REALISM VERSUS ANTI-REALISM

Dr. Mohammad Ashraf Adeel

INTRODUCTION

 The difficulties encountered by an architect, while constructing a house from make-believe bricks, are difficulties of principle, unless he is engaged in constructing a make-believe castle of a child’s reveries. It is this kind of difficulty, according to Dummett, that a realist theory of meaning is confronted with, and not simply the difficulties of detail. This claim, plainly, is not music to the realist ears. Hence the ensuing realist/anti-realist controversy that, by the way, has remained at the heart of the philosophical project down the ages. In more recent literature, Dummett has argued that the primary character of this controversy is semantic. How do we understand the sentences of a given class, e.g., sentences about physical objects, mathematical sentences, sentences about mental states or events, sentences about theoretical entities of science, and past tense and future tense sentences? The realist and the anti-realist give different answers to this question.

I seek here first, to state the anti-realist argument in its contemporary form, specifically as it has appeared in the writings of Michael Dummett, and then to examine various realist responses to it. My strategy is to establish various points of contact, where possible, between the contemporary and Plato’s responses to the anti-realist, and then to argue that realist arguments, old or new, are attended by difficulties. The anti-realist position, therefore, is not effectively destroyed by them.

Both Dummett and Davidson seem to agree that a theory of meaning is concerned with explaining a speaker’s practical ability to understand and use his language. It is a theoretical representation of the implicit knowledge that a speaker has of his language. But Dummett argues that, “since what is being ascribed to a speaker is implicit knowledge, a theory of meaning must specify not merely what it is that a speaker must know, but in what his having that knowledge consists, i.e., what counts as a manifestation of that knowledge. Without this not only we are left in then dark about the content of ascribing such knowledge to a speaker, but the theory of meaning is left unconnected with the practical ability of which it was supposed to be a theoretical representation.”[1]

Now a realist theory of meaning takes truth to be the central semantical notion. Sentences in the given class, on such a truth conditional theory of meaning, have a determinate truth-value, regardless of whether we actually do or do not know their truth-condions. Realism, therefore, is wedded to the principle of bivalence or an analogous multivalence, since some philosophers (Aristotle for example) are inclined to think that future tense sentences are neither true nor false. Bivalence, however, is not a sufficient condition for a realist theory of meaning. One needs, in addition to bivalence, a “certain conception of the manner” in which our sentences come to acquire their truth-value. And, as Dummett puts it, this “embodies an appeal to the notion of reference as an indispensable notion of the semantic theory”.[2] However, within these broad features of the realist program, there are at least three variations of realism that can take place. These variations Dummett refers to as naive realism, sophisticated realism, and semi-naive realism.

A naive-realist believes in what Dummett calls an irreducibility thesis regarding the statements in the given class.[3] What it amounts to is that no non-trivial general answer can be given to the question: ‘What makes a sentence of the given class true, if it is true?’ A trivial answer to such a question would be something like Tarski’s disguotation schema: ‘P’ is true if P, where P is the sentence of the given class. In addition to the irreducibility thesis, naive realism, as traditionally understood, involves an epistemological component. Dummett offers, as an illustration of this component of naive realism, the standard view of sentences in the past tense. This view is committed not only to the principle of bivalence but also to some kind of direct acquaintance with the past events through the memory. The reason is that, on a naive realistic interpretation, there is no non-trivial answer to the question: ‘What makes a sentence about the past true, if it is true?’ And the trivial answer, to the effect that, a past event makes a sentence about the past true, when it is true, automatically commit a naive realist to the claim that we have a direct acquaintance with the past events.

A sophisticated realist, on the other had, holds a reductive thesis for the sentences in the given class. This Dummett distinguishes from reductionism, the claim that sentences in the given class can ° be translated into the sentence in the reductive class. Such a claim involves well-known difficulties and has been avoided by most philosophers. The reductive thesis, however, makes the more modest claim that “no statements of the given class can be true unless some suitable statement or statements of the reductive class are true, and, conversely, that the truth of those statements of the reductive class guarantees the truth of the corresponding statement of the given class”.[4] The basic idea of a reductive thesis then, is that when we accept as true the sentences of the given class we thereby implicitly accept the related sentences in the reductive class.

Some reductive thesis, however, depend for their plausibility on the admission of subjunctive conditionals into the reductive class. Phenomenalistis forienstance, reduce sentences like ‘There is a moon in the sky’ into a subjunctive conditional of the form: ‘If someone were to look at the sky (under appropriate conditions), he will see a moon’. This, of course, is only the first step, and through further transformations, in which the subjunctive conditional form persists, phenomenalists try to get the equivalent of the original sentence in terms purely of sense-data.

The case of constructivism in mathematics is different however. The plausibility of the reductive thesis here does not depend on the admission of subjunctive conditionals into the reductive class. What makes a mathematical sentence true, when it is true, is the existence of a proof for it. There is no need here to assume that a mathematical sentence is true if, under appropriate conditions, there would exist a proof for it. Such subjunctive conditionals are simply irrelevant.

As for the semi-naive realist the only thing that differentiates him from the naive realist is his denial of the epistemological component of naive realism. This, obviously, is a difficult position to be in. Given his adherence to an irreducibility thesis all that a semi-naive realist can do is to resort to analogy. Sentences of the given class, he claims, have features analogous to some more primitive sentences which, in their turn, are explainable in the naive realist fashion. This, of course, does not seem to be a very consistent position --- but, that is what it is.[5]

I have belabored these Dummettian characterizations, before setting out his anti-realist argument, simply to avoid confusion. For instance, it should be clear from the foregoing that admission of a reductive thesis for sentences of the given class, in and of itself, does not constitute a rejection of realism. Indeed, a reductive thesis is quite compatible with realism. In addition to a reductive thesis one needs to reject a truth-conditional account of the meanings of sentences in the given class in order to embrace an anti-realist position.

II

The fact that Dummett’s anti-realist argument has been subjected to such invective as ‘circular’[6] ‘lunatic’[7] and ‘non-argument’ by recent authors on the subject makes it urgent that I state the argument in its entire here for us to be able to cast a hard look it. Following is what I consider to be a faithful presentation of Dummett’s argument:

1. A speaker’s implicit knowledge of the meanings of sentences in his language may be manifested in either (a) a verbal explanation of the meaning of a given sentence in terms of an equivalent but non-trivial reformulation of it or (b) in carrying out an actual verifactory procedure that issues in a decision as to the truth or falsity of the sentence.

2. It is circular to assume that a speaker’s understanding consists in an ability to express each sentence of his language in terms of another, equivalent but distinct, sentence of the same language.[8]

3. The most primitive parts of language, its lower levels, cannot be explained by the means suggested in 1 (a).

4. Natural language has certain features that lead to “formation of sentences not in principle decidable: the use of qualification over an infinite or unsurveyable domain (e.g., over all future times); the use of subjunctive conditional, or of expressions explainable only by means of it; the possibility of referring to regions of space-time in principle inaccessible to us”.[9]

5. A language comprised of decidable sentences alone will continue to have this property when enriched by expressions introduced through verbal explanations.

6. Natural language operators mentioned in 4 would belong either to the verbally explainable part of our language or to its primitive part not so explainable. .

7. If the natural language operators belonged to verbally explainable part of our language and if they were applied to sentences similarly explainable, they won’t yield undecidable sentences.

8. Since they do produce undecidable sentences they must either belong to the primitive part of our language or be applied therein.

9. In either case our language contains undividable sentences in it’ primitive part.

10. A speaker’s knowledge of. the primitive parts of his language does not/cannot consist in an actual verification as to the truth or falsity of some sentences in that part. In slightly different word’s as the conditions of truth for undecidable sentences are recognition-independent no one can say what procedures would; actually confirm or infirm these sentences.

11. We are left in the dark about the content of the knowledge/understanding of these sentences ascribed to a speaker. Which, obviously, means that a truth-conditional theory fails to provide us with an acceptable account of a speaker’s understanding of his language.[10]

This is Dummett’s anti-realist argument. It must be borne in mind, as Dummett himself insists, that the argument advocates a negative position.[11] Also we must realize that although it is imperative that an anti-realist propose a viable alternative to the realist theory of meaning, the success or failure of the anti-realist argument, does not depend simply on the success or failure of such proposed alternatives. The reason being that even when the anti-realist proposals for a theory of meaning would have failed completely, the task still would remain for the realist to make good his claim that a speaker’s understanding of undecidable sentences of his language consisted in a grasp of recognition-independent truth-conditions.

Demmett’s proposed alternative to the truth-conditional theory of meaning is that of acceptability-conditions. This proposal takes its inspiration from the work of later Wittgenstein. Wittgenstein said that the meaning of a word/sentence is its use. The central features of such use for assertoric sentences, according to Dummett, are two: (i) the circumstances in which an assertion is made, and (ii) the evidence that we take as verifying or falsifying the assertion.[12] These two features of its use constitute, for Dummett, the assertibility conditions of a given sentence. Dummett no longer insists that the sentences be verified falsified conclusively.[13] Verification needs be only defensible.

Clearly, then, sentences no longer possess a determinate truth-value under the anti-realist theory of meaning. The classical bivalence (or any analogous multivalence) goes by the board. Of course this requires adjustments in the tenets of classical logic.

The advantage of an assertibility-conditions theory over a realist theory of meaning, in Dummett’s own words, is “that the condition for a statement’s being verified, unlike the condition for its truth under the assumption of bivalence, is one which we must be credited with the capacity for effectively recognizing when it obtains, hence there is no difficulty in stating what an implicit knowledge of such a condition consists in -- once again, it is directly displayed by our linguistic practice”.[14]

It is this anti-realist position that has come under virulent attack in recent literature. In what follows I shall examine some of the more “powerful” attacks on anti-realism.

III

Richard Boyd, I believe, has made a serious effort recently to defend realism.[15] I am inclined, therefore, to examine his argument for scientific realism in some detail. The target of his argument is radical underdeterminism --the view that experimentally indistinguishable theories can produce incompatible interpretations of the causal relations between theoretical (unobservable) entities that the theories in question quantify over; and that the choice between two such theories is a matter simply of convention and not of experimental evidence. In other words, for a radical underdeterminist, there is no fact of the matter that determines which theory we should choose. This, obviously, is an anti-realist position.[16] Let us, therefore, turn to the mechanics of Boyd’s argument against it.

As an example of two theories with the above-mentioned characteristics Boyd cites the paradigm-case introduced by Reichenbach:‑

Let F be current physical theory, and in particular, let F contain a “catalogue” of the sorts of forces which operate in ‘the physical systems let G be the geometrical principles which are true if “straight line” is taken as “trajectory of an (idealized) point mass upon which the resultant of the forces acknowledged by F is zero”. Let G’ be an alternative set of (suitably comparable) geometrical axioms, and let F’ be the physical theory which results from the addition to F of laws governing an additional, universal force f with the following property: f’ is so defined (rigged, as it were) that G’ is the correct physical geometry if the physical interpretation of “straight line” is amended so that the relevant trajectories are those of point masses upon which the forces acknowledged by F together with the force f have resultant zero.[17]

Now Boyd informs us that it is a widely acknowledged fact today that most theories arising in physical sciences have no non-trivial observational consequences unless combined with other laws (consistent with them) which he calls “auxiliary hypotheses”. Hence what the radical under determinist claims, in effect, is that “F and G” and “F’ and G” will have exactly the same observational consequences when taken together with other currently accepted theories respectively compatible with them. This means that “F and G” and “F’ and G” are experimentally indistinguishable. Also, because F’ adds a new force f’ to our “catalogue of physical causes” it appears that “F and G” and “F’ and G” provide us with mutually incompatible account of the causal relations between theoretical entities. This Boyd believes is a striking example for a radical underdeterminist to present in favour of his position. For the radical underdeterminist wants to argue that the choice between such theories as “F and G” and “F’ and G” is a matter not of experimental evidence but simply of convention. Because, after all, both the theories are experimentally indistinguishable.

Boyd thinks that such a view of the situation is utterly mistaken. The radical under determinist relies on the following principle for his position:

(1) If two theories have exactly the same deductive observational consequences, then any experimental evidence for or against one of them is evidence of the same force for or against the other.[18]

But this principle, Boyd claims,” is absurd if the antecedent is taken as referring to the observational consequences of the theories themselves (i.e., with auxiliary hypotheses not employed in the deductions), since it would claim that the experimental evidence for classical mechanics is exactly as good as that for special relativity, if only both theories are stated abstractly enough”.[19]

But what sort of auxiliary hypotheses, consistent with the theories in question, can be combined with these theories? Presumably there are three legitimate possibilities: (a) any set of auxiliary hypotheses, (b) “currently accepted” laws and generalizations, and (c) laws and generalization that “will eventually be accepted (and not thereafter rejected)”. Each of these possibilities, of course, yields a separate version of the principle (1). But Boyd goes on to argue that all the three resultant versions are false. Following are the three versions:

(1) If ‘T’ and T’ are each consistent and have exactly the same observational consequences no matter which set of possible auxiliary hypotheses is employed with both in the course of the deductions, provided only that the auxiliary hypotheses are consistent with T and T’, then T and T’ are equally supported or disconfirmed by any possible experimental evidence.

(1”) If T and T’ are each consistent and have the same observational consequences when one is allowed to employ with each of them as auxiliary hypotheses any set of currently accepted laws or generalizations which forms together with the theory a consistent set, then T and T’ are equally supported or disconfirmed by any possible experimental evidence.

(1”‘) If T and T’ are each consistent, and if, when one is allowed to employ with each of them as auxiliary hypotheses whatever laws or generalizations will eventually be accepted (and not thereafter rejected) in the course of scientific research, T and T’ have the same observational consequences, then T and T’ are equally supported by any possible experimental evidence.[20]

Version (1’), according to Boyd, is true “provided at least one of the theories has some non-observational terms and provided that the set of observational consequences which these theories each yield with no auxiliary hypotheses leaves some observational question unsettled”.[21] As long as the two theories in question satisfy these restrictions, version (1’) is true. But then, Boyd contends, the two theories turn out to be exactly the same. And hence version (1’) cannot help the radical underdeterminist. Version (1”) is false, according to Boyd, because it is always possible for us to discover new laws or generalizations in the light of experimental evidence, such that, these laws and generalizations when employed as auxiliary hypotheses with the two theories question would enable us to “derive contradictory, observational testable”[22] consequences from them.

As for the version (1”) it doesn’t offer much hope for the radical underdeterminism either. For Boyd contends that even if we grant that there is a non-empty set of laws and generalizations that will be eventually accepted and never rejected thereafter we have no logical or historical technique at our disposal to show the antecedent of (1’) as true except where the two theories in question are identical.

But, Boyd continues, that a radical underdeterminist can insist that version (1”) cannot be rejected simply on the grounds that some laws and generalizations to be discovered in future would enable us to derive contradictory consequences from the two theories in question. For example we cannot claim that there is experimental evidence for us to choose, between the above-medntioned F and G and F and G” on the assumption that some future theory T would prove F and .G experimentally wrong. Therefore the realist must show that F and G and F and G are experimentally distinguishable, here and now, in order for his claim to . work. In other words the radical underdeterminist invokes the following version of (1”):

(1”a) If T and T’ are each consistent- and have the same observational consequences when one is allowed to employ with each of them, as auxiliary hypotheses, any set of currently accepted laws or generalizations which forms, together with the theory a consistent set, then T and T’ are equally supported by any possible experimental evidence, .provided that this evidence does not indicate the acceptance of some new law, or the disconfirmation of an old one.[23]

Obviously then the success of Boyd’s realist argument depends now on whether or not he can prove (1”a) to be false. His strategy is to offer us the following methodological principle P, and then to argue that we could never apply the principle P fruitfully unless the collateral information, in the light of which we test a new hypothesis, was already (approximately) true.

(P) a proposed theory T must be experimentally tested under situations representative of those in which, in the light of collateral information, it is most likely that T will fail, if it’s going to fail at all.[24]

It is obvious that no hypotheses are tested in a vacuum. But unless the collateral information in the light of which they are tested is already somehow true, we should never be able to collect the right kind of evidence that would confirm a given hypothesis. (Boyd cites some convincing examples).[25] The reason is that if our collateral information is false we will never know for sure under what circumstances the given hypothesis can be tested effectively.

From this Boyd concludes that (1”a) is also false. For assuming that our currently accepted laws and generalizations are (approximately) true (for otherwise we cannot successfully test any hypotheses), the same experimental evidence cannot equally support both T and T’. Reverting to Reichenbach’s case “F’ and G” must be taken, here and now, as experimentally false. Hence the choice between “F and G” and “F’ and G” is not a matter of convention, it is determined by experimental evidence, by facts of the matter.

The picture of science that emerges from Boyd’s argument is strikingly different from the Kuhn-Feyerabend mode. The scientific enterprise is not a matter of paradigm-shifts but a matter of getting better and better theories. So, what Boyd’s argument for realism says, in effect, is as follows: ‘surely we are making progress. Such progress would be impossible unless we were already in possession of some (approximate) truth. Therefore we are in possession of such a truth’. In other words sentences of a currently accepted scientific theory have a determinate truth-value.

Now at this general level this argument for realism is not a completely new one. It appears in Plato’s Cratylus (386 a-c) and The aetetus (166b-183b)) as well as in Aristotle’s Metaphysics (Book 4, chap. 4-5). Let us concentrate on the argument in the Theaetetus.

In the Theaetetus this argument really has a grand design. Socrates first establishes that Protagras and Heraclitus are committed to the same underlying position in their respective doctrines, and then goes on. to force Theodorus to grant, on pain of self-refutation, that there is such a thing as good and bad practices with individuals and among communities (177-c-179b). Indeed, Socrates claims, Protagoras’ own profession as a teacher will make no sense without such a distinction. Having made these two gains Socrates proceeds to give the following argument against Protagoras and Heraclitus.

1. All things are in flux.

2. Nothing is constant enough for us to give it this or that name. Not just white things are changing, but whiteness itself is changing (to blackness or non-whiteness).

3. Any particular perception - seeing of example - is no more seeing than not-seeing.

4. Perception is knowledge.

5. Knowledge is no more knowledge than not-knowledge.

6. We can say of anything that it is so and it is not-so.

7. Language is impossible because what is so ceases to be so in change, and the same goes for what is not-so.

According to this argument, the flux theorist cannot claim any fixity at all and hence he cannot talk. However the flux theorist need not remain silent. It is possible for him to claim that the possibility of language is not contingent upon there being any fixities, determinate facts of the matter, to the world or to our understanding or to anything Language as well as all human practices, just flow out of our nature. Language is a “form of life”. We are nothing different from bees and ants. If it does not make any sense to ask how bees and ants can justify their practices? it makes even less sense to ask how language ultimately, is possible. It is just there.

An anti-realist need not insist here that we can draw an ultimate distinction between language and non-language, meaningfulness and meaninglessness. Speech is just an expression of our brute fluctuating urges. There is no fixed meaning to any word or sentence. Meaningfulness depends on when, under our brute urge, we assert a sentence and what appears to us as appropriate evidence for asserting it at that time. The search for a fixed meaning for a word or a sentence is worse than a wild goose chase, it is an illusion.

Just as with meanings so with other matters: there are no fixed rules governing any human practice. A brute urge is all that we follow. And hence, it is senseless to talk about good and bad practices. We never improve.

This means that an anti-realist need not grant what our gullible Theodorus grants Socrates in the Theaetetus. An anti-realist can maintain that there is no such thing as good views and bad views. Even the distinction between views and non-views is not a fixed one. We follow a brute urge and test witch-theories at one time and theories of successful bomb-making at another. The point is that if witch-theories strike as ridiculous today, our theories may strike future generations as ridiculous. Getting struck, one way or the other, by a theory is itself a matter of theorizing, which, after all, has no fixed charter to it. An anti-realist, therefore, can make a case to pull these notions of improvement and fixities from underneath the Plato-Boyd argument.

IV

Other recent authors on the subject have claimed that anti-realism is incoherent and hence, self-defeating. William P. Alston, for instance, in his Presidential Address (delivered to the Western Division of the American Philosophical Association on May. 20, 1979) makes the following claim:

The real crusher for the anti-realist is the question “How are we to interpret the statements to which you apply your concept of truth?” What is crushing about this question? Well, the point is that on a natural intuitive way of understanding statement content (of specifying what is being asserted in a given statement), that content carries with it the applicability of the realist concept of truth.[26]

Essentially the same point is made, I believe, by C.B. Martin in a more recent paper. He says:

Sentences used in claiming for oneself or for others an understanding of a sentence or range of sentences are to be explicated, on Dummett’s account, in terms of a vaguely indicated range of capacities to recognize - as-confirming. These, in turn, generate a vaguely indicated range of counter factual sentences.

The problem for Dummett is how to find for this whole bag of sentences some other set of sentences by which the former can be explicated (Dummett often uses the term “reduced”) that will be “effectively decidable” and determinate in truth-value. Otherwise, there will be no determinate fact of the matter” as to our understanding.[27]

From these kind of considerations both Alston and Martin conclude that anti-realism cannot be maintained consistently. In Alston’s words: “it is incoherent to say” What I asserted was that snow is white (or what I did in my assertion was to refer to snow and say of it that it is white)., but the truth of my assertion does not ride on whether snow is white”.[28]

This argument also takes its cue from Plato. In the Theaetetus Socrates gives the following summary-statement of it.

Protagoras, for his part, admitting as he does that everybody’s opinion is true, must acknowledge the truth of his opponents’ belief about his own belief, where they think he is wrong, (171a).

The argument is fallacious and one way to expose its fallacious character is as follows: Protagoras is being portrayed as a half ‘hearted anti-realist here. In other words if no-thing (no truth-conditions) make Protagoras’ opinion true and if ‘man is the measure of all things’ then he must grant that his opinion is false when it appears false to others. The question for the anti-realist, however, is that of assertibility and not that of truth or falsity. Protagoras asserts his opinion and others assert an opinion about his opinion. In other words these ‘others’ are at a metalinguistic level viz-a-viz Protagoras’ opinion. Just as there is no fact of the matter that makes Protagoras’ opinion true, so there is no fact of the matter that will make his opponents’ opinion true. Both opinions are just assertible. And if we throw away the principle of bivalence and keep our eyes open to the defeasible character of assertibility conditions, Protagoras cannot be refuted on the grounds that he allows his opponents’ opinion about his own opinion to be assertible.

The above mentioned Alston Martin argument, as Alston readily acknowledges,[29] is based on a similar .half-hearted characterization of anti-realism. Why in the world should Dummett, ory anti-realist for that matter, acknowledge that there is any “determinate fact of the matter” to our understanding of the sentences either in the given or in the reductive class. There is no such thing as a “determinate fact of the matter” either to the world or to our understanding.

V

I turn now to McGinn’s and Currie-Eggenberger arguments for a realist theory of meaning.

McGinn Argument: McGinn relies on Putnam’s brains-in-the vat analogy in order to make his point that content of a speaker’s understanding of any sentence in his language cannot be explained simply in terms of assertibility conditions.[30] We have these vat people on a twin earth who are victims of a systematic hoax. They speak and behave like normal people overtly but in reality they are under systematic hallucinations or under the complete control of some super-genius biologist who induces all these sensations into their system.

Now McGinn asks us to imagine four possible cases:

(A) Vat people have, in their language, a word similar to one in our language, for example the word ‘water’. The difference, however, is that ‘water on twin-earth refers to a stuff which, although similar in taste and looks to water on earth, has XYZ rather than H2O as its chemical constitution. (B) The term water on the twin-earth is empty. (C) All terms on the twin-earth are empty. (D) Twin earthians have a psychological term that is empty.

From all the tour cases McGinn contrives to get the same conclusion, namely that assertibility conditions cannot explain the content of a speaker’s understanding of his language. Let us look at his argument in Case A for now:

1.   Let speakers on earth and twin earth have the same recognitional capacities and manifest them in the same conditions of evidence.

2.   There is stuff on the twin-earth which tastes and looks exactly like water but is XYZ in its chemical constitution rather than H2O.

3.   Twin-earthins, in their hoax, call this stuff ‘water’. So their sentences containing the term ‘water’ have assertibility conditions similar to the corresponding sentences of normal speakers on earth who are equally ignorant of chemistry.

4.   But, obviously, the content of twin-earthian sentences is different from those of earthian sentences - they are sentences about different things.

5.   Similarity in assertibility conditions or use does not lead to similarity in content.

6.   We have to invoke truth-conditions in order to account for the content of the sentences.

McGinn repeats the same argument, with appropriate adjustments, for all of his cases. This is a bizarre argument. It places the speakers behind a veil of ignorance and then proceeds to assign different contents to their sentences on the basis of supposed difference in truth-conditions. One wonders how recognition-independent truth-conditions can contribute to the content of a speaker’s understanding of a sentence. To postulate recognition-independent truth-conditions for sentences first, and then proceed to claim that these truth-condtions cannot be recognized in use is hopelessly question-begging. And this goes for all the cases constructed by.McGinn.

Currie-Eggenberger Argument: Gregory Currie and Peter Eggenberger in a recent joint paper present the following argument against what they call “the central and most problematic contention within Dummett’s argument,” i.e., “the claim that an ascription to a person of knowledge of meaning is without content unless we can specify some item of behavior which maifests that knowledge.”[31]

 1. The distinction between a person’s knowledge of the meaning of a sentence and the observational behavior in which this knowledge is manifested is similar to the distinction between theoretical sentences and observational sentences of a scientific hypothesis:

2. Dummett’s view demands that each occurrence of a theocratical sentence must be associated with an observational sentence.

3. Ordinarily, the presence of theoretical sentences in a scientific hypothesis is justified by their “overall contribution to the coherence and explanatory power”[32] of the hypothesis.

4. Theoretical sentences of the form ‘knowledge of the truth-conditions of sentence X’ in a meaning theory need not be each associated with observational sentences of the theory, i.e., observable behavior of the speaker - verbal explanations or effective verficatory procedures. All that is required of these theoretical sentences, on the most widely accepted and successful model of scientific theories, is that they make an “overall contribution to the coherence and explanatory power” of our theory of meaning.

5. Dummett’s demands on a meaning theory are illegitimate. Currie and Eggenberger acknowledge that Dummett’s model is more austere but go on to claim that such austerity is uncalled-for according to our current Model of a theory.

This argument involves a dubious equation. The distinction between a peron’s knowledge of meaning and his observable behavior in which such a knowledge is manifested cannot be equated with the distinction between theoretical and observational sentences of a scientific hypothesis for the following reason: Although we need not, in a scientific hypothesis, associate each occurrence of a theoretical sentence with an observational sentence, we must always be able to determine that our observational and theoretical sentences do not contradict each other. (This is what Currie and Eggenberger call ‘coherence’ of a theory). Which means that we must be able to associate theoretical sentences with observational ones at least indirectly. Which means that theoretical sentences should be indirectly deducible from the observational ones. Now in case of in principle decidable sentences, it is perfectly possible for us to associate their truth-conditions with the recognizable behavior of the speaker in this indirect fashion and make sure that our theoretical sentences do not contradict our observational sentences in a meaning theory. The problem, however, is regarding those sentences in our language which are not decidable in principle. We have no way of determining whether or not their completely recognition-independent truth-conditions are compatible with the observational behavior of a speaker. It follows, therefore, that the distinction between a person’s knowledge of the meaning of a sentence and his observable behaviour is not exactly similar to the distinction between theoretical and observational sentences of a scientific hypothesis. The Currie-Eggenberger argument, therefore, seems defective.

VI. Conclusion

The foregoing analysis of some of the realist arguments shows that they are attended by various sorts of difficulties. We can conclude, therefore, that anti-realism is still in the field as a viable alternative to realism.

NOTES