268 33 2MB
English Pages 144 Year 1973
FORM AND CONTENT
BERNARD HARRISON
Published in th e U.S.A . 1973 by H A R P E R & R O W P U B L IS H E R S . IN C . B A R N E S & N O B L E IM P O R T D IV IS IO N ISBN 06-492736-9 All R ights Reserved. N o p a rt o f this p u b lication m ay be reproduced, stored in a retrieval system , o r transm itted, in any form o r by an y m eans, electronic, m echanical, photocopying, reco rd in g o r otherw ise w ithout th e p rio r perm ission o f Basil Blackwell & M ott Lim ited.
P rinted in G re a t B ritain
It is evident, th at even different simple ideas may have a sim ilarity o r resem blance to each other; nor is it nccessary that the point o r circum stance of resemblance should be distinct o r separable from that in which they differ. Blue and green are different simple ideas, but are m ore resem bling than blue and scarlet; though their perfect simplicity excludes all possibility of separation o r distinction. It is the same case with particular sounds» and tests and smells. These adm it of infinite resem blances upon the general appearance and com parison, w ithout having an y com m on circum stance the same. A nd of this we m ay be certain, even from the abstract term s simple idea. They com prehend all simple ideas under them . These resem ble each other in their simpli city. A nd yet from their very nature, which excludes all com position, this circum stance, in which they resem ble, is not distinguishable or separable from the rest. It is the same case with all the degrees in any quality. They are all resembling, and yet the quality, in any individual, is not distinct from the degree. David Hum e, Treatise of Human Nature , Bk 1, P art 1, V II, footnote
T h e senses have . . . become theoreticians immediately in their praxis. M arx, Economic and Philosophic Manuscripts (1844)
Contents
Preface 1 N am ing and discrimination
xi 3
2 Schlick on form and content: a Tractarian excursion
33
3 A model o f colour naming
53
4 The limits o f scepticism
93
Appendix
143
Index
145
Preface
T he chief influences on this book are, as will be obvious, W ittgenstein— the early W ittgenstein as m uch as the late—and N oam Chom sky. 1 have profited as well from the conversation of m any teachers and colleagues, including most recently T . L. S. Sprigg, A. D. N uttall and G abriel Josipovici. 1 owe a particular d ebt of gratitude to A aron Sloman and John M epham , who read an earlier draft of the m anuscript and discussed it with me a t great length. If the book is less obscure and groping than it was in that version the credit is mainly theirs. 1 am also grateful to the University of Sussex for freeing me from teaching an d m ost adm inistrative duties in the Summer T erm of 1970-1, during which m ost of the first draft was written. A trial run of some of the ideas put forw ard here appears in Inquiry for 1967 under the title ‘On Describing C olours’. A m uch fuller exposition of the theory of meaning presupposed by m uch of w hat I say here will be found in my book M eaning and Structure, published by H arper & Row. Kingston near Lewes, Sussex July 1972
PART I
Naming and Discrimination
1. Naming and Discrimination i It is often held th at language can express only the form but not the content of experience. I shall call this thesis ‘ the thesis of the inexpressibility of content’ o r ‘the inexpressibility thesis’ for short. I shall argue th at on both the interpretations of it that are current in contem porary philosophy1 it is false, and that its falsity o n these interpretations has im portant epistemological consequences, particularly in connexion with scepticism about the content of the experience enjoyed by other minds. T he two versions of the thesis that I have in m ind are those put forward respectively by M oritz Schlick, in a discussion which is still, to m y m ind, the best on the topic in the literature of analytic philosophy2 and m ore recently, by a num ber o f writers includ ing J. J. C. S m art5 an d B. A. F arrell.4 1 shall discuss the latter set of argum ents first. 1 Contem porary discussions have, I think, genuinely superseded traditional ones on this and related topics, so that we need not refer back, for exam ple to Descartes’ and G alileos’ theories o f primary and secondary qualities or to L ocke’s distinction between sim ple and com plex ideas; although it w ill help if the reader is able to bear in mind throughout what follow s these and other philosophical antecedents o f the inexpressibility thesis, not least because they help one to understand the general nature o f the meta physical im pulses which have repeatedly led philosophers to advance such a thesis in its various historical form s and variants. 2 M oritz Schlick, ‘Form and Content, an Introduction to Philosophical Thinking' in G esa m m elte A n fsa ize , 1926-1936, G erold and Co., Wien, 1938. * J. J. C. Smart, ‘Sensations and Brain Processes’, P hilosophical R eview , LX V III (1959), 141-56; ‘C olours’, P h ilo so p h y, X X X V I (1961), 128-142. 4 B. A. Farrell, 'Experience’, M in d , M X (1950), 170-98.
4
NAMING AND D IS C R IM IN A T IO N
II A ccording to Sm art and Farrell, the m ost fundam ental level of discourse about experience is the level at which we state the facts, so far as they are known to us, about our own and other's ability to discrim inate between particular types of pre sented stimuli. If we ask w hether there is not some more funda mental level at which we can describe the phenom enal character of the experiences between which we discrim inate, the reply must be that there is not. Experience talk boils down ultim ately to discrim ination talk. Let us call this ‘the discrim ina tion thesis’. S m art’s and F arrell's ground for holding the discrim ination thesis differ som ew hat. Sm art's argum ent, if I understand him correctly, is th at acceptance of the thesis must follow acceptance of the doctrine that m eaning is ‘use’, and not ‘a m ental ex perience which evokes and is evoked by a w ord’,5 once we fully understand w hat is involved in the latter doctrine. It is evident, Sm art thinks, th at a blind m an who had the oppor tunity to study how norm ally sighted people sorted sam ples identified by physical and chem ical criteria, and how they applied colour w ords to the resulting sets of samples, would eventually com e to use colour words exactly as the sighted sorters use them. T h at is, he would use ‘orange’, for exam ple, to mean something like ‘resem bling anything which a norm ally sighted person would, under norm al conditions of illum ination, sort into the same com partm ent as oranges, fashionable Swedish iron casseroles, . . and so on. We can still say, perhaps, that the blind m an does not ‘know w hat colours are’, but this only m eans that he cannot use him self as a ‘norm al percipient’, o r to put it bluntly cannot see, which is trivially true. Will this do? T h e bald assertion that ‘M eaning is u se ’ no 5 ‘Colours*, p. 141.
NAMING AND DISCRIM INATION
5
longer counts, it seems to m e, as the expression of a philo sophical insight of the first w ater, even when the word ‘use’ is prudently enclosed, as Sm art encloses it, in inverted commas. Before anything can be m ade of such a claim one needs to have a plain, preferably non-m etaphorical, account of w hat the speaker understands by ‘use’, an d this Sm art does not provide. W hen he says that the blind m an's ‘use’ of colour words would be equivalent, in the context of his exam ple, to the ‘use’ m ade of them by the norm ally sighted, w hat Sm art seems to mean is that for any statem ent ab o u t colour m ade by the sighted, an analytically equivalent statem ent could be found which would be capable of being know n from experience to be true by S m art's blind man. If this were true, then the discrim ination thesis would follow from it as a necessary consequence, but why should we accept it as true? If we are to accept that for every statem ent ab o u t colour which the sighted observer can m ake, an analytically equivalent one capable of being indepen dently asserted by the blind m an can be found, then we cannot, presum ably, confine ourselves to attributing a contingent deficiency of expressive pow er to the particular natural language in which it happens th at the sighted observer m ust articulate his rem arks ab o u t colour. W'hat we shall have to claim is that there exists a necessary deficiency of expressive pow er which (being necessary) afflicts every possible natural language, and which m akes it im possible for language (language per se that is) to penetrate, as it were, below the level of recording discrim ina tions and failures to discrim inate, to the level of phenom enal description; description, th at is, o f w hat Sm art would call the qualia of experience. But this raises obvious difficulties. How can we possibly hope to dem onstrate any necessary truth about any subject m atter as m anifestly em pirical and contingent as ‘language’ or ‘experience*? It m ight be argued at this point that the reason why language is necessarily incapable of being used to describe the qualia of experience is that there are sim ply no such things: our
6
NAMING AND DIS C R IM IN A T IO N
experience possesses structure but no content. But this will not do, for to assert this am ounts sim ply to reasserting Sm art’s claim about the analytical equivalence in principle of the dis course of the sighted and the blind, and this is the very claim for whose assertion we are trying to find grounds. I cannot find in S m art’s discussions any clearly stated reason for accepting this claim. Farrell, however, does offer quite a powerful argum ent in its favour, and one o r two rem arks of Sm art’s suggest that he too has this argum ent in the back of his mind. Farrell sees clearly that w hat the inexpressibility theorist must defend is the necessary inexpressibility of content. ‘W hen we say “ We can ’t describe the experience of seeing a red patch” (in the raw feel11 sense) the “ can ’t” is logical.’7 He argues that the reason why the experience of seeing a red patch cannot (logically) be described, is th at such an experience is ‘feature less’. ‘I shall say that the experience of X which we are alleged to be leaving out [when we describe the totality of X ’s verbal and behavioural response to a given stimulusl is featureless . . . The experience of X is featureless because there is nothing about it that X can discrim inate. If he does discrim inate som ething that appears to be a feature of the experience this som ething at once becomes, roughly, either a feature of the stim ulus in the sort of way that the saturation of the red in a red shape is a feature of the red shape, o r a feature of his own responses to the shape.’8 W hy can no features of raw feels be discrim inated? T he answer is to be found in F arrell’s definition of a raw feel. W hen a given observer, confronted by a given stim ulus pattern, has m ade all the discrim inations which he is capable of m aking with respect to is, and has m arked each of these discrim inations 6 Farrell means by ‘raw feel* the phenom enal character o f an experience. The term is adapted from Tolm an, P u rposive B eh a vio u r in A n im a ls and M en. 7 Op. cit.. p. 18. s Op. cit., p. 178.
NAMING AND DISCRIMINATION
7
by som e appro p riate piece of verbal o r non-verbal behaviour, w hat, by definition, rem ains to be described, if per impossibile such description were possible, is the ‘raw feel’ of the objects of perception between which he has been discriminating. Farrell am usingly docum ents the anguish of psychologists over their inability to describe such ‘bare' experiences, and thus to com plete in an objective and im personal way (he psychological description of experience. F arrell's point, however, is that just because when we reach the point at which we encounter raw feels, all possible discrim inations h a vc.ex hypothesit been made, it is senseless (logically absurd) to dem and of the observer that he m ake further discrim inations with respect to those raw feels, and thus senseless to dem and any kind of description, scientific o r otherw ise, of them .0 T hus, if the argum ent works, Farrell has established both the inexpressibility thesis and the discrim ina tion thesis, while avoiding the support of any such suspiciously corroded staff as the verification principle or the use theory of m eaning, by dem onstrating that the inexpressibility of the con tent of experience is, in a quite simple way, a logical con sequence o f o ur concept of experience itself. 9 Farrell's argument at this point has general affinities with Locke. For Locke, as for m ost later empiricists, to state the meaning o f an expression is in elTcct to treat it as an analysandum to be replaced in discourse by som e analysans. T he linguistic nature o f this operation is obscured in Locke by the fact that for him. analysans and analysandum are conceived as psychological entities: ideas. But, nonetheless, to explain the meaning of the nam e o f a com plex idea is in clTcct, for Locke, to replace the com plex idea in question with som e set o f simple ideas which can be displayed as standing in clear and com prehensible relationships to one another. A simple idea. now. just is an idea which can be subjected to no further analysis of this kind. Hence such sim ples are ultimate in our conceptual scheme, in the sense that they correspond to the concepts in terms o f which all explications o f m eaning must ultim ately be framed. Sim ple ideas then, in a sense, make possible alt com m unication between men—and yet, paradoxically, just because they are sim ple (i.e. insusceptible o f analysis) their nature remains itself incom m unicable. They can perhaps be identified (by pointing, by reference to the contexts in which they occur) but they cannot be described. The extrem e oddity o f this position has not. I think, been generally appreciated even today.
8
NAMING AND DIS C R IM IN A T IO N
III
But docs F arrell’s argum ent w ork? W hat we m ust first notice about this elegant and persuasive argum ent, I think, is th at it derives the inexpressibility thesis from two prem ises, one con cerning the nature of experience an d the other the nature of linguistic description, and th at it will n ot yield a proof of the thesis unless both these premises are in fact true. T he two premises in question m ay, I think, be stated as follows: (i) Perceptual discrim ination has em pirically discoverable limits: any observer, in the process of m aking finer and finer discrim inations between stim uli of a given sort m ust reach a point at which he is unable to discrim inate between certain stimulus-occurrences, which he m ust therefore adm it to be qualitatively identical, although not, of course, num erically identical, since to say th at tw o stimuli are qualitatively identical is precisely to say th at the sam e stim ulus, qualita tively speaking, can recur. (We shall say th at when an observer is acquainted with such a stim ulus occurrence he is acquainted with an instance of an in d isc rim in a te recurrent) (ii) W hen an observer m arks a perceptual discrim ination behaviourally, by responding differentially, say by buttonpressing, to discrim inable stim uli, he is doing som ething which is not essentially o r im portantly different from m arking such a discrim ination verbally, by assigning to the discrim in able stimuli different nam es o r descriptive phrases. (i) seems to me to be relatively unexceptionable;10 (ii) to be false. 10 Problems arise from the fact that for three closely similar colours, A , B, C, A m ay be discriminable from C although it cannot be discriminated from B nor B from C. Farrell could, however, do justice to this possibility by quite sim ple modifications to his thesis, so for the sake o f sim plicity I shall ignore it in what follows.
NAMING AND D IS C R IM IN A T IO N
9
W e can see this if we look m ore closely at the way in which (ii) is involved in F arrell's argum ent, and at some of the con sequences which flow from it if it is taken strictly and literally. W hat F arrell’s argum ent certainly shows is that there must be indiscrim inablc recurrents. It only shows, however, that indiscrim inable recurrents are necessarily indescribable if des cribing som ething essentially am ounts to discrim inating some feature of it. T his will be true if, and only if, a name o r descrip tive phrase functions sem antically (that is, functions in its role as an elem ent of a natural language) as label, or tag, for a given determ inate stim ulus in the same way that a behavioural response such as button-pressing m ay do. If describing is a form of behavioural tagging of discrim inations then, manifestly, when there are no m ore discrim inations to be tagged, descrip tion m ust have an end. But if description in language cannot be understood sim ply as a variant of behavioural tagging then it rem ains possible that some sort of linguistic description of the qualitative character of instances of indiscrim inable recurrents m ight be feasible, and, since F arrell’s argum ent contains no m achinery for excluding this possibility then, in th at case, F arrell will have failed to dem onstrate the logical indescribability of raw feels and instead will simply have reduced the issue to one turning upon a H um ean challenge: 'Show m e how raw feels are to be described.’ A nd this is a challenge which I think, despite the contrary weight of three hundred years of C artesian an d British Em piricist dogm a, adm its of an answer, although a rath er com plicated one.
IV Is it, then, reasonable to suppose th at description in a natural language is essentially a form of behavioural tagging of stimuli? L et us call a language conceived in such a way a discrimination language. F o r the purpose of setting up such a language we
10
NAMING AND DIS C R IM IN A T IO N
have to conceive of its correlative extra-linguistic w orld— that world with which the sem antic rules of the language connect its lexemes11— as an array of stimuli between any two of which an observer who uses the language can either discrim inate or not discrim inate. T he assignm ent of m eanings to the lexemes of a discrim ination language in terms of the contents of this extra-linguistic world proceeds as follows: if the observer can not discrim inate between two stim uli he tags them with the same lexeme; if he can discrim inate between them he tags each with a different lexeme. This specification gives us a language with an extremely simple and restricted structure of sem antic rules,12 and it may be objected that it is obvious that no natural language could possibly be so restricted in its sem antic m echanisms. This, I think, is true. My reasons for nonetheless proceeding to discuss discrim ination languages are first, that this, as we have seen, is the model of the semantics of a natural language that one gets if one takes absolutely rigorously and literally the account of describing which is implicit in F arrell's argum ent, which itself is, in one form or another, very widely accepted am ong philo sophers as sound; and second, that it is im portant to get clear not only that this model is inadequate as a theoretical repre sentation of the sem antics of a natural language, but also why it is inadequate. In a discrim ination language a given lexeme is presum ably in practice given a m eaning by being associated with one o r more instances of some indiscrim inable recurrent. T he utterance of a given lexeme L a t T, in response to some total stimulus51 1 wish to use ‘lexem e’ as a term denoting indifferently the members o f any class o f entities postulated at any level o f linguistic description at which semantic rules m ay be supposed to operate. In my usage it is thus neutral with respect to degree o f structural com plexity: lexem es, for exam ple, need not be. though they may be, morphemes. 12 F or exam ples o f languages with more com plex systems o f rules see m y M eaning an d Stru ctu re, ‘Studies in Language.’ ed. N oam Chom sky and Morris Halle. Harper and Row. N ew York, 1972; ‘Category M istakes and Rules o f Language’. M in d , 1965.
NAMING AND DISCRIMINATION
pattern Sj thus has (as its use, as Sm art would say) the function of directing a hearer's attention to some previous stimuluspattern S0 by reference to which, at T 0t L was, for both speaker and hearer, assigned a m eaning. If we think of the utterance of L as constituting the utterance of w hat Q uine has called a ‘oneword sentence' then the sentence might be regarded as equivalent to (as asserting the same proposition as) some English sentence of the form: w
(1) So and Si are in d iscrim in a te. T his account of the signification of L is open to the objection that ( 1) fails to specify in respect o f what type o f feature S 0 and 5 t m ust be in d isc rim in a te , in order that their indiscrim inability m ay justify the utterance of L. It can be replied, however, that a well-designed discrim ination experim ent evades this difficulty by its design. We can, for exam ple, test discrim ination by using a specially designed array of colour chips, each of which is so sim ilar to its fellows in shape, size, texture of surface and so on, that only their hues can in practice be used to discrim inate one chip from another. We can thus, it seems, by operating a t the em pirical limits of the ability to perform sensory discrim ina tions, use these very lim its to single ou t from the ordinary con fusion of sensory experience the type of sensory feature which we wish the lexemes of o u r discrim ination language to denote. W hen we try to apply this model of the sem antic rules governing the denotation of basic sensory terms to natural languages, we encounter the following series of objections. The first is the relatively trivial one that it is difficult to differentiate hue from subjective brightness w ithout the use of verbal direc tions. We will, however, simply assum e that this difficulty is not insuperable. Secondly, and more seriously, in the universe of experience in which the colour vocabularies of natural languages are learned and used, well-designed arrays of colour chips are extrem e rarities. We shall assum e that this objection.
12
NAMING AND D I S C R IM IN A T IO N
too, can be answered and that som ething of the sort often described by psychologists in some such conventional phrase as ‘the patterning of stim uli and reinforcem ents by the linguistic com m unity’ fulfils the function of singling out types of sensory feature which is perform ed in a different way by the design of a discrim ination experiment. There is, however, a third objection which I think is fatal to the thesis that natural languages, o r those parts of them which handle reference to im m ediate sensory experience, are discrim ination languages in the sense so far defined. It is th at the colour concepts m ade available by a discrim ination language are radically different in type from those m ade available by an ordinary natural language such as English. T he point which I wish to m ake is not easy to explain, for reasons which will become clear in a m om ent, w ithout intro ducing a technical term, and this I shall now do, stifling in the interests of clarity a reasonable distaste for neologism in philo sophy. I w ant to speak of a particular hue, presented in a certain definite degree of saturation and tonality,15 as a colour presentation. T h e cover of this book, on this terminology, exhibits a certain colour presentation. Different areas of the cover, assum ing the cover to be hom ogeneously coloured and not yet dappled with w hat a structuralist philosopher of my acquaintance felicitously calls ‘the stains of hum an use’, exhibit the same colour presentation. T h e cover of the next book on the shelf, however, in all probability exhibits a quite different colour presentation. In other words, colour presentations com prise one species of in d isc rim in a te recurrents. A ‘colour’ in English, and so far as I know in all other natural languages which have received careful study, is a set or category of colour presentations. T hus ‘red ' in English is equally correctly applicable to a very large num ber of distinct colour presentations, including, for exam ple, scarlets, crim sons, some 18 The latter terms are standard in the description o f colour, but in any case will be explained here at a later point.
NA MI NG AND D IS C R IM IN A T IO N
13
deep pinks, som e reddish purples, and so on. In everyday English we d o some justice to this fact by speaking of ‘shades’: we say things like ‘a rath er bluish shade of purple’ for example. But ‘shade’ is not equivalent in m eaning to ‘colour presentation’ as we have defined it above. People very often use ‘shade’ to m ean sim ply ‘h u e’: so that the sam e ‘shade’ of red could be presented in different degrees of saturation and tonality to yield w hat, in o u r term inology, w ould be different colour presenta tions. F u rth er possibilities of confusion arise from the fact that, while ‘a colour’— e.g. red— is in English a category of colour presentations, the expressions ‘the colour of x’ and ‘the same colour as x ’ are norm ally used not to refer to colour category but to colour presentation. T hus when I speak of ‘the colour of this bo o k ' I d o not generally m ean to refer to the English colour category (e.g. blue) into which the colour presentation exhibited by the cover falls: I m ean to refer to the particular colourpresentation which the cover exhibits. A nd similarly it would be very odd to say of two very differently coloured red things (one scarlet, one crim son, for exam ple) that they were ‘the same colour’, for to say that two things are ‘the same colour’ norm ally m eans that they exhibit the same colour presentation o r closely sim ilar colour presentations. Since technical term s carry their own possibilities of con fusion, there is very little th at one can d o with ordinary English except use it an d hope for the best. I shall therefore use the w ord ‘co lo u r’ throughout w hat follows as it is ordinarily used, trusting to the term ‘colour presentation’ to help us to evade the possibilities of confusion im plicit in ordinary usage. M y point against the thesis th at the colour vocabularies of natural languages are discrim ination languages is, now, pre cisely th at the colour term s in a discrim ination language are nam es for colour presentations, whereas the colour terms in natural languages are nam es for categories of colour presenta tion.
14
NAMING AND DIS C R IM IN A T IO N
T his fact is closely connected with a feature of natural language which has been exhaustively em phasized over the past ten years by linguists interested in the construction of genera tive gram m ars;14 namely, that o u r use of language exhibits creativity or originality. Linguists have prim arily concerned themselves with syntactic originality: they have pointed out that a native speaker’s ability to construct and interpret syn tactic structures is creative in the sense of not being lim ited in its exercise to the set of structures which have figured expressly in past learning situations. Som eone who knows English has acquired, as a result of a finite period of language learning, the ability to assign o r deny syntactic structures to a potentially infinite num ber of putative English sentences.15 But it seems clear th at o u r ability to nam e colours is creative in a closely related sense. T h at is, I do not need to be expressly told that a particular colour presentation is one of the group of colour presentations called, in English, ‘blue* in order to recognize it as blue, and this rem ains true even if I have never before come across that particular colour presentation— if for exam ple it is the colour of a newly synthesized dyestuff. As I adm iringly watch the new com pound precipitating in the retort I m ay in deed need to be told its m olecular form ula o r the process by which it is m ade, but I shall hardly need to be told that it is blue. 14 But o f which Wittgenstein was well aware at the time o f writing the Tractatus. T he same point was discussed in the Thirties by writers o f the Vienna Circle, particularly, as we shall see. by M oritz Schlick. 15‘The central fact to which any significant linguistic theory must address itself is this: a mature speaker can produce a new sentence o f his language on the appropriate occasion, and other speakers can understand it immedi ately, though it is equally new to them . Most o f our linguistic experience, both as speakers and hearers, is with new sentences; once we have mastered a language, the class o f sentences with which we can operate fluently is so vast that for all practical purposes (and. obviously, for all theoretical purposes) we m ay regard it as infinite. Norm al mastery o f a language involves not only the ability to understand immediately an indefinite number o f entirely new sentences, but also the ability to identify deviant sentences and, on occasion, to im pose an interpretation on them., N oam Chomsky, C urrent Issues in L inguistic T h eo ry, M outon & C o., The Hague. 1964, p. 7.
NAMING AND DISCRIMINATION
15
I m ay even say, for exam ple, ‘T h a t is a shade of blue the like of which I have never in my life seen.' T o possess the concept blue (to know the meaning of the English colour term 'b lu e') is, then, to possess the capacity to apply the term ‘blue' to colour presentations which I have in no sense been taught to regard as blue. It seems clear that the description of a discrim ination language which we have before us contains no explanatory m achinery which could render explicable the acquisition of such a capacity. T he speaker of such a language who has been taught to associate a particular lexeme with anything identical in colour with some specifically designated colour sam ple will be a fortiori prevented by the rules of his language from applying that lexeme to anything which is discrim inable in colour from the sam ple in question, however sim ilar in colour to that sample it m ay be. He will possess, not a colour nam e, but a nam e for one particular colour presentation. A nd he will be unable to attach any nam e to a colour presentation, unless the colour presentation in question is indistinguishable from th at exhibited by some colour sample with which he has been m ade acquainted during the process of being taught the language, and with which a lexeme has been expressly associated. It is im portant to see, too, that we cannot arrive a t the sort of colour-vocabulary characteristic of natural languages by postulating a very long and detailed process of discrim ination-language teaching, in which great num bers of discrim inable colour presentations are associated with lexemes en bloc, m any hues to one lexeme. T he postulation of such a process can in principle explain any finite particular sequence of associations of single colour nam es with hues by some finite set of speakers of a natural language over a finite period of time; but it cannot explain the open-ended capacity to name unfam iliar hues which is acquired in learning to speak a natural language. M oreover the postulation of such a process offers no explanation of the assignm ent of com pound colour names to unfam iliar hues: for exam ple ‘a greenish blue’, ‘a deep
16
NAMING AND DISCRIMIN ATIO N
purpleish red’. We shall see later that the possibility of giving such characterizations is an im portant feature of the colour description system of natural languages.1®
V A t this point it is w orth pausing, however, to show that a certain alternative account which m ight seen prima-facie rather attractive will not do. It m ight seem that w hat is w rong with a discrim ination language is sim ply that it m akes the location of nam eables depend upon w hether or not successive colour pre sentations can be barely discriminated. Perhaps a m ore ade quate model could be based upon the noticing of sim ilarities between colour presentations, since, after all, it seems not unreasonable to suppose that w hat links colour presentations together into the sets denoted by English colour term s are the relationships of resem blance which subsist between them . It might seem quite easy to restate the inexpressibility thesis in terms of this model. C olour words, we m ight argue, refer to patterns of sim ilarities between colour presentations. A pattern of sim ilarities is an array of form al relationships of a certain sort, and it seems logically possible, for exam ple, th at two sets of colour presentations possessing no com m on m em ber m ight nevertheless each exhibit internally identical sets of sim ilarity 16 The a d h o c character— the generative infertility— o f discrimination languages perhaps explains w hy natural languages use other mechanisms o f colour reference. A generative system affords m aximum com municative power in return for minimal com plication in the system o f rules which must be assimilated by the learner. In an a d h o c colour language based upon inability to discriminate as its fundam ental notion, the occasion to use a colour word w ould rarely recur. Block association (m any-one association) would to som e extent overcom e this, but would still leave the learner open to the possibility that som e new hue (a new dyestuff e.g.) might be unnam eable, and would, in any case, be intrinsically much m ore complicated than, for exam ple, the structure o f rules presented in Part 2 o f the present volume.
NAMING AND DISCRIMINATION
17
relationships am ongst their respective mem bers; so that in such a case the difference in content between the two sets would be in principle inexpressible in language. T his move, however, exposes the inexpressibility theorist to a further set o f difficulties, which a m ore straightforw ard account, in term s of discrim ination rather than similarity, avoids. W e can best approach these by way of some further reflections on nam ing. N am ing, as we have seen, can be regarded as a device for m aking tem poral cross-references between stim uli, and this is how the inexpressibility theorist does regard it. T o attach a lexeme to a given total stim ulus pattern is implicitly to assert (indeed, if we are prepared, like Q uine, to countenance onew ord indicative sentences, we can regard it as explicitly asserting) that th at total stim ulus pattern is in some sense the sam e as any other total stim ulus pattern to which the same lexeme could be correctly attached. T h e trouble is, in w hat sense? Trivially, the answ er to this question m ust be ‘the same for the purpose of attaching this lexem e’; but w hat m akes two stim ulus patterns a p t for the attachm ent of the same lexeme to each? T here seem to me to be two possibilities. O n the one hand, this question m ay be answ ered by the natural articula tions of the given-in-experience, so th at once we attach a lexeme at To to a particular com ponent of o ur experience the natural articulations o f things in our experience which define things o f that sort as distinct com ponents o f our experience themselves suffice to determ ine at w hat points in time T u T 2f. . .* T n we are to reattach the lexeme to the shifting fabric of our experience. In this event we shall be able to speak of our experience as containing, o r consisting of, natural nameables. T he other alter native is th at we ourselves, through the stipulation of rules or criteria of some sort, d o some of the work of determ ining w hat consequences for o u r future lexem e-attaching behaviour at T lt 7'2, . . T a are im plicit in o u r attachm ent of a given lexeme at T 0 and thus, per contra, of course, of determ ining what type of
18
NAMING AND DISCRIMIN ATIO N
entity it was to which we attached the original lexem e at T 0. T o take this latter view is, I shall say, to hold that the lexemes to which it applies designate constructed nameahles. Talk of ‘natural nam eables’ perhaps m akes it sound as though I want to claim th at nam ing m ight be thought of som e how as a ‘n a tu ra l’ process, w ithout any elem ent of linguistic convention; perhaps as som ething hum an beings do as a mere causal consequence of the way in which they perceive the world. T his is not a t all w hat I have in m ind. Even if I can perceptually discrim inate som ething as a clearly defined feature of my en vironm ent prior to my acquisition of any linguistic capacities, the mere act of pointing to that thing and uttering som e lexeme carries in itself no im plication whatsoever for any future act of pointing and lexeme-uttering. We can only extract such im plica tions from the act in question if we suppose it to be carried out by some person, o r persons, who have adopted some such linguistic rule as the following: (2) A ttach lexeme / to any object o n of type T which is indis tinguishable from (sim ilar in respects a %6, c . . . to) the object, o,, of base attachm ent of /. It is im portant to see that (2) is a rule, and that w ithout it there would be no (linguistic) procedure of lexeme ‘attach m en t’ at all. We cannot reduce the conventional elem ent in language to nothing. O n the other hand we can certainly im agine it, at least, progressively reduced, and the notion of a language all of whose basic lexemes denote natural nam eables can be thought of as the theoretical limit of this process of m aking more parsim onious the system of rules which define constructed nameables. T he furthest we can go in this direction is, perhaps, a language in which the fundam ental rules governing the attach m ent and reattachm ent of lexemes have the general form of (2), but in which the crucial expressions, ‘object of type T \ ‘indistinguishable from ’, ‘sim ilar to in respects a, />, c, . .
NAMING AND DISCRIMINATION
19
which occur in (2) are defined in terms of the criteria employed by the perceptual m echanism s which enable us to discrim inate things of the type being nam ed as distinct elements in our per ceptual field. T h e reattachm ent of lexemes would thus depend only m inim ally upon anything describable as ‘linguistic con vention', and m axim ally upon those features of our experience which enable us to divide experience into distinct separable elem ents prior to, an d quite independently of. any linguistic capacities which we m ay be im agined to acquire later on. The referents of the basic term s of such a language would thus in effect be defined as distinct objects of reference (nameables) independently of linguistic convention, and it is such objects of reference that I shall refer to as ‘natural nam eables' in what follows. T he referents of the terms in a discrim ination language are natural nam eables in this sense. T h at is, once we have defined a nam ing procedure by stipulating some such rule as (2), our experim ental design— the use of standard colour chips and so forth— really does single o u t a class of natural nam eables (i.e. colour presentations, considered as indiscrim inable recurrents) which are sim ply given in experience. T h a t is, we d o not, as we have seen, given the design of our experim ental situation, face the question (a) ‘W hat constitutes a h u eV o r (b) ‘W hat entitles us to say that two objects are o f the sam e h u eV F or a hue is sim ply a m em ber of that class of properties by reference to which alone our special colour chips can be told apart, while two hues are the same when, and only when, no subject can tell ap a rt the chips which exhibit them. Given our experim ental arrangem ents, that is, which define a certain limited field of ex perience, hues by their nature define themselves with respect to that field both as a type of property and as individual properties of that type. T h e notion of a constructed nam eable can best be under stood by contrast with that of a natural nam eable. T he objects of reference corresponding to the terms of a natural language w
w
20
NAMING
AND DISCRIMINATION
are constructed nam eables just in case the term s ‘object*, indis tinguishable from ’, ‘sim ilar in respects a , b>c, . . whi ch occur in (2) are defined not merely by reference to the conditions of our pre-linguistic perceptual sensibility but also in p a rt by reference to further linguistic conventions; that is, conventions going beyond (2) itself. VI If the referents of colour term s in a natural language can be construed as natural nam eables, then it follows that the inexpressibility thesis is correct. F o r if the referents of colour terms are natural nam eables, then w hat counts as a colour, and w hat colour presentations fall into the group of presenta tions to which I later learn to ascribe a particular colour nam e (for exam ple ‘red’) are questions fully determ ined by the condi tions of my pre-linguistic perceptual sensibility. In other words, the fabric of my experience, prior to the point a t which I begin to learn language, is already divided into objects of various types (perceptual m odalities, particular colours, and so on) between which I can discrim inate. It now looks as though F arrell’s account of language is quite correct. L anguage merely supplies conventional m arkers for these pre-existing perceptual discrim inations by setting up a system of differential verbal responses, the utterance of which depends by convention upon the discrim inability or indiscrim inability of successive stimuli. And as we saw earlier, the inexpressibility thesis follows immediately from this account of the nature o f language. O n the other hand, if the referents of the term s in our ordinary colour vocabulary are constructed nam eables, then the inexpressibility thesis cannot be dem onstrated in this short and satisfying way. It m ay still of course be dem onstrable in other ways, but in order to arrive a t a dem onstration we should have to exam ine the additional rules (additional to rules of the same type as (2), th at is) by m eans of which language resolves
NA MI NG AND D IS C R IM IN A T IO N
21
the questions ab o u t the com position and limits of classes which rules of type (2) leave open* but which m ust be settled if we are to explain how a given speaker of a natural language applies term s within roughly the same limits of application as those observed by other speakers of the sam e language. It m ight tu rn ou t th at a thorough exam ination of these linguistic m echanism s would leave us still in a position to form ulate some version of the inexpressibility thesis. But then again, it m ight not. So far o u r argum ent has run as follows. F arrell's argum ent presupposes a view of languages according to which the basic colour term s in o u r language designate natural nameables. This is not im plausible if we can suppose that the basic colour terms in our language designate particular colour presentations. F or a colour presentation m ight seem to have a rather good claim to the status of a natu ral nam eable. T he necessary criterion of lexeme reattachm ent in the case of determ inate colour terms is provided by o ur actual incapacity to distinguish between the objects sim ply by appeal to their colour when both exhibit the same colour presentation. T he determ inable ‘colour’* as we have seen, gives us a little m ore trouble. We have to discover a criterion for recognizing som ething as a colour. As we rem arked earlier this can be done by using a special set of objects (standard colour chips) so constituted th at in practice colour is the only criterion by reference to which one such object can be distinguished from another. And although it seems grossly im probable at first sight that the actual teaching o f colour language is carried ou t in conjunction either with such sets of objects o r with natural analogues of them, perhaps ingenuity m ight m itigate this initial implausibility. T he trouble with this, however, is that, however plausible the claim th at colour presentations are natural nam eables, it is dem onstrably false that the basic colour terms in our language designate colour presentations. T hey designate colours, and colours are classes of colour presentations. V
w
w
22
NAMING AND D IS C R IM IN A T IO N
Different speakers of the sam e language assign the same colour term to (with m inor variations) the same class of colour presentations, and the question confronting us now is: w hat accounts for this capacity? W hat are the criteria which the speaker uses to assign a given colour presentation to a given colour category (e.g. ‘r e d ’) and w hat are his criteria for determ ining the limits of application of the term ‘colour’? (How, in other words, docs he answ er, for particular colour terms and for the determ inable ‘colour’, the questions pre supposed by rules of type (2)?). T he suggestion th at lies before us is that a sufficient answer to these questions is supplied by the natural (i.e. pre-linguistically d isc rim in a te ) resem blances subsisting between colour presentations. A colour is, on this view, simply a class o f natural resem blances, and the term ‘colour’, used as the nam e of a determ inable universal, stands for the class o f all such classes. T h e trouble with this suggestion, I think, is th at resem blance can serve as a principle upon which to connect things with one another b u t not as a principle upon which to separate things from one another. But it is the latter function which it is being called upon to serve here. W e are told th at a colour is a certain class of resemblances. W hich class? Classes o f resem blances are by their nature indefinitely extensible. L et us suppose th at I exhibit two red colour presentations, ru r 2, and propose as a criterion for determ ining the lim its of application for the term ‘re d ’ some such rule as the following: (3) (i) Any colour presentation is a m em ber of the class of red things if it resembles red colour presentations to the degree that red colour presentations resem ble each other, (ii) and r2 are red colour presentations. It is evident, I think, th at this rule will entitle us to add m ore and more colour presentations to the category ‘re d ’ in serial order, but gives us no criterion w hatsoever for refusing to adm it
NAMING AND DISCRIMINATION
23
a colour presentation to the category. T h e process of adding colour presentations to the category ‘re d ’ will thus go on until the colour universe has been exhausted. And it is not at all obvious how any criterion for stopping this process at a certain point can be extracted from the bare notion of resemblance. T he notion of a ‘class’ of resem blances thus stands as much in need of explication as the notion of a class of colour presenta tions. Sim ilar problem s arise with ‘colour' construed as the name of a determ inable universal. If we cannot define a colour as a class of resem blances, we cannot define ‘colour' as the class o f all such classes. A nd equally clearly if, choosing several diversely coloured objects, we attem pt to indicate the scope of application of the term ‘colour’ by stipulating that any object is coloured which resembles these in just the degree and respects in which they resem ble each other, we shall not succeed in distinguishing the application of ‘colour’ from that of ‘brightness', o r ‘surface', o r for that m atter, from any other feature, quite unrelated to colour, in which an arbitrarily chosen set of paradigm objects m ay resem ble one another (this latter possibility becom es stronger the larger the set of para digm s we use). It m ight perhaps be thought th at we could evade these difficulties by specifying th at the crucial class of sim ilarities is precisely that class, attention to which would lead a subject to arrange colour sam ples exactly as a com petent speaker of English does when he arranges such sam ples under the headings ‘re d ’, ‘green’, ‘yellow’ etc. But this is a wholly unenlightening move. W hat we wish to know is, precisely, w hat the com petent speaker of English does when he assigns colour nam es, and that involves knowing how he singles ou t (by reference to w hat criteria he singles out) certain classes of colour presentations from am ongst the indefinite num ber of possible groupings and regroupings into which the experiential m aterial to which he has access in
24
NAMING AND D IS C R IM IN A T IO N
perception might equally well be ordered. O ur problem is not, of course, the psychological one of whether, o r how, the subject sees colours. T he first of these questions we can answ er by appeal to discrim ination tests, the second would take us into the anatom y and physiology of vision. O ur question concerns the nature of w hat is learned by someone who learns to assign the m em bers of a certain class of words to certain features of his experience in ways which always turn out, barring colour blindness or disagreem ents ab o u t very fine distinctions o f hue, to coincide with the assignm ents m ade by other native speakers. And this, as we shall argue in a m om ent, is a question not about his psychology but ab o u t the ruie-structure o f his language. L et us sum up this prelim inary stage of the discussion. It seems clear that while colour presentations can, up to a point at least, be regarded as natural nam eables, colours (that is to say the entities denoted by the colour term s of ordinary English) can not. A colour is not a self-individuating elem ent of our ex perience,17 whose very n atu re settles o u t of hand all questions about the present and future attachm ent of colour names. And hence, since natural languages with colour vocabularies ob viously exist, so that such questions m ust clearly be settled by some m eans, we m ust look elsewhere to discover w hat it is that settles them . A nd the only alternative answ er seems to be that they are settled by some stipulated rule o r convention to which all speakers of the language adhere: that, in short, colours are constructed nam eables. T here is at stake here a point of general philosophical im portance. Philosophers are accustom ed to the idea th at ex perience presents us— o r seems to present us— with various types of objects: m aterial things such as tables o r clouds; 171 am not, o f course, saying that c o lo u r is unreal or in any w ay ‘con structed’ by us, and still less— what would surely be transparently absurd— that our experience o f colour is a sort o f figment o f our linguistic conventions. C olour— what we perceive— is the total set o f colour presentations. A colour (red, green, for exam ple) is the denotatum o f a colour term in a particular language.
NAMING AND D I S C R IM IN A T IO N
25
sounds; pains; pangs and tingles; the sort of things I have described as colour presentations; perhaps even (although few philosophers would care to go this far) persons, and so on. T here is a good deal of discussion about which of these sorts of object m ust be treated as prim itive and fundam ental and which can be treated as ‘constructions' in some logical or trans cendental sense of th at term . But I think m ost philosophers take it for granted th at there is no difference a t all between speaking of ‘objects' of experience in this sense and speaking of ‘objects’ of reference of term s in a language. They assume, th at is, that the articulations, o r structures, which divide experience into distinct classes of perceptual object, and in dividuate objects within these classes, are identical with the articulations which divide experience for the purpose of lin guistic reference. T his conflation is no doubt m ade easier by the undim inished popularity of associative/ostensive theories of m eaning. If w ords acquire m eaning merely by being ostensively associated with features of experience, then the individuation of the features in question m ust be prior to, and independent of, language. If we are correct, however, m atters m ay be m ore com plicated than this. We have begun to discover grounds for thinking th at objects of experience and objects of reference m ay be radically different types of entity. O n this view, objects of reference are linguistic entities, in the sense that their individ uation is accom plished not m erely by the nature of our prelinguistic perceptual sensibility (however the latter may be described) bu t through the addition to this, as it were, con ceptual substratum , of some system of linguistic rules. We shall m ake use of this suggestion in w hat follows.
V II Som eone m ight object th at o u r discussion so far has been essentially concerned with w hat happens when a child learns
26
NAMING AND D IS C R IM IN A T IO N
his native language, and th at this is not a philosophical problem , but a problem for em pirical psychology. In fact we arc not a t all concerned with em pirical questions about language learning, and neither, for th at m atter, are Sm art and Farrell. W hen the inexpressibility theorist puts forw ard, explicitly o r by im plication, a thesis ab out the nature o f m ean ing and reference in language, we should not, unless we merely wish to score a trivial debating point a t his expense, take him to be putting forw ard a causal thesis. He should not for exam ple, that is, be taken to be claim ing that as a m atter o f fact the repeated utterance by a teacher of a certain lexeme in circum stances in which a learner is attending to a given sam ple discrim inableonly in hue from other sam ples in whose presence another lexeme or no lexeme is uttered by the teacher (or the process described by any other rigm arole which the reader m ay care to insert here) is causally necessary a n d /o r causally sufficient for the successful learning of the m eaning of lexemes of the type in question. W e ought not, in charity, to put such a construction on the inexpressibility theorist’s claim because (i) no evidence for any such factual claim is ever presented and it is in fact rather hard to see how any such evidence could be acquired, and (ii) even if such evidence were to be adduced the claim itself would be theoretically quite trivial and unenlightening. It is possible that eating m ango pips m ight turn out to be causally sufficient a n d /o r necessary to confer ex nihilo upon unsuspecting British diners a com plete grasp of the G ujarati language. Such a discovery w ould in itself be of no theoretical interest w hatsoever for the student of language, since it would throw no light upon the one question which, in these o r any other circum stances, he would wish to have answered: w hat is it th at the rem arkable pips, in conferring a grasp of G ujarati, confer? We can in fact distinguish two quite different questions about ‘the nature of language*, viz.
NAMING AND DISCRIMIN ATIO N
27
Qi W hat is acquired in acquiring the ability to use and understand a natu ral language ab initio? Q z W hat are the causal m echanisms, o r the causally neces sary a n d /o r causally sufficient conditions of the initial learning o f a natural language? M ost philosophical ‘theories of meaning* have attem pted to answ er b oth questions sim ultaneously w ithout clearly distin guishing between them: this accounts both for the peculiar unsatisfactoriness of discussion on the topic and for its resistance to ‘therapeutic* dissolution. F or obvious reasons of the sort given above, Q 2 does not seem to me to be a question which can be fruitfully tackled by philosophers, o r for that m atter by psy chologists if they approach Q 2 w ithout having in mind any glim m ering of an answ er to Q x. Q x on the other hand is an intelligible an d perhaps answ erable question, and I think we m ust construe the inexpressibility theorist’s thesis, so far as it is not m erely a foredoom ed attem pt to d o em pirical psychology by the light o f neither theory nor experim ent, as an attem pt to answ er QiB ut once we state Q x clearly, certain quite strict criteria of adequacy for proposed answers to it m ake themselves apparent, including at least the following one. ( C A J A n account of w hat is learned in learning a first language, if it is to possess any explanatory force, m ust avoid using as an unanalysed prim itive notion, any concept c, if the application of c can only be explained within the context of the theory by being exemplified through the presentation of sam ple expressions in some natural language. Such an account m ay, however, m ake use o f such concepts to locate the linguistic phenom ena which it proposes to explain. ( C A J is perhaps not im m ediately intelligible as it stands. My intentions in form ulating it is this. T h e ability to use a language, like the ability to ride a bicycle, is not entirely perspicuous to
28
NAMING AND DISCRIM INATION
the possessor. W e can know w hat balancing is and stil! not know w hat exactly we do when we balance, and we can know w hat a general nam e is and still be unsure about w hat it is that we d o in applying general nam es (a state of ignorance to which much of the developm ent of the theory of universais bears witness). W e norm ally explain notions like that of a general name (or a sentence, o r an article, o r a definite description) by reference to examples of word-use o r sentence construction. It is obviously essential th at we should use such concepts in any theory in which we wish to talk ab o u t general term s, sentences, and so on, since otherwise we should have no m eans of locating o r referring to the linguistic phenom ena which interest us. But if a theory which sets out to explain w hat is learned in learning to use a general term can evade certain difficulties o r lacunae in the explanation which it offers only by assum ing the truth of some proposition which uses the everyday intuitively based con cept of a general term, then we shall have illicitly assum ed that the capacity to use general terms is perspicuous to its possessor in the very way in which, if there is anything in the m atter for us to theorise about in the first place, it is not. A nd hence our theory will be trivial because circular, having com m itted a sort of petitio principii. T h e rationale of C A x is, then, the avoidance of this sort of circularity. W hen I say, now, th at the theoretical specification of a strict discrim ination language singles ou t a class of natural nameables, w hat I m ean is that it is free from this sort of circularity, at least as far as the questions (a) and (b) are concerned. W hat constitutes a colour, and w hat it is for two sam ples to be o f the sam e colour can be explained by reference to the design of the m atching experim ent, and to an operationally defined prim itive capacity of any subject to m atch colour sam ples o f the specified sort in ways which conform to the practice of oth er subjects, w ithout using any concept which we should need to explain by reference to linguistic examples: w ithout using, for exam ple, the concept of being able to use the word *hue*
NAMING AND DISCRIMINATION
29
But if we turn to the w eaker and m ore liberal form ulation of the inexpressibility theorist's case in terms of the notion of resem blance between, we find th at it does not avoid the sort of circularity which ( C A X) prohibits. T he limits of classes of colour presentations cannot be specified by appeal to the bare notion of resem blance, and the theorist is reduced to specifying the resem blances he has in m ind merely as those by attention to which one speaker is enabled to apply colour terms in ways w hich m atch their application by other speakers. I am not suggesting, of course, th at there is any difficulty in understanding how som eone recognizes an object to be the sam e o r sim ilar in hue to one which he has seen elsewhere; not, of course, because there are not difficulties (of a psychological and physiological character) involved in understanding such phenom ena of m em ory and recognition, but because the phenom ena in question are entirely irrelevant to our present inquiry. We are concerned, not with how a m an recognizes objects as sim ilar o r identical in hue to one another, but with how he m anages system atically to apply colour terms in such a way as to m atch the usage of other native speakers of his language.
PART 2
Schlick on Form and Content: A Tractarian Excursion
2. Schlick on Form and Content: A Tractarian Excursion i It is not by any m eans obviously silly to attem pt to identify colours with systems of natural resemblances. Such a theory springs from a sensible enough thought: that w hat enables us to identify colours as a distinct and clearly delineated category o f qualities, distinct from , say, degrees of brightness o r conditions of surface texture, is th at a large variety of relationships subsist between different colour presentations, in virtue of which all colour presentations hang together in a single rela tional system when considered as exem plars of hues, in a diff erent relational system when considered as exem plars of relative degrees of brightness, in a third when considered as exem plars of relative degrees of saturation, and so on. T he interest of M oritz Schlick’s lectures on Form and Con tent* lies partly in the fact that Schlick takes this thought seriously. Instead of taking the notion of a ‘set* o r ‘system’ of natural resem blances for granted, he m akes a serious effort to get clear ab o u t w hat sorts of relationship actually do subsist between colour presentations. L et us . . . exam ine the verbal expressions of our ordinary language, i.e. the sentences and the w ords by which I give a description of o ur particular green colour. We easily discover an essential feature which they all have in common: they 1 Loc. cit.
34
S C H L I C K ON F O R M AND C O N T E N T
assign ‘green ’ a certain place within a com prehensive system of shades, they speak of it as belonging to a certain order of colours. They assert, for instance, th at it is a bright green, o r a rich green, o r a bluish green, th at it is sim ilar to this, less sim ilar to that, equally dark as that, and so on; in oth er words, they try to describe the green by com paring it to other colours. Evidently it belongs to the intrinsic nature of our green that it occupies a definite position in a range of colours and in a scale of brightness, and this position is determ ined by relations of sim ilarity an d dissim ilarity to the oth er ele ments (shades) of the whole system. These relations which hold between the elem ents of the system of colours are, obviously, internal relations, for it is custom ary to call a relation internal if it relates two (or more) terms in such a way th at the term s cannot possibly exist w ithout the relation existing between them — in other words, if the relation is necessarily implied by the very nature of the terms. T hus, all relations between num bers are internal: it is in the nature of six and twelve th at the one is half the other, and it would be nonsense to suppose that instances m ight be found in which twelve would not be twice as m uch as six. Similarly, it is not an accidental property of green to range between yellow and blue, but it is essential to green to be related to yellow and blue in this particular way, and a colour which was not so related to them could not possibly be called ‘green* unless we decide to give to this word an entirely new meaning. In this way every quality (for instance, the qualities of sensation; sound, smell, heat, etc., as well as colour) is interconnected with all others by internal relations which determ ine its place in the system of qualities. It is nothing but this circum stance which I m ean to indicate by saying th at the quality [the particular shade of green, i.e.l has a certain logical structure.2 *Schlick, op cit., pp. 161-2. Schlick w as not, o f course, the first to notice that internal relationships obtain between sim ple sensory qualities, although
S C H L I C K ON FORM AND C O N T E N T
35
T o describe these relationships as logical might seem an odd misuse of the word ‘logic’: T h e explanation is th at Schlick’s thought in these lectures is profoundly influenced, as we shall see, by W ittgenstein’s Tractattts Logico-Philosophicus. ‘Logical structure’ for Schlick has associations with the T ractarian notion of ‘logical form ’. T h e theory of m eaning which Schlick puts forw ard in the lectures, and on which his version of the inexpressibility thesis is based, has close affinities with the socalled ‘Picture T heory of M eaning’ in the Tractatus. Schlick’s version, however, differs from W ittgenstein's in that Schlick m akes a resolute attem p t to convert w hat in W ittgenstein is an alm ost wholly abstract discussion of the general nature of signification into an account of the m anner in which language is connected with experience. Schlick begins by anticipating a num ber of points about the peculiar character of natu ral languages which have recently been m ade fam iliar, at least am ong linguistics, by the work of N oam Chom sky. Like Chom sky he distinguishes beween a language and a repertory of signals,3 and says th at the essential he was, perhaps, first to notice or to think it important that they form systems, or matrices, within w hich individual qualities intrinsically possess definite locations. H um e at times com es close to seeing this (see, for exam ple, the footnote to the chapter on ‘Innate Ideas*, Treatise, 1. 1. vii). s ‘Expression is entirely different from mere Representation, it is much more and cannot be derived from it. G enuine speech is som ething entirely new as com pared with the sim ple repetition o f signs w hose meanings have been learned by heart . . . . . . If language were nothing but a system o f signs with fixed significations it w ould never be capable o f com m unicating new facts. If its function consisted solely in representing thoughts or facts by means o f sym bols it could represent only such thoughts or facts to which a sign had been attached; a new fact would be one to which no symbol had been assigned, it w ould therefore be im possible to com m unicate it . . . This state o f affairs is made very clear by what is often called the ‘language’ o f certain animals such as bees and ants . . . The signals o f bees and ants represent or indicate certain occurrences, they do not express them. T hey arc restricted to these particular states o f affairs, and cannot represent anything else. The essential characteristic o f languages, on the other hand, is its capability o f expressing
36
SCHL1CK ON FO R M AND C O N T E N T
feature of the form er is th at it can com m unicate new facts by allowing the construction (as distinct from the express stipula tion) of new signs which express those facts; whereas the latter is merely a list o f associative symbols with no com binatory potential, and can hence express only the fixed range of facts with which the individual m em bers of the list have been associatively linked. F urtherm ore Schlick, like Chom sky, locates the distinguishing line between anim al and hum an ‘language’ in the transition from repertory ‘language’ to languages with com binatory o r generative potential. Schlick introduces, to express this distinction, the contrasting term s ‘representation’ and ‘ex pression’. A system of signs ‘represents’ states o f affairs in the world sim ply by associating a sign with each state of affairs. Such a language can only represent the states of affairs with which the individual signs com posing it have been linked: it cannot represent new states of affairs. A language 'expresses’ states of affairs, on the other hand, if it allows the possibility of constructing, from the simple signs of the language, complex expressions which correspond to new facts. Schlick’s discussion here exactly parallels a (far m ore con densed and cryptic passage of the Tractatus: 4.02 4.024
. . . we understand the sense of a propositional sign w ithout its having been explained to us . . . T o understand a proposition m eans to know w hat is the case if it is true. (One can understand it, therefore, w ithout knowing whether it is true.)
facts, and this involves the capability o f expressing new facts, or indeed any facts. A schoolboy opens his copy o f X enophon’s Anabasis, and by reading the first sentence o f the book he learns the fact, w hich Get us assume) is entirely new to him, that K ing Darius had tw o sons. H e knows what particular fact is expressed by that particular sentence, atthough he never cam e across that sentence before and certainly did not know the fact before. H e therefore cannot have learned that the on e corresponds to the other.*— F o rm a nd C o n te n t, loc. cit., pp. 155-6.
S C H L I C K ON FORM AND C O N T E N T
4.025
4.026
4.027 4.03
37
It is understood by anyone who understands its constituents. W hen translating one language into another, we do not proceed by translating each proposition of the one into a proposition of the other, but merely by translating the constituents of propositions. (A nd the dictionary translates not only substantives, but also verbs, adjectives, and conjunctions etc., and it treats them all in the same way.) T h e m eanings of simple signs (words) m ust be ex plained to us if we are to understand them. W ith propositions, however, we m ake ourselves understood. It belongs to the essence of a proposition that it should be able to com m unicate a new sense to us. A proposition m ust use old expressions to com m unicate a new sense.
Both Schlick and W ittgenstein see the distinction beween nam es (or sequences of names) an d propositions as consisting in the fact th at a proposition can express ‘a new sense’. Both hold that the central problem of the theory of meaning is to explain how this is possible. T he capacity of language to express new senses, Schlick goes on to argue, clearly depends upon the possibility of re-ordering the elem ents of com pound signs to form new com pound signs. T hus, for exam ple, if we understand the m eaning of: (1) T h e ring is lying on the book then, if we rearrange the parts of (1) to form (2) T h e book is lying on the ring ‘we understand the m eaning of the second proposition im m ediately, w ithout explanation’.4 4 Op. cit., p. 157.
38
SCH LI CK ON FO RM AND C O N T E N T
T he fact that language allows this kind o f re-ordering seems, to Schlick, to entail two general propositions about the nature of language. T h e first is th at w hat m akes a row of w ords into a proposition — w hat m akes it express a fact— m ust be a correspondence in structure between the expressive fact (the written o r spoken sentence) and the fact which it expresses.5 F o r structure is the only thing that differentiates (1) an d (2); hence it is by reference to the structure of the two sentences, and to th a t alone, that we are able to identify the very different states of affairs which the two sentences express. T h e astute reader will notice th at in the last half-page o r so we, and Schlick, have passed from speaking of propositions expressing senses to speaking of them as expressing facts, and also that we have begun to use ‘proposition’ and ‘sentence* as if they were synonymous. T h e explanation of both transitions is the same. Once we have been persuaded th at difference in syn tactic structure alone can produce two sentences expressing quite different propositions it becom es easy to regard structure alone as the determ inant of propositional content in sentences. If we ask how this can be, the obvious answ er which suggests itself is that the structure of the sentence somehow locates a corresponding structure ‘in the w orld’ and that this structure identifies w hat it is th at the statem ent asserts to be the case. It is now very tem pting to reify the second structure— the one ‘in the w o rld ’— and to speak of such structures as ‘facts’ o r ‘states of affairs’. It now also becom es relatively indifferent w hether we speak of the vehicle of assertion as a sentence o r as a propo sition. It is the structure of the sentence, and the correspon dence of that structure with the structure of some possible state of affairs, which m akes it express a proposition; but by the 3 'One and the same fact may be expressed in a thousand different lan guages, and the thousand different propositions w ill all have the same structure, and the fact which they all express will have the same structure, too. for it is just for this reason that all those propositions express just this particular fact’, op cit., p. 158.
S C H L I C K ON FORM AND CO N T E N T
39
sam e token if the sentence did not express a proposition it would n o t be a sentence but merely a sequence of words. T his account o f the difference between words and sentences, however, confronts us with a problem , which in turn will lead to the second of the general propositions m entioned above, and, incidentally, back to Schlick's reflections on the internal rela tions of colours. W hat enables me to identify, merely by inspect ing the structure of a new sentence, the stale of affairs (fact) which th at sentence expresses? It cannot, clearly, be a conven tional association between that structure and the state of affairs which it expresses, for ex hypothesi no such association has been set up (the sentence w ould not be ‘new ' in the required sense if it had). A nd equally, I cannot discover w hat fact it is that the sentence expresses by carrying ou t any em pirical test o r verifica tion procedure. Em pirical tests m ay indeed establish the truth of a sentence once I know w hat it asserts, but until I know w hat it asserts (w hat fact it expresses) the sentence is merely a string of w ords which sim ply has no p o in t d'appui in the extra-lin guistic world. Plainly, p art of the answ er m ust be that the words which m ake up the sentence locate at least the features of the world which enter into the fact which the sentence expresses. Sentences thus express, in effect, m odes of relationship of the features o f the world denoted by the nam es which enter into their com position. B ut from this it follows that in order to con struct new sentences we m ust know in advance the possible m odes of relationship into which the features in question can enter. Hence the basic objects of reference with which names are correlated m ust stand to one ano th er in an a priori order of possibilities of relationship. T he study of the ways in which words can be com bined or recom bined to form significant sentences is gram m ar: or, if the naked term ‘g ram m ar’ seems to carry too strong a suggestion th at the possibilities of recom bination are determ ined by mere linguistic convention, logical gram m ar. T h e argum ents we have been outlining thus lead to the view that gram m ar (the order of
40
S C H L I C K ON F O R M AND C O N T E N T
possibilities of significant recom bination of words) is a direct reflection of the structure of facts: that is to say, of the natural order of possibilities of com bination of objects (the things with which basic names are correlated) in states of affairs. T his position, while very m uch simplified, approxim ates in outline to the theory of m eaning which W ittgenstein puts for ward in the Tractatus. ‘Logical space”1 in T ractarian term s is the order of possibilities of relationship (in sachverhalten) be tween the objects of reference (dingen)1 of basic nam es in a language. T he fundam ental relationship linking language with the world holds, not between nam es and their referents, but between a sentence as a structured sequence of signs (a sequence of signs exhibiting, in W ittgenstein's term s, a logical form* and so m t group o t things in the world standing in some possible set of relationships to one another, so that the relationships holding between the m em bers of the group (the structure of the fact, o r better, state of affairs) is m irrored by the structure (the logical form) of the sentence. This, I think, is w hat W ittgenstein means by the second sentence of the Tractatus: ‘T he world is the totality of facts, not of things.’® T he logical priority of ‘facts’ as against ‘things* is guaranteed in the Tractatus by the fact th at things cannot be conceived independently of their possibilities of relationship within states of affairs. If I know an object I also know all its possible occurrences in states of affairs (Every one o f these possibilities m ust be part of the nature of the object) A new possibility cannot be discovered later (2.0123) T h e relationship between an object and its possibilities of occurence in states of affairs is thus internal. It is im portant to see, however, th at W ittgenstein is not saying th at the relatione L. Wittgenstein, T ractatus L ogico-P hilosophicus, tr. D . F. Pears and B. F. M cGuinness, Routledge and Kcgan Paul, 1961. 7 Op. cit. 2.01. 8 Op. cit. 3.14-3.142. »O p. cit. 1.1.
S C H L I C K ON F O R M AND C O N T E N T
41
ships which link objects into states of affairs are internal rela tionships for if this were so it would remove the possibility of asserting contingent propositions about states of affairs. The sort o f things W ittgenstein has in mind when he says that possibilities of com bination are intrinsic properties of object is further explained in 2.013-2.0141. 2.013
Each thing is, as it were, in a space of possible states of affairs. T his space I can im agine em pty, but I can not im agine the thing w ithout the space 2.0131 A spatial object must be situated in infinite space. (A spatial point is an argum ent-place). A speck in the visual field, though it need not be red, must have some colour: it is, so to speak surrounded by colour-space. Tones m ust have som e pitch, objects of the sense of touch som e degree of hardness, and so on. 2.014 O bjects contain the possibility of all situations 2.0141 T h e possibility of its occurring in states of affairs is the form of an object.
T he nature of objects is not further specified in concrete terms, and objects are never identified with any particular class of objects of experience. T he distinction between an order of relationships in the world which guarantees the possibility of the states of affairs expressed by sentences in a language, and the circum stances which establish the truth o r falsity of a given sentence, leads directly to the celebrated— indeed notorious—T ractarian dis tinction between w hat can be said, o r asserted, about the world, and w hat can only be show n by o ur use of language. If a sentence— o r rath er its structure, o r logical form, expresses a given state of affairs, according to W ittgenstein, then w hat the sentence states is th at that state of affairs obtains in the world, o r ‘is the case’. T he sentence is true just in case the state of affairs does obtain (is ‘the case’). T h e relationship of possibility
42
S C H L I C K ON F O R M A ND C O N T E N T
of com bination between dingen (basic objects of reference) which the sentence reflects in its own structure, is not, however, asserted by the sentence, but merely show s itself in the fact that we can seriously raise the question of the truth o r falsity of the sentence. If, now, we try to construct a sentence which asserts this relationship between dingen— which says som ething like 'dingen can be related in the following way, am ong others, viz, ------ ’, we run into an obvious theoretical difficulty. We can only refer to dingen by nam ing them. We can only identify them for the purposes of nam ing them by reference to the relationships which can obtain between them. W e can only refer to such a relationship by constructing a syntactic model of the relation ship in question; th at is, a sentence which m irrors the relation ship in its syntactic structure. But now we have o u r original sentence, which shows but does not assert the relationship which guarantees its possibility (its ‘m aking sense’); and not a t all the sentence we w anted. In the end we can only gesture tow ards the relationships between dingen which constitute the fabric of initial possibilities on which in turn the possibility of assertion rests; or, w hat comes to the sam e thing, exhibit (show) their existence by the very sureness with which we construct and interpret new sentences in o u r language.
II It is notorious that W ittgenstein never identifies dingen with any definite class of objects of experience, despite the fact that dingen are conceived as the referents of nam es in fully analysed sentences, and th at fully analysed sentences m ust presum ably stand in some relationship to w hat we actually experience. Dingen, in the Tractatus, rem ain in a certain sense purely theoretical objects: th at is, they are postulated in order to satisfy certain theoretical requirem ents of the theory of m eaning (that there should be ‘definiteness of sense’; th at we should know a
S C H L I C K ON F O R M AND C O N T E N T
43
priori and with certainty the possibilities of recom bining sentence-elem ents to form new significant sentences) and they are specified no further than is absolutely necessary in order for them to satisfy these requirem ents. O ddly enough, Schlick s discussion in Form and C ontent, profoundly T ractarian though it is, contains no theory of objects. T he whole problem of how nam es refer, and what it is that they refer to, receives very little m ention. Schlick does how ever, cleave firmly to the T ractarian distinction between, on the on hand, an order o f external relationships between things in the world which constitute the ‘structure* which contingent propositions ‘express' (in T ractarian terms, ‘m irror’); and on the other hand, an order of internal relationship which deter mine logical gram m ar, o r the possibilities of com bining names in discourse. It is clear that in speaking of colours o r other ‘qualities’ we can refer to them only as external properties of something: we have to define a certain flavour as the sweetness of sugar, a certain colour as the green of a m eadow . . . . . . it becom es evident th at propositions express facts in the world by speaking o f objects and their external properties and relations. A nd it would be a serious m isunderstanding . . . if you believed that propositions could speak o f logical structures o r express them in the same sense in which we speak of objects and express facts. Strictly speaking none of our sentences ab o u t the green leaf expresses the internal structure of the green; nevertheless it is revealed by them in a certain way, o r— to use W ittgenstein’s term — it is shown forth by them . T h e structure of ‘green’ shows itself in the various possibilities o f using the word *green\ it is revealed b y its gram m ar. A language does not, of course, express its own gram m ar, but it shows itself in the use of language. If the internal relationships which relate colour presentations
44
SCH LI CK ON FORM AND C O N T E N T
determ ine ‘the various possibilities of using the w ord green’, it might reasonably seem to follow that Schlick identifies colour presentations— that is colour content, o r qualia, with W ittgen stein’s dingen. N othing could be further from Schlick’s mind. W hat Schlick wishes to show is not merely th at content is inexpressible, but that there is no such thing as content; that the very word ‘content’ is w ithout meaning. T here is no proposition about content, there cannot be any. In other words: it would be best not to use the word ‘c o n te n t’ at all, there is no need for it, and my only excuse for using the word (even in the title of these lectures) is that this for bidden road seemed to me to be the easiest way of taking the reader to a point which will allow him to get a first view of the land before him. He will now be able to turn back and find the right road which will actually take him to the promised land. I shall continue to use the term ‘co n ten t’ now and then, but the reader will understand that a sentence in which this word occurs m ust not be regarded as a proposi tion about som ething called ‘co n ten t’ but as a sort of abbreviation of a m ore com plicated sentence in which the word does not occur.10 This, in a way, is a tem pting way out. W ittgenstein’s account of dingen is m etaphysical and obscure; it seems impossible to connect them with anything in o u r experience. W hat better than to transm ogrify them into internally related contents and then somehow reductively analyse them out of existence? T here are, of course, severe difficulties about the introduc tion of colour contents to fill the theoretical role played in the Tractatus by dingen: that of accounting for the possibilities of com bination exhibited by nam es in discourse. D ingen, like colour contents, are indeed internally related, but the internal relationships in question seem to be of radically different types. « O p . cil.. p. 176-7.
S C H L I C K ON F O R M AND C O N T E N T
45
D ingen are internally characterized by their possibilities of com bination in states of affairs. C olour contents are internally related into sequences of hue, o r relative brightness or darkness, in virtue of their qualitative character. It is very difficult to see how these relationships could possibly govern the ‘gram m ar’ of a colour w ord as Schlick suggests, and Schlick does not elabor ate o r clarify his suggestion. A nd yet Schlick has accepted so m uch of the argum ent that leads W ittgenstein in the Traetatus to postulate dingen that he can hardly do w ithout them. His way out of this apparent im passe is to claim that the argum ent itself is a tissue of trivi alities. ‘ In order to be able to describe the world we m ust be able to speak of all possible facts including those which do not exist at all, for language m ust be able to deny their existence. O ne m ight think that in saying this we are m aking rather bold a priori statem ents about the world. F or are we not implying that all possible things o r events in the world must conform to certain conditions, m ust possess a certain kind of order which will enable us to grasp them by m eans of our expressions? A nd would this not m ean a m etaphysical pre supposition which could never be justified? It is of the highest im portance to see that in m aintaining that facts m ust have a structure we arc not m aking any pre suppositions ab o u t the facts at all, we are saying only that facts are facts, which is, as will probably be adm itted, saying nothing ab o u t them .11 All th at Schlick has earlier said about the logical structure of colour is thus dismissed as vacuous or tautologous. The intrinsic im plausibility of the claim is som ew hat m itigated by Schlick’s prom ise to show that all talk of content is a mere façon de parler, useful only in so far as it m ay help to carry the 11 Op. cit. p. 158.
SCH LI CK ON FORM AND C O N T E N T
46
reader to a point from which he may clearly see the vacuity of such talk. Proposition 6.54 of the Tractatus12 is clearly in the back of Schlick’s mind. But the resolute paradoxicality of 6.54 is foieign to Schlick’s straightforw ard m ind, and his words on pp. 176-7, which I quoted a few pages ago, suggest som ething m ore on the lines of a reductive analysis of sentences involving the notion of con tent. And this is, in the end, w hat Schlick offers. T o understand the turns of his thought at this point we m ust rem em ber that structure for him m eans not only the internal relationships of ‘logical structure’, but the external relationships which con stitute facts. It is only the latter that propositions can express, so that a fortiori all propositions are contingent propositions. So far as Schlick fills in, later on in the lectures, the claim that there can be no propositions ab out content, it would appear that he intends it to be construed as follows. A sentence is meaningless unless we can ‘define the use' of the symbols it contains.13 Defining a symbol involves exhibiting the rules according to which it is to be used, and ‘These rules m ust be taught by actually applying them in definite situations, th at is to say, the circum stances to which they fit m ust actually be shown . . . And this can be done only by certain acts, as for instance gestures, by which o u r words and expressions are cor related to certain experiences.’14 N ow , to correlate a sentence with the world is to place it in just such a relationship with some fact that between it and the fact in question there exists that structural parallelism which Schlick regards as constitutive of the sem antic function of the sentence. But to say this am ounts precisely to saying that to give a meaning to a sentence is always to give it a m eaning in term s of structure. ‘A proposi w
12 ‘M y propositions serve as elucidations in the follow ing w-ay: anyone w ho understands me eventually recognizes them as nonsensical, when he has used them—as steps—to climb up beyond them. (H e must, so to speak, throw away the ladder after he has clim bed up it).’ 18 Op. cit., p. 180. 1« Op. cit.. p. 180-t.
S C H L I C K ON FORM AND C O N T E N T
47
tion will be verified, the truth will be established if the structure is the sam e as the structure of the fact it tries to express.’1“ Hence no sentence can have as its truth conditions any state of ‘c o n te n t’, and hence there can be no ‘propositions about con tent’; and any sentences in ordinary language which apparently express propositions of this sort do so in appearance only and are in fact meaningless. We have thus arrived at a new version of the inexpressibility thesis. But Schlick's argum ent seems to me to have serious defect. How can we place sentences in structural correlation with facts? How do we know w hat ‘structure’ in the world a given sentence picks out? A sentence, after all, is no more than a collocation of words. T he obvious answer which springs to m ind, I think, is that when we place the nam es in a sentence in correlation with things in the w orld, the external relations in which those things stand to one ano th er can be represented in various ways by the external relations between the words in the sentence. T his is very m uch the direction Schlick takes. A ccording to Schlick, a proposition is ‘given a logical structure and a m ean ing . . . by assigning definite significations to its parts’, and that, to see how this is done, ‘we m ust im agine all the words of the sentence to be replaced by their definitions, the terms occurring in the definition m ust be replaced by sub-definitions and so on until we reach the boundary of ordinary verbal language when it ends in gestures o r prescription to perform certain acts.’18 In other words, a proposition acquires logical structure and m eaning when we give, by m eans of deictic gestures, ostensive definitions of the words which it contains. W ittgenstein says very sim ilar things in the Tractatus. 3.202 T h e sim ple signs em ployed in proposition are called names. 15 O p. cit., p. 228.
’ « O p . cit., p. 228.
48
S C H L I C K ON F O R M AND C O N T E N T
3.203 A nam e m eans an object. T he object is its meaning.