WOLFGANG HEYDRICH

A RECONCEPTION

OF MEANING s

The problems [the paper 'On Likeness of Meaning'] deals with do not seem to me to have quite the paramount importance that is commonly attached to them these days. (Goodman, 1954, p. 54) The subversive power of a radical reorganization of categories should not be underestimated. (Goodman, 1983, p. 2) 2 ABSTRACT. Nelson Goodman's proposal for a reconception of meaning consists in replacing the absolute notion of sameness of meaning by that of likeness of meaning (with respect to pertinent contexts). According to this view, synonymy is a matter of degree (of interreplaceability) with identity of expression as a limiting case. Goodman's demonstration that no two expressions are exactly alike in meaning is shown to be unsuccessful. Although it does not make use of quotational contexts for the test of interreplaceability, it .is tantamount to their acceptance. Goodman rejects quotational contexts; I argue that they should be accepted. This move offers two advantages. Firstly, and mainly, it allows interlinguistic comparison of meaning, something that has not been deemed possible in the received version of Goodman's account. Secondly, it restores the full scale of likeness of meaning damaged by the renunciation of those contexts that guarantee difference in meaning for diverse expressions.

Reconceptions require problems and projects. Often the starting point is a problematic concept or family of concepts, and the project is to point to the direction of a less problematic account. Pretheoretic and traditional conceptions of meaning are problematic in several respects. The ontological and epistemological status of meanings taken as obiects is insecure and their conditions of identity are unclear. Whereas semantic relations like synonymy (sameness in meaning), hyponomy (subordination in meaning), and antinomy (contrast in meaning) seem to offer the prospect of dispensing with talk of meanings as objects, these notions themselves turn out to be notoriously hard to define. Some philosophers of language have argued that there is a sharp contrast and unbridgeable gap between the theory of reference and extensions on the one hand and the theory of meaning and intensions on the other.

Synthese 95: 77-94, 1993. © 1993 Kluwer Academic Publishers. Printed in the Netherlands.

78

WOLFGANG HEYDRICH

Here now is what I think is Goodman's project of reconception with respect to meaning and meaning-relations.

1.

GOODMAN'S PROJECT

I cite a rather early formulation of what seems to me the central idea of Goodman's account. It is a quotation from Quine's essay 'The Problem of Interpreting Modal Logic' (1947, ft.4). This paper was published in 1947, thus antedating Goodman's own papers 'On Likeness of Meaning' (1949) and 'On some Differences about Meaning' (1953). Quine writes: Dr. Nelson Goodman has suggested (in conversation) the dismal possibility that what we think of as synonymy may be wholly a matter of degree, ranging from out-and-out orthographical sameness of expressions on the one hand to mere factual sameness of designation (as in the case of "nine" and "the number of the planets") on the other. (Quine, 1947, p. 45)

What might very well be a 'dismal possibility' with regard to the problem of interpreting modal logic can nevertheless provide a more pleasant view on matters of meaning. And I suspect that it is more or less along the lines of the project formulated here that Goodman developed his ideas with respect to meaning. I say 'more or less' since as we shall see - later on Goodman seemed to be somewhat hesitant about whether sameness of expression should really be accepted as a (limiting) case of a semantic relation. Anyhow, the core idea is that to deal with questions of meaning requires a reconception of synonymy as a matter of different degrees and, as may be added, different kinds: Goodman proposes to substitute the unqualified notion of sameness of meaning by that of likeness of meaning. In contrast to the former, the latter comes with a hidden parameter: likeness of meaning must always be judged with respect to a relevant discourse and pertinent context. The need for this substitution is motivated in a concise passage inserted into 'On Likeness of Meaning' (1952)3: [I]t is commonly supposed that a satisfactory definition of synonymy must meet two requirements: that some predicates be synonymons with others, and that either of a pair of synonyms be replaceable by the other in at least all non-intensional contexts without change of truth-value. B u t . . . these two requirements are incompatible. The sound course seems to be to construe degree of synonymy as, so to speak, degree of interreplaceabili-

A R E C O N C E P T I O N OF M E A N I N G

79

t y . . . and to recognize that the relation of exact synonymy between diverse predicates is null. (Goodman, 1972, pp. 229-30)

Given some relevant discourse, each non-intensional context appropriate for the test of interreplaceability relative to that discourse provides a specific kind of likeness of meaning, and several such kinds accumulate to degrees of synonymy.4 If Goodman uses the term 'interreplaceability', he thinks of interreplaceability salva ordinary extension. Speaking about non-intensional contexts, he excludes contexts like 'thought of ( a ) . . . ' , 'concept of ( a ) . . . ' , 'attribute of ( a ) . . . ' , and 'meaning of ( a ) . . . ' , i.e., contexts which (with a name or noun inserted) become compound predicates denoting intensional entities like thoughts, concepts, attributes, and meanings. He clearly does not exclude contexts like 'picture of ( a ) . . . ' , 'description of ( a ) . . . ' , or 'is hunting ( a ) . . . ' , i.e., contexts that become compound predicates denoting ordinary objects like pictures, inscriptions, and persons after insertion. This last point needs mention, since there is a broader - and perhaps more common - use of 'intensional', which covers both kinds of contexts, since in each case the ordinary extension of the noun to be filled in does not determine the extension of the predicate as a whole. Goodman's explications of likeness of meaning take into account only ordinary extensions (no Fregean senses, Carnapian intensions, or the like). Thus, his project might be called extensionalistic. But he does not exclude from consideration each and every context that is intensional in the broader sense (only those that are intensional in the more restricted sense). Thus, the project might be called intensionalistic as well. This is, it seems to me, how Goodman proposes to link the extensional and the intensional and tries to deal with the theory of meaning within the bounds of the theory of reference. Now, it should be stressed that Goodman did not - and probably did/does not intend to - develop a comprehensive theory of meaning for natural languages, 5 nor did he design a semantic description of at least a restricted fragment of some natural language (like, e.g., Montague's fragment of English in 'The Proper Treatment of Quantification in Ordinary English'). Instead, his contributions concentrate on olaf-

80

WOLFGANG

HEYDRICH

ification of the fundamental concept of semantic equivalence (likeness of meaning) as applied mainly to one-place predicates 6 and names. Additionally, there is a certain preoccupation with the second part of the 'sound course' recommended in the passage quoted a b o v e , 7 i.e., to recognize: (1)

The relation of exact synonymy between diverse predicates is null.

In Goodman's approach this has to be defended by arguing: (2)

For each pair of diverse predicates P and Q there is (for some type of discourse) a pertinent context C not allowing the extensional interreplaceability of P and Q.

Since (2) follows logically from the following thesis, it is sufficient to show: (3)

There is (for some type of discourse) a pertinent context C such that, for each pair of diverse predicates P and Q, C does not allow the extensional interreplaceability of P and Q.

The truth of (3) gives us, so to speak, zero-degree difference in meaning, kinds of likeness of meaning according to which each predicate is semantically equivalent exclusively with itself. This construes sameness of expression as one of the two limiting cases of synonymy along the lines of the project formulated in the quotation from Quine's paper. (The other limiting case is provided by extensional contexts, whose denotations are determined by the extensions of the expressions to be inserted.) Obviously, everything depends on which contexts are considered to be pertinent. If, e.g., contexts like (4)

qnscription of 'X '~

are admitted as pertinent with respect to some discourse, (3) turns out to be trivial, and likewise (2) and (1). Note that (4), although intensionat in the broad sense, is not intensional in the narrow sense, since after insertion of an arbitrary expression for 'X' it becomes a compound predicate which denotes ordinary objects, namely, inscriptions.

A RECONCEPTION

81

OF MEANING

In 'On Likeness of Meaning' and some passages of 'On some Differences about Meaning' Goodman tries to defend (3) in a way that does not make it trivial. But at the end of the latter paper he remarks that contexts warranting the truth of (3) are somewhat exceptional anyhow, whether they are quotational or not. Thus, he advocates complete exclusion of such contexts and abandons (3). This move requires an independent argument for (2), if the result (1) is to be maintained. Additionally, it gives up those kinds of likeness of meaning that provide the limiting case of zero-degree difference in the original project. In what follows, I want to argue, firstly, that Goodman's proposals to secure the truth of (3) in a non-trivial way are unsuccessful, and, secondly, that this should not lead him to the abandonment of (3), but to the acceptance of its triviality: I advocate the inclusion of quotational contexts like (4). The advantage is that the limiting case of zero-degree difference need not be given up, and - more importantly perhaps that the relation of likeness of meaning can not only be applied intralinguistically (between predicates of one and the same language) but interlinguistically as well (between predicates of different languages). So, let me explain this in more detail.

2.

wHY

QUOTATIONAL

CONTEXTS

SHOULD

NOT

BE

EXCLUDED

In 'On Likeness of Meaning' Goodman makes use of inscriptions of form (5) in order to show the non-trivial truth of (1): (5)

rp that is not a Q~

He claims that each inscription of form (5) is a P-description, but not a Q-description, if P and Q are different. Thus, according to this

first proposal, (6)

rP-description~

is a context that maximally differentiates meanings and verifies (3). Obviously, there are clear cases that support Goodman's claim: an inscription of (7)

oculist that is not a dentist

82

WOLFGANG

HEYDRICH

describes oculists and not dentists; additionally, and more to the point, it is an oculist-description and not a dentist-description. Note that inscriptions describing oculists are not bound to be oculistdescriptions. If, e.g., all physicians living in my neighbourhood are dentists or oculists, an inscription of (8)

physician living in my neighbourhood that is not a dentist

may very well describe oculists without being an oculist-description. Rather, it will be, say, a physician-living-in-my-neighbourhood-description. A P-description need not be a Q-description only because all Ps are Qs. Accordingly, Goodman stresses that a centaur-description such as (9)

centaur that is not a unicorn

is not a unicorn-description, although (because of their non-existence) all centaurs are unicorns. There are cases, however, of all Ps being Qs such that all P-descriptions are Q-descriptions as well. For example: not only are all physicians living in my neighbourhood physicians, but all physicians-living-in-myneighbourhood-descriptions are physician-descriptions as well - or so it seems. 8 What about the inscriptions of expression (7)? Are they physiciandescriptions as well? Here our intuitions begin to fade away. I tend to answer yes. Are inscriptions of expression (7) eye-doctor-descriptions? I hesitate, but tend to answer yes again. Are all oculist-descriptions eye-doctor-descriptions? I hesitate still more, but again tend to answer affirmatively.9 Goodman, however, says no! For him, each description of (10)

oculist that is not an eye doctor

is an oculist-description and not an eye-doctor-description. This is due to some stipulation concerning the applicability of predicates of form (6) in cases where intuition does not guarantee unambiguous results. With respect to the question whether inscriptions of form (11)

rp that is not a P~

A RECONCEPTION

OF MEANING

83

are P-descriptions, Goodman writes that direct appeal to ordinary usage may yield no firm decision . . . . [In order to decide the question,] we must formulate rules that fit ordinary usage where it is clear and that can be projected to decide these doubtful cases, There is no correct way of doing this, but a reasonable rule covering the present question runs as follows: Any phrase of the form "-that i s . . . " is both a -description and a . . . description; and a not-a-soandso-descripfion is not a soandso-description unless required to be by the first ciause of this rule. (Goodman, 1972, p. 236) l°

In fact, some stipulation seems to be required to avoid the selfdefeating result that, given some predicate P, an inscription of form (11) both is a P-description and is not a P-description. However, Goodman's rules are stronger than necessary. Besides (i)

Any inscription of the form ' - - t h a t is.o .' is both a --description and a...-description,

it is sufficient to stipulate (ii)

The transition from being not-a-P-description to not being a P-description is blocked for inscriptions of logically inconsistent predicates.

This modification prevents inscriptions of form (11) from being both P-descriptions and not P-descriptions, since they are inscriptions of logically inconsistent predicates. Clause (ii) is non-committal, however, about inscriptions of logically consistent predicates. Thus, it does not urge us to admit that any inscription of (10) is an oculist-description and not an eye-doctor-description. To be sure, each inscription of (10) is, according to (i), a not-an-eyedoctor-description; but it is compatible with (ii) that it is an eye-doctordescription, as well as that it is not an eye-doctor-description, since (10) is a logically consistent predicate. In this way the intuition might be respected, that all oculist-descriptions are eye-doctor-descriptions. I do not claim that these modifications are superior to Goodman's original rules. My point is merely that Goodman's stipulations are stronger than required in order to take measures against contradictory results. And it is this additional strength that makes (6) a context of maximal differentiation of meaning. Without it we have no basis for

84

WOLFGANG

HEYDRICH

claiming that inscriptions of form (5) are P-description without being Q-descriptions. Thus, according to this first proposal, the truth of (3) results from a stipulation which is not independently motivated, and may be called ad hoc. There is, however, a second proposal of Goodman to defend the truth of (3) without recourse to quotational contexts. In 'On some Differences about Meaning' he discusses contexts of the form (12)

rliteral English A-word 7,

because he thinks that predicates of this form have more easily specifiable ranges of application than predicates of form (6), which - as we have seen - require more or less complicated stipulations to regulate their applicability in intuitively unclear cases. According to Goodman, contexts like (12) are sufficient to show the truth of (3) as well: (12) "may be taken", he writes, "as applying just to those inscriptions that are tokens of" A (Goodman, 1972, p. 233). Given that different English words correspond to different classes of inscriptions, and predicates of form (12) apply to inscriptions, then, obviously, the extensions of the latter will be different in case different words are substituted for 'A'. This result depends on interpreting (12) just the way indicated by Goodman. 'May be taken', however, suggests that there are other possibilities. In 'Some Questions concerning Quotation' (1978), Goodman himself makes use of analogous predicates of the form (13)

rA-termT.

He requires that these predicates apply not only to inscriptions of the term A but also to its paraphrases, "where a paraphrase of a term preserves not only the primary but also the requisite secondary extensions of that term ''11 (Goodman, 1978, pp. 45-46). In my view, it is far from evident that context (12) should be given the strict interpretation, while context (13) is to be taken more liberally. Accordingly, it is not clear to me why 'literal English oculist-word' must be understood as applying to inscriptions of 'oculist' but not

A RECONCEPTION

OF MEANING

85

to inscriptions of 'eye doctor', whereas 'oculist-term' may apply to inscriptions both of ~oculist' and of 'eye doctor'. Surely, (12) m a y be interpreted the way proposed by Goodman, but, if so, this seems to be just in order to secure the truth of (3). So, once again, rendering (3) true here depends on an otherwise not sufficiently motivated stipulation. Incidentally, it may be remarked that admission of context (12) which is n o t quotational - is, under the strict interpretation proposed, tantamount to admission of context (4) - which is quotational - as long as we restrict X to English words. So, why not go right ahead and admit quotational contexts as extreme cases of pertinent contexts from the start, facing the triviality of (3)? In 'On Likeness of Meaning' no exclusion of quotational contexts is formulated. In 'Splits and Compounds' (1984), Goodman gives a necessary condition for relevant contexts, which likewise does not exclude contexts like (4): IT]he only compounds relevant to meaning are such that the term compounded (say, "unicorn") serves as modifier of the compounding addition (say, "picture of a"), (Goodman, 1984, p. 79)

It might very well be argued that in (4), each compounded expression X functions as modifier of the compounding addition °inscription of ' . . . " . Just like, semantically, the embedding of 'unicorn' into ~picture of a . . . ' restricts the set of all representational pictures to the set of all unicorn-pictures, its embedding into 'inscription of L . . " restricts the set of all inscriptions to the set of all 'unicorn'-inscriptions. In 'On some Differences about Meaning', however, Goodman excludes quotational contexts - unfortunately without explicit argument. But recognizing that the admission of contexts like (6) or (12) makes (3) hardly less trivial than the admission of quotational contexts, he advocates, as has already been remarked above, the exclusion of all contexts that verify (3), thus he abandons (3) altogether. By this "perhaps welcome amendment" (Goodman, 1972, p. 231) to his account, he tries to

86

WOLFGANG

HEYDRICH

honor the feeling or principle that the interesting differences between two terms are just those that are not shared by every two terms. (Goodman, 1972, p. 238)

Now, as before, the truth of thesis (1) that no different predicates are absolutely synonymous is not precluded. But arguing for it can no longer consist simply in constructing one of those exceptional contexts that highlights each and every difference between expressions. Instead, one has to show differences in meaning between every two predicates by means of a number of pertinent contexts such that each of them allows extensional interreplaceability for at least one pair of predicates. To be sure, excluding quotational contexts along with all others that verify (3) has some perhaps less welcome side-effects. H e n d r y (1980) and Elgin (1983) remark that likeness of meaning as explained in 'Some Differences about Meaning' can only be applied intralinguistically. Elgin writes: The requirement that parallel compounds [i.e., those originating from the same context by insertion] be coextensive is quite strong. It is much stronger, for example, than the requirement that each compound of one term be coextensive with some compound of the other. Synonymyrequires.., that there be a point-by-point agreement . . . . But the price of this agreement is high. Synonymy, so construed, is strictly intralinguistic. The definition gives no standard for interlinguistic translation, for the various replacements in compound constructionsmust belong to the same language. The compounds that result from replacing "dog" with "Hund" or "chien" by and large belong to no language. (Elgin, 1983, pp. 55ff.)12 If the price is that high, then one hesitates, wondering whether Goodman's reconception of meaning is worth it. What one gets for giving up the wonderful prospect of a Platonistic paradise of language-independent (although hardly distinguishable) intensional objects is a bundle of equivalence relations for each separate language. Is that a fair exchange? 13 In the passage cited above, however, the exclusion of quotational contexts adopted in 'On some Differences about Meaning' is silently presupposed. Clearly, if (4) is admitted as a pertinent context, we get by insertion (t4)(i)

inscription of 'dog'

A RECONCEPTION

OF MEANING

87

(ii) inscription of 'Hund' (iii) inscription of 'chien' as parallel compounds with different ranges of application. Granted, this demonstration that 'dog', 'Hund', and 'chien' are not absolutely alike in meaning is not very exciting; it simply restates that they are different expressions. In any case, it could not have been given without a quotational context like (4), and it gives us back the limiting case of zero-degree difference in meaning excluded by Goodman. The other limiting case can be accounted for by allowing (15)

rdenotation of 'X '~

as a pertinent context. Let us suppose for a moment that it is only with respect to the limiting cases that we take recourse to quotational contexts, and that (4) and (15) are the only ones adopted. Then, in dealing with cases in between the extremes, we face the problem highlighted by Elgin at the end of the quotation above again: that, e.g., 'oculist' and 'eye doctor' are atike in meaning with respect to the intermediary context (16)

rP-picture~

means that all oculist-pictures are eye-doctor-pictures and vice versa. And that's true, I suppose. What one would like to show additionally, however, is that some such kind of likeness of meaning obtains among the predicates ~dog', 'Hund', and 'chien' as well - in contradistinction, say, to the predicates 'centaur' and 'Einhorn'. One could try to do without further quotational contexts, inserting each of these nouns into context (16) and testing the resulting parallel compounds with respect to their ranges of application. But this does not work. Not, however, because the test provides a negative result; rather, because it is not applicable. Whereas 'dog-picture' and 'centaurpicture', are rather well behaved English predicates, 'Hund-picture', 'chien-picture', and 'Einhorn-picture' by and large do not belong either to English, German, French, or some other language, as Elgin has it. As long as (4) and (15) are the only quotational contexts at our disposal,

88

WOLFGANG

HEYDRICH

they are not of much use for a theory that investigates different kinds of likeness of meaning across different languages. So, let us try to design other quotational predicates. Obviously, (17)

r'X'-picture ~

will not be very helpful. I must confess that I have no very precise ideas about what exactly pictures of expressions are (perhaps some Magrittes are 'pipe'-pictures, and perhaps all inscriptions are expression-pictures), but I am rather sure that, provided expression-pictures exist, it is not the case that, say, all 'Hund'-pictures are 'chien'pictures. 14 A n o t h e r quotational predicate, however, does the job. Scheffler (1979) has called attention to the relation of mention-selection, which obtains between predicates and names on the one side and certain pictures and descriptions on the other. If a predicate or name X is applied to an X-picture or X-description instead of to what X denotes, he calls this use mention-selective and quasi-denotative, and he points to the fact that this kind of applicational practice is not only constitutive for our getting the point about fictional predicates and names but is also quite common and natural in general; A child is also often asked to point out trees, dogs, and automobiles in picture books and magazines. A n d in our own typical labeling of a picture of a man "Man" (rather

than "Man-picture"), we ourselves apply the term "Man" to select not a man but a picture; we here apply the term not to what it denotes but rather to a mention thereof. (Scheffler, 1979, p. 35) Concentrating on the special case of mention-selection, where a predicate is correctly and quasi-denotatively applied to a picture, let us say that, under those circumstances, the picture is a depiction-selection of the predicate. Then, by taking the context (18)

rdepiction-selection of 'X '"

we get what we need. All and only oculist-pictures are depiction-selections of 'oculist', and

A RECONCEPTION

OF MEANING

89

all and only eye-doctor-pictures are depiction-selections of 'eye doctor'. In view of the fact, however, that oculist-pictures are eye-doctor-pictures and vice versa, all and only depiction-selections of 'oculist' are depiction-selections of 'eye doctor'. In the case of Elgin's interlinguistic example, we get that all and only dog-pictures are depiction-selections of 'dog', 'Hund', and °chien'. The difference in language does not matter: whereas the words 'dog', 'Hund', and 'chien' are expressions in English, German, and French, respectively, the pictures they select are in none of these languages. Thus, context (18) does not point to a difference in meaning either in the 'oculist'-'eye doctor'-case or in the 'dog'-'Hund'-'chien'-case. However, in the 'centaur'-'Einhorn'-case, a difference in meaning is highlighted, since there are centaur-pictures that are not unicorn-pictures, and the latter are just the pictures selected by 'Einhorn'. Accordingly, 'depiction-selection of 'centaur" and 'depiction-selection of 'Einhorn" have different ranges of application. Consequently, context (18) does just what it is supposed to do. It discriminates intra- and interlinguistically just the way (16) only discriminates intralinguisticatly. Corresponding to other cases of quasi-denotation we may introduce (19)

rdescription-selection of 'X '~.

Both (18) and (19) discriminate meanings more finely than (15) and less finely than (4), provided, of course, we do not adopt such a strict regimentation for (19), like the one advocated by Goodman for (6). ~s If we introduce rquasi-denotation of 'X 'n for the genera! case including (18) and (19) and - just in order to get a uniform terminology - rautodenotation of 'X 'n for (4), we get the following sequence of contexts that highlights decreasing degrees of difference and increasing degrees of likeness of meaning: (20)(i) Fdenotation of 'X '1 l( Cdescfiption-selection of 'X '7 (ii) ~quasi-denotation of 'X '~ / Fdepiction-setection of 'X '7 (iii) rauto-denotation of 'X'" (i) and (iii) provide the limiting cases of a notion of synonymy that, ~uoting Quine again, turned out to be "wholly a matter of degree,

90

WOLFGANG

HEYDRICH

ranging from out-and-out orthographical sameness of expressions.., to mere factual sameness of designation". (ii) is the most interesting case between the extremes. It is well motivated by the pervasive use of mention-selective application of expressions in natural language. The question might be asked here whether these five (quasi-)semantic relations between expressions and ordinary objects provide a sufficiently rich basis for the systematic development of a semantic theory for natural languages (or some fragments of them, or, at least, some artificial languages sharing interesting intensional features with natural languages 16) along the lines of Goodman's reconception of meaning. 3. SUMMARY In summary, we can say that admitting quotational contexts seems to provide two advantages. Firstly, and mainly, it allows interlinguistic comparison of meaning, something that has not been deemed possible in the received version of Goodman's account. Secondly, it restores the full scale of kinds of likeness in meaning damaged by the renunciation of pertinent contexts that guarantee difference in meaning for diverse expressions. The claim that there are no two absolute synonyms in natural language becomes a triviality. 17 This, however, should not be seen as a disadvantage. Rather, it seems to me quite natural that, once the old central notion of absolute synonymy is superseded (or 'aufgehoben', as Hegelians might have it) within an account of degree of interreplaceability, its new place is marginal and just a limiting case. Limiting cases are powerful theoretical devices. Admittedly, they are often somewhat at odds with pretheoretic use or intuition. Take the example of the reconception of the part-of relation in mereology. We start with intuitions such that for each individual there are some others it is part of and still others it is not part of. But developing the account systematically, we may find it convenient and theoretically fruitful to postulate one individual that is part of no individual (except itself) (the fusion of everything) and one individual that is part of every individual (including itself) (the null-individual), is Something not too different happens in the case of the reconception of the relation of synonymy in the theory of likeness of meaning. We start with intuitions such that for each expression there are some

A R E C O N C E P T I O N OF M E A N I N G

91

others it is coextensive and synonymous with, and still others it is coextensive and n o t synonymous with. But developing the account systematically, we may find it convenient and theoretically fruitful to postulate o n e kind of likeness such that, according to it, each expression is synonymous with n o coextensive expression (except itself), and o n e kind of likeness such that, according to it, each expression is synonymous with e v e r y coextensive expression (including itself).

NOTES 1 Thanks to Oliver R. Scholz and Lorenz Lorenz-Meyer for valuable discussions and advice. Thanks to Geoff Simmons for checking my English. Israel Scheffter and Rolf Eberle both read a preliminary version of this paper. My thanks for their criticism and encouragement. The first quotation is from the foreword to the 1954 version of 'On Likeness of Meaning'; the second from the foreword to Elgin (1983). 3 First published in Linsky (1952) and retained in Goodman (1972, pp. 221-30). 4 Perhaps this r6sum6 goes a little bit beyond the letter of Goodman's writings; however, it gets the gist, I hope. As for the idea that there are different degrees and different kinds of iikeness of meaning, cf,, e,g., Elgin (1983, pp. 54-58). As for the idea that what is a pertinent context may depend in part on type of discourse, cf. Goodman's 1954 foreword (pp. 203-06) to 'On Likeness of Meaning', retained in Goodman (1972, pp. 221-30). How to calculate degrees of synonymy of expressions given their kinds of likeness of meaning in several types of discourse is a question which, as far as I see, has not yet been discussed. 5 Since 'On Likeness of Meaning', Goodman's own interests went more in the direction of the general theory of systems, representations, and representational systems than in that of semantics for natural language. It is interesting, however, that his ideas have not been taken up by others, linguists or philosophers of language (with - as far as I see the only exceptions being Eberle (1978a; 1978b), Brown (i984), and Heydrich (1985)), in order to develop a viable semantic theory, which could provide an extensionalistic and ontologically parsimonious alternative to possible worlds semantics, situation semantics, DRT, or the language-of-thought account. Perhaps the cues provided by 'On Likeness of Meaning' and 'On some Differences about Meaning' were not sufficient to inaugurate such a direction of research. Perhaps, as well, Goodman's 1954 foreword to 'On Likeness of Meaning' (see the first epigraph) had a rather discouraging effect. One of the problems here concerns the principle of compositionality: according to Goodman's account, it might be argued, the denotation of a compound expression will not always be determined by the semantics of its components (and its structure); rather the other way round: the semantics of an expression depends on the denotation of (some of) its compounds (and their structures)~ What 'unicorn' means is explained (in part) by means of the denotation of 'picture of a unicorn' (given some or several syntactical parsings of that compound) - not vice versa. This turning things upside down might be seen in conflict with a semantic approach that, in following the principle of compositional-

92

WOLFGANG HEYDRICH

ity, postulates semantic rules mimicking syntactic composition. (Some aspects of this problem have been taken up in the exchange between Wollheim (1970), Ebefle (1978a), and Goodman (1972, pp. 122-25; 1984a, pp. 77-80).) According to this view, a systematic semantics elaborating Goodman's basic idea has to be - at least in part - non-Fregean. (By the way, it seems to me far from easy to formulate a knock-down argument against the possibility of such an elaboration by consideration of questions of learnability. On the contrary, there seems to be prima facie evidence that grasping what 'unicorn' means presupposes knowing how to apply 'picture of a unicorn', 'description of a unicorn', and predicates like that, rather than the other way round. (Cf. Goodman (1968, pp. 24-5) and Scheffler (1979, pp. 34-6).) Perhaps, it is reasonable to reconceive matters in such a way that compositionality interacts with some converse of that principle.). However, as Roll Eberle has pointed out to me in correspondence, things are not quite clear-cut. Eberle stresses the difference between the meta-language use of 'unicorn-picture' (in order to explicate what 'unicorn' means in the object-language) and the use of 'picture of a unicorn' within the object-language itself. Given some way, then, of specifying the semantics of 'picture of' and 'a unicorn' (or of 'picture of a' and 'unicorn') along the lines of Goodman's ideas, the semantics of 'picture of a unicorn' might be obtained entirely compositionally. 6 Goodman's examples of predicates in 'On Likeness of Meaning' and 'On some Differences about Meaning' are primarily non-relational nouns. 7 However, this preoccupation does not, according to Goodman himself, coincide with his primary goal: "Proving that every two terms differ in meaning is not part of my primary goal. The paramount problem is to deal with comparisons of meaning without reference to intensions, attitudes, or modalities" (1972, p. 233). s Cf. Goodman (1972, p. 236): "'[I]sosceles triangle', 'triangle with angles totalling 110 degrees', and 'triangle that is not a trilateral' are all triangle-descriptions according to ordinary usage"i Even "triangle that is not a triangle" is, according to Goodman, a triangle-description. 9 And so does Eberle (1978b, p. 366). lo Analogous rules were added in a footnote to the 1952 version of 'On Likeness of Meaning' - presumably in reaction to criticism like that advanced by Church (1950). In order to get the intuitively welcome result that any inscription of 'physician that is a dentist' is both a physician-description and a dentist-description, one should perhaps add the rule that each a-P-description is a P-description. n A requisite secondary extension of A is the extension of a compound expression formed by inserting A into a context which is pertinent to the discourse within which A occurs. If rpicture of A 7 is such a pertinent context and the term under discussion is 'unicorn', then all pictures of unicorns are in the secondary extension of 'unicorn'. 12Hendry (1980) observes that there is a substantial theoretical difference between 'On Likeness of Meaning' and 'On some Differenees about Meaning'. In the first paper, Goodman defines: "[T]wo terms have the same meaning if and only if they have the same primary [i.e., ordinary] and secondary [see Note 11] extensions" (1972, p. 227). This corresponds to Elgin's less strong and insufficient requirement. In the second paper, however, Goodman strengthens it to "point-by-point agreement in both primary and secondary extensions" (Elgin, 1983, p. 55). The 'On Likeness of Meaning'-version of the theory does not prohibit interlinguistie application. As demonstrated by Hendry (1980,

A R E C O N C E P T I O N OF M E A N I N G

93

p. 323), it has, however, the consequence that, contrary to what Goodman intended to show, each two coextensive predicates have the same meaning! ~3 There are, however, Platonists as well who, in their endeavour to defend intensional objects, are willing to pay the price of intratinguistic restriction. See, e.g., Ktinne (1983, p. 254). a4 If all 'Hund'-pictures are 'chien'-pictures, because there are no expression-pictures (hence no 'Hund'-pictures), this need not bother us. In that case, there are no 'centaur'pictures either, all such pictures are 'Einhorn'-pictnres, and context (i7) will not discriminate properly. ,s In that case, (19) would just do the work of (4). ,6 Eberle's proposals (in 1978b) seem to go in that direction. Cf. as well Heydrich (1985). 17 The claim is, at least among linguists, almost a commonplace, anyway. Cf. Gauger (1972). ~s To be sure, Goodman's own mereological account (i.e., his calculus of individuals cf. Goodman and Leonard (1940) and Goodman (1951)) does not countenance the nullindividual. See, however, alternative approaches like Martin's (1943).

REFERENCES Brown, B. E.: 1984, Representational Semantics, Ph.D. dissertation, University of Rochester, New York. Church, A. : 1950, Review of 'On Likeness of Meaning', Journal of Symbolic Logic 15, 150-51. Eberle, R. A.: I978a, 'Goodman on Likeness and Differences of Meaning', Erkenntnis 12, 3-16. Eberle, R. A.: 1978b, 'Semantic Analysis without Reference to Abstract Entities', The Monist 61, 363-83. Elgin, C. Z.: 1983, With Reference to Reference, Hackett, Indianapolis. Gauger, H.-M.: 1972, Zum Problem der Synonyme, Ttibinger Beitr~ige zur Linguistik, Tt~bingen. Goodman, N.: 1949, 'On Likeness of Meaning', Analysis 10, 1-7. Goodman, N.: 1951, The &ructure of Appearance, Harvard University Press, Cambridge (3rd ed.: 1977, D. Reidel, Dordrecht). Goodman, N.: 1952, 'On Likeness of Meaning', in L. Linsky (1952), pp. 67-74. Goodman, N.: 1953, 'On Some Differences about Meaning', Analysis 13, 90-96. Goodman, N.: 1954, 'On Likeness of Meaning', in M. Macdonald (ed.), Philosophy and Analysis, Blackwett, Oxford, pp. 54-62. Goodman, N.: 1968, Languages of Art, Bobbs-MerriI1, Indianapolis (2nd ed.: 1976). Goodman, N.: 1972, °On Likeness of Meaning', in N. Goodman, Problems and Projects, Bobbs-Merrill, Indianapolis, pp. 221-30. Goodman, N.: 1978, 'Some Questions Concerning Quotation', in N. Goodman, Ways of Worldmaking, Bobbs-Merrill, Indianapolis, pp. 41-56. Goodman, N.: 1983, 'Foreword', in C. Z. Elgin, With Reference to Reference, Hackett, Indianapolis, pp. 1-2. Goodman, N.: 1984, 'Splits and Compounds', in N. Goodman, Of Minds and other Matters, Harvard University Press, Cambridge, pp. 77-80.

94

WOLFGANG HEYDRICH

Goodman, N. and H, S. Leonard: 1940, 'The Calculus of Individuals and its Uses', Journal of Symbolic Logic 5, 45-55. Hendry, H. E.: 1980, 'Nelson Goodman's Two Theories of Meaning', Philosophical Studies 38, 321-24. Heydrich, W.: 1985, 'Logic of Representation', presented at the conference "Texts, Dialogues, and Interpretation", University of Pec, Hungary, unpublished. KOnne, W.: 1983, Abstrakte Gegenstdnde,Semantik und Ontologie, Suhrkamp, Frankfurt a.M. Linsky, L. (ed,): 1952, Semantics and the Philosophy of Language, University of Illinois Press, Urbana. Martin, R. M.: 1943, 'A Homogenous System for Formal Logic', Journal of Symbolic Logic 8, 1-23. Quine, W. V. O.: 1947, 'The Problem of Interpreting Modal Logic', Journal of Symbolic Logic 12, 43-48. Scheffler, I.: 1979, Beyond the Letter, Routledge & Kegan Paul, London. Wotlheim, R.: 1970, 'Nelson Goodman's Languages of Art', Journal of Philosophy 67, 531-39. Germanisches Seminar Universit~it Hamburg Von-Melle-Park 6 D-2000 Hamburg 13 Germany