Brandom

10/7/09

Topics from “Grammar and Existence” and Plan for Seminar

1. In NS, WS addressed the suggestion that ‘Fa’ should be read as ‘a exemplifies F-ness’. GE addresses “categorizing contexts”: ‘F-ness is a quality’. It offers a metalinguistic expressivism about them (WS’s “syntactic strategy”, taken from Carnap). GE then further addresses the principal objections to Carnap’s flat-footed version of the strategy.

Plan of the essay (from Section XIV):

“I began by arguing that 'existential quantification over predicate or sentential variables' does not assert the existence of abstract entities. [This is (2) below, which takes us through Section X.] I then suggested that if the only contexts involving abstract singular terms of the forms 'f-ness', 'K- kind', and 'that-p' which could not be reformulated in terms of expressions of the forms 'x is f', 'x is a K', and 'p' were categorizing statements such as 'f-ness is a quality', 'K-kind is a class', 'that p is a proposition', then we might well hope to relieve Platonistic anxieties by the use of syntactical therapy. I then examined a context which has been thought to correlate words with extralinguistic abstract entities, namely the context ‘‘—’ (in L) means …’, and found that it does not do so. Encouraged by this, I proceeded to examine the distinction between the material and the formal modes of speech to see if the idea that such categorizing statements as 'Triangularity is a quality' have the force of syntactical statements such as '"triangular" is an adjective' can run the gauntlet of familiar objections, with what I believe to be hopeful results.” [This takes us through Sections XI to XIII (XIV and XV first summarize and then point forward).] [161]

2. The roadmap for the first part of the essay (up through Section X) is set out like this:

“Now it is important to realize that Geach gives two accounts of the term 'property'; one of which, though cautious, is based on a simple grammatical mistake, while the other is derived from Frege's account, and is more difficult to expose.

a) The cautious account is contained in the passage quoted above, in which he stipulates that 'property' is to be equivalent to 'something that an object is or is not'.

b) The Fregean account is the one in which properties are introduced as what predicates stand for.” [138]

In effect, the thought of (b) is to introduce properties and concepts as what predicates mean.

In each case WS’s objection is of the same general form:

“And can we not therefore legitimately introduce the common noun 'concept' as having the force of 'something which a predicate stands for'? The answer is, as before, No; not, however, because it is incorrect to say that there is something which 'triangular' stands for (or bedeutet), but because the expression 'something which a predicate stands for' like the

1 0567579d14da788cab8cfdc5fc0bacef.docx Brandom expression 'something which an object is or is not' does not play the sort of role which would make it proper to introduce a common noun as its stipulated equivalent.” [143]

Specifically, however, he thinks this is easy to see for (a), and hard to see for (b).

3. The argument against (2a) is found in sections IV, VI, and VII, where the question of when it is appropriate to see a common noun as having been introduced. There Sellars distinguishes between “existential” quantification and existential commitment. The overall point is that only when an individuating sortal is explicitly indicated is there an existential commitment. There are a number of different considerations in play here:

a) Predicate/sortal/mass-noun: Individuating sortals (‘dog’) vs. predicates (‘red’). The former have both criteria (and consequences) of application and criteria of identity and individuation (which settle the answers to the questions “is this the same K as that, or a different one?”); the latter have only criteria (and consequences) of application. Q: What about mass nouns such as ‘gold’ or ‘water’? A: They belong in a box with sortals, rather than predicates, but are not counted but measured. They require units (we could—and linguists sometimes do—call them ‘quantifiers’, if that term were not already in use for something else), such as ‘pound of’ or ‘cup of’. Cf. Japanese classifiers, which are required for sortals as well as mass nouns. All three have circumstances and consequences of application. Sortals also have criteria of identity and individuation. Mass nouns need units to be specified, and then have criteria of identity and individuation, which is what we mean when we say that they are measured rather than counted. (The fact that the units may permit real numbers— or even more general measures—and not just natural numbers, is a further complication.)

b) Quine’s contrary claim: quantifiers are the locus of existential commitment (“To be is to be the value of a variable,”), with its commitment to working in a more expressively powerful metalanguage in which domains of quantification can be explicitly specified (even in the uncountably infinitary case) vs. the possibility of free logics, which separate the roles of particular quantifiers and expressions of existential commitment. Cf. “quantifying out”.

c) The issue of sortally restricted vs. unrestricted (Fregean) quantification. (Quine is on the side of unrestricted quantification—which again is the side of .) Geach, on behalf of natural languages, thinks only sortally restricted quantification makes sense.

d) One of the reasons Geach thinks that is that he is a fan of sortally relative identity: Should not say ‘x=y’, but ‘x is the same K as y’. I think this is a mistake (Gupta’s

2 0567579d14da788cab8cfdc5fc0bacef.docx Brandom

Logic of Common Nouns is locus classicus for an analysis of this situation). I think it is a mistake because I am an absolutist about identity: identity is identity. By that I mean that if x is identical to y, then they must share all their properties. This has fairly radical consequences: the statue Goliath ≠ the clay Lumpl, no passenger is a person, and I am not identical to the mereological sum of any collection of particles.

e) The reason to insist on sortally restricted quantification, in spite of rejecting relative identity, is that ‘thing’, ‘object’, and ‘unit’ are pseudosortals: they do not come with criteria of identity and individuation. One cannot count the objects in the room (do things like shadows and stains count?). One response: start with basic things (e.g. subatomic particles) and insist that everything else is a mereological sum of them. But this strategy is not available to absolutists about identity. My theory is that they are pseudosortals because they are prosortals: they are place-holders for other sortals, requiring one to be supplied in order for counting to go on.

f) What is ‘existential commitment’? My answer in MIE (Chapter 7) is in terms of substitution and canonical designators (examples: physical, numerical, and fictional existential commitments). But this answer does not allow existential questions that have the kind of generality characteristic of ontological questions, such as nominalists address. For if we ask: do abstract entities (or universals) exist? It is not clear how one might go about specifying a set of canonical designators. So I regard all such questions as only apparently determinate. A form of question that makes perfectly good sense locally has been illegitimately generalized beyond the boundaries of the contexts within which it has a definite sense. It is grammatical nonsense of the sort the later Wittgenstein points to (“Where was the pain before it was in my tooth?”, by analogy to “Where was the gold before it was in my tooth?”)

g) Q: But is WS right to think not just that sortal relativity is necessary for the issue of existential commitment to arise, but also that it is sufficient for existential commitment to be undertaken? I do not see that he is. I can ask how many unicorns are represented on the tapestry without being existentially committed to the (physical) existence of unicorns.

4. The argument against (2a) is epitomized in this passage from section VI: The key point to notice is that unlike existence statements proper, the statement (26) There is something which Jack and Jill both are begins not with 'There is a …’, not with 'There is a something …', but simply with 'There is something …'. If it began with 'There is a something …', thus using 'something' as a common noun, one might well look for a common noun, such as 'property', to pinpoint just what sort of ‘something' 'there is' which Jack and Jill both are. We could then have

3 0567579d14da788cab8cfdc5fc0bacef.docx Brandom

3 (26 ) There is a property which Jack and Jill both are. But all this, as by now should be obvious, is logical nonsense. 'Some- thing' is not a common noun, and it is incorrect, therefore, to introduce 'property' as equivalent to 'something which an object is or is not'. The term 'property' has, as a common noun, the form '— is a property' whereas, unless 'something' is to be construed as a common noun, the supposed equivalent has the form '— is something which an object is or is not', thus (28) Tall is something which an object is or is not and not ‘— is a something which an object is or is not'. Only if the expression 'something which an object is or is not' were a common noun expression (which it is not) would it be correct to introduce the common noun 'property' as its stipulated equivalent. In short, this way of introducing the term 'property' is simply a mistake. [139]

This argument is then confirmed by the discussion of ‘somequale’ (adjective) and ‘somequid’ (common noun) in Section VII: “Suppose we had begun with an example which involved the common noun 'man', instead of the adjective 'tall', say (29) Tom is a man. The corresponding generalization, as we have represented it, would be, (30) Tom is a something where the fact that the 'something' comes after the indefinite article makes it clear that 'something' is, so to speak, quantifying over a common noun variable. [140] Now the question-word 'what?' plays a number of roles in English which might well be split up among a number of interrogatives. In particular, we might introduce the interrogative 'quale?' to indicate that the answer is to be in terms of an adjective, and the interrogative 'quid?' to indicate that the answer is to be in terms of a common noun. Then we would have the question-echoing counterparts 1 (31 ) Tall is quale Tom is: Tom is who is tall, 2 (29 ) A man is quid Tom is: Tom is who is a man. To the first of each of these pairs there would correspond a general statement which would bear the mark of its origin, thus, (34) There is something which is quale Tom is (i.e. tall). (35) There is something which is quid Tom is (i.e. a man) or, more concisely, 1 (34 ) There is somequale which Tom is (i.e. tall). 1 (35 ) There is somequid which Tom is (i.e. a man).

5. What mainly does the work for WS here, I think, is the claim that it is at least a necessary condition of carrying existential commitment that a particular quantification be sortally restricted. In the Geach case, somequale is not. So, he concludes, no thing, nothing with identity and individuation criteria, is being introduced. We just have a variant use of a predicate, not a case where a new kind of thing, a property, is being referred to, even obliquely. Challenge: Unpack this argument. There is clearly something to it, but it needs to

4 0567579d14da788cab8cfdc5fc0bacef.docx Brandom

be broken down into steps and confronted with all the alternatives that (3) above puts into play.

6. The argument against (2b) commences in VIII. The move he will complain about goes from 2 (45 ) Triangular is the meaning of 'triangular', which involves the common noun 'meaning'. It is then a simple step to stipulate that 'concept', 'property', 'nature', and 'form' are to be general terms for the meanings of adjectives and common nouns. [143-4] At this point, his analysis of ‘means’ as illustrating-classifying (he does not yet have dot-quotes) is his response.

7. “Can we, then, say what is said by 'That S is triangular implies that S has three sides' (47) and 'That anything is triangular implies that it has three sides' (461) without committing ourselves to singular terms formed from statements? Surely it will be said, all we need to do is to make use of the familiar symbol ‘É’ which was specifically designed to be the non- predicative core of the predicative term 'implies’. We would then have 1 (47 ) S is triangular  S has three sides and 2 (46 ) (x) x is triangular  x has three sides and if this move is successful, we should have freed ourselves (temporarily, at least) not only from expressions of the form 'that-p', but also, unless we find other reasons for reintroducing them, from un- saturated singular terms of the forms 'that x is f' and 'that x is a K'; and hence from 'f-ness' and 'K-kind'. We would indeed have extricated ourselves from Plato's beard.” [146-7] What we have instead is relations among sentences or statements.

8. Sections XI to XIII present the argument that categorizing contexts, which resist the paraphrase strategy just indicated, also can be dealt with, by going metalinguistic, á la Carnap:

“[T]he truth or falsity of such categorizing statements as (11) Triangularity is a quality, (54) Dog-kind is a class, (55) That Chicago is large is a proposition, is ascertained not by 'examining' triangularity, betweenness, dog-kind, or that Chicago is large, but by reflecting on the role in discourse of the corresponding expressions. This is the insight contained in Carnap's contention (in The Logical Syntax of Language) that the above statements are in the 'material mode of speech' and are the 'quasi-syntactical' counterparts (roughly—for I am following the general spirit, rather than the letter of Carnap's account) of 2 (11 ) 'Triangular' (N .B. : not 'triangularity') is an adjective (in English),

5 0567579d14da788cab8cfdc5fc0bacef.docx Brandom

1 (54 ) 'Dog' (N.B.: not 'dog-kind') is a common noun (in English), 1 (55 ) 'Chicago is large' (N.B.: not 'that Chicago is large') is a sentence (in English).” [148-9]

9. The next challenge is to face the “two obvious objections” to the metalinguistic move:

a) We’ve just traded nonlinguistic universals for linguistic ones: word-repeatables.

b) The original statements don’t talk about or mention English words.

10. On (8a): “But surely, it will be said, the word 'triangular' is just as abstract an entity as triangularity. Where is the 'nominalistic' gain? Is not the term ‘"triangular"’, as much a singular term as 'triangularity', and 'adjective' as much a common noun as 'quality'?” [149]

And the answer he gives is this:

The answer is simple and straightforward. ‘"Triangular"’, is not a singular term, but a common noun, and the gain arises in that we can hope to equate (113) with something like:

(x) x is ‘triangular’  x is an adjective.

11. “Now the way out of this labyrinth consists in recognizing that it is incorrect to say that 'dreieckig' means a word, and equally incorrect to say that it means a non-word, for the simple reason that 'means' is not a transitive verb. Not that it is an intransitive verb, for it is neither, and the attempt to fit it under one or the other of these headings, on the supposition that they are not only mutually exclusive but jointly exhaustive, is the cause of the puzzle. “ [152]

12. [W]hile it would be incorrect simply to say that there are no such things as meanings, or Frege's concepts, or Geach's properties, to trace the common noun 'meaning' to its source in the translation rubric ‘‘—’ (in L) means ...’ (66) is to make what amounts to this point in a less misleading and dogmatic way. The upshot of the foregoing discussion of meaning with respect to the primary theme of this article can be summed up by saying that the translation context (66) does not properly take a singular term on the right-hand side unless the expression of L which is placed in the single quotes of the left hand side is itself a singular term. In other words, this context does not of itself originate a commitment to abstract entities. [154]

13. On (8b):

6 0567579d14da788cab8cfdc5fc0bacef.docx Brandom

How can 'Triangularity is a quality' (11) have something like the force of '"Triangular" (in 2 English) is an adjective' (11 ) in view of the fact that (11) makes no reference to the English language? [156] Again, how can the truth of (11) be ascertained by reflecting on the use of the word 'triangular' if, were a German to say (78) Dreieckigkeit ist eine qualitaet, aber es gibt keine Englische Sprache, his colleagues would recognize that his statement was only contingently false? For if his statement is only contingently false, it might have been true, and if it had been true, he could have made a true statement, namely (11g) above even though there was no English language in general, nor, in particular, such an English word as 'triangular'.

14. Summary: “To know how to use singular terms ending in ‘-ity’ is to know that they are formed from adjectives, while to know how to use the common noun ‘quality’ is (roughly) to know that its well formed singular sentences are of the form ‘__is a quality’ when the blank is appropriately filled by an abstract noun.” This statement (among others) makes it very clear that what Sellars is doing here is saying what one must do in order to mean various things (‘triangularity’, ‘quality’). [158] In my BSD terms, he is offering a pragmatic metavocabulary for abstract object talk.

15. Further summary (158): “Thus, while my ability to use ‘triangular’ understandingly involves an ability to use sentences of the form ‘__is triangular’ in reporting and describing matters of physical, extralinguistic fact, my ability to use ‘triangularity’ understandingly involves no new dimension of the reporting and describing of extralinguistic fact—no scrutiny of abstract entities—but constitutes, rather, my grasp of the adjectival role of ‘triangular’.” I could put this in BSD terms by saying that the ability to use ‘triangularity’ can be algorithmically elaborated from the ability to use ‘triangular’. It is not, however, explicative of that use. It is only derivative from it. [158]

16. “Ontological categories are the material mode of speech for syntactical categories.” [159]

17. [W]e are in a position to answer our original question. For we have now located a difference between the 'material' and the 'formal' modes of speech which enables us to see how they can 'have the same force' without one being a simple rewriting of the other. For while it would be incorrect to say that 'That — is triangular is a particular-gappy state of

7 0567579d14da788cab8cfdc5fc0bacef.docx Brandom

1 affairs' (82 ) is a mere rewriting of' ‘‘— is triangular’ (in English) is a singular-term-gappy 6 attributive sentence’ (11 ), it is at least a reasonable next step in the direction of the truth to interpret it as a rewriting which presupposes that 'writer' and 'reader' are able to use as well as mention sentences of the form ‘— is triangular'. [160]

18. WS is desperately in need of dot-quotes here, and will get them in place 3 years later in the 1962 “Naming and Saying”

19. Propositions are to be done away with in favor of assertions of truth, which are in turn understood as endorsements of inferences (corresponding to T-sentences), that is, to doings. (Cf. the end of NS) This is an appeal to pragmatics. But properties are understood in terms of predicates (not predicating), a matter of content rather than force. This is a semantic issue. Q: How is he going to reconcile these two, given that he is committed to understanding subsentential expressions in terms of the contribution they make to what one is doing in using sentences? (Perhaps this is the resolution, in favor, ultimately, of pragmatic metavocabularies.) The issue is the relation between pragmatic and semantic metavocabularies. Cf: The 'propositional' reading of Carnap's definition becomes, under the pressure of the demand for consistency, (77) T is a true that-clause (in L) = there is a proposition, that-p, such that T designates Df that-p (in L) and that-p is the case. and while I do not wish to impugn the consistency of the notion, thus introduced, of the truth of a that-clause, I do wish to insist that this notion is philosophically unsound in so far as it rests on the mistaken idea that the truth must be defined in terms of propositions, and leads to the mistaken idea that the truth of statements is derivative from that of that-clauses. [155]

20. I am, in the special sense of BSD, together with WS in endorsing what amounts to a metalinguistic expressivism about all of logical, modal, normative, and intentional vocabularies. WS extends this to ontological vocabulary, including not only talk of properties, concepts, and universals, but also propositions, hence facts. Here we part company, because I want to understand the relation between object, fact, and law—the concepts we use to articulate the structure of the objective world, on the one hand, and referring (singular term), asserting (declarative sentence), and inferring (conditionals) as one of sense dependence (not reference dependence). If one does that, one can allow that properties and concepts, as well as facts, really articulate the objective world, even though the concepts of such things cannot in principle be understood except in a context that includes linguistic activity. Thus, I am not a nominalist in Sellars’s sense.

8 0567579d14da788cab8cfdc5fc0bacef.docx Brandom

21. On metalinguistic expressivism as a strategy—independently of whether Sellars takes it to far: Huw Price on subject vs. object naturalism, and the suggestion of global expressivism, even about empirical describing.

After:

22. Locutions that might lead us to think we were committed to the existence of properties:

a) Predications (if we have to think of them as put more perspicuously by ‘a exemplifies F-ness’, a la Bergmann)—discussed in NS.

b) ‘Existential [really: particular] quantifications over predicates’: “There is something Tom and Tim both are, namely tall,”—(2a) in GE.

c) “What predicates stand for (not name),” Geach in the Fregean mode—(2b) in GE.

d) Categorizing contexts: “Triangularity is a property.”—in GE.

e) Might add ‘means’ locutions more generally.

WS means to treat all the locutions that one might think carry such commitment, and to show first that they do not commit us to abstract object kinds, and second, that we can in each case give a metalinguistic expressivist (his ‘quasi-syntactical’) account of the actual use of those locutions, which does not involve reference to such things in the world, but understands them rather as expressing necessary conditions of the possibility of saying anything about objects.

So, if one asks: do the objects that are really, objectively, in the world, have properties? The answer is Yes. But what one is saying (doing?) when one says that is properly understood to involve a metalinguistic admixture that is not involved in saying that there are individual objects (as opposed to universal properties).

Q: Why isn’t it strictly parallel: “a is an object” = as are singular terms?

A: We can and should say that. But that is a point about the concept object.

9 0567579d14da788cab8cfdc5fc0bacef.docx