first contrasted, as a paradigm case of uninvented entities, with the theoretical constructs or entia rationis of the scientist, are then treated as being themselves entia rationis.l One possible reaction to this argument is, of course, " So much the worse for the causal analysis of perceiving "; but, as an alternative, the argument itself may be challenged, and I shall conclude by mentioning, without attempting to evaluate, some ways in which this might be done. (l) It may be argued that it is quite incorrect to describe rnany of my perceptual beliefs (e.g., that there is now a table in front of me) as " inferences " of any kind, if this is to be taken to imply that it would be incumbent upon me, on demand, to justify by an argument (perhaps after acquiring further data) the contention that what appears to me to be the case actually is the case. When, in normal circumstances, it looks to me as if there were a table before me. I am entitled to say flatly that there is a table before me, and to reject any demand that I should justify my claim until specific grounds for doubting it have been indicated. It is essential to the sceptic to assume that any perceptual claim may, without preliminaries, be put on trial and that innocence, not guilt, has to be proved; but this assumption is mistaken. (2) The allegedly 'fundamental' case (which is supposed to underlie other kinds of case), in which a perceptual claim is to be establishable purely on the basis of some set of sense-daturn statements, is a myth; any justification of a particular perceptual claim will rely on the truth of one or more further propositions about the material world (for example, about the percipient's body). To insist that the ' fundamental' case be selected for consideration is, in effect, to assume at the start that it is conceptually legitimate for me to treat as operl to question all my beliefs about the material world at once; and the sceptic is not entitled to start with this assumption. (3) It might be questioned whether, given that I accept the existence of M on the evidence of certain sense-impressions, and given also that I think that M is causally responsible for those sense-impressions it follows that I accept the existence of M on the grounds that THE CAUSAL TIIEORY OF PERCEPTION its existence is required in order to account for the senseimpressions. (4) The use made of the smoke-fire model in the sceptical argument might be criticized on two different grounds. First, if the first point in this paragraph is well made, there are cases in which the existence of a perceived object is not the conclusion of a causal inference, namely those in which it cannot correctly be described as a matter of inference at all. Secondly, the model should never have been introduced: for whereas the proposition that fires tend to cause smoke is supposedly purely contingent, this is not in general true of propositions to the effect that the presence of a material object possessing property P tends to (or will in standard circumstances) make it look to particular persons as if there were an object possessing P. It is then an objectionable feature of the sceptical argument that it first treats non-contingent connexions as if they were contingent, and then complains that such connexions cannot be established in the manner appropriate to contingent connexions. The ruon-contingent character of the proposition that the presence of a red (or round) object tends to make it look to particular people as if there were something red (or round) before them does not, of course, in itself preclude the particular fact that it looks to nre as if there were something red before me from being explained by the presence of a particular red object; it is a non-contingent nratter that corrosive substances tend to destroy surfaces to which thcy are applied; but it is quite legitimate to account for a particular case of surface-damage by saying that it was caused by some corrosive substance. In each case the effect might have corrrc about in some other wav. 7 I conclude that it is not out of the question that the following vcrsion of the CTP should be acceptable: (1) It is true that X lrclccives M if, and only if, some present-tense sense-datum slltcrncnt is true of X which reports a state of affairs for which ,4/, in a way to be indicated by example, is causally responsible, irrrtf (2) a clairn on the part of Xto perceive M,lf it needs to be irrstiliccf at all, is justified by showing that the existence of M is r ct;uirccl ii the circumstances reported by certain true r5l r52 H. P. GRICE sense-datum statements, some of which may be about persons other than X, are to be causally accounted for. Whether this twofold thesis deserves to be called a Theory of Perception I shall not presume to judge; I have already suggested that the first clause neither obviously entails nor obviously conflicts with Phenomenalism; I suspect that the same may be true of the second clause. I am conscious that my version, however close to the letter, is very far from the spirit of the original theory; but to defend the spirit as well as the letter would be beyond my powers.