You are on page 1of 23

Dummett's Intuitionism Is Not Strict Finitism Author(s): Samuel William Mitchell Reviewed work(s): Source: Synthese, Vol.

90, No. 3 (Mar., 1992), pp. 437-458 Published by: Springer Stable URL: http://www.jstor.org/stable/20117007 . Accessed: 07/01/2013 22:05
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Synthese.

http://www.jstor.org

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

SAMUEL

WILLIAM

MITCHELL

DUMMETT'S

INTUITIONISM FINITISM1

IS NOT

STRICT

ABSTRACT:

Michael

Dummett's

anti-realism

is founded given

on

the

semantics

of natural

he argues, language which, It has been objected that into strict finitism. does not My Dummett confirmation happen finitists decide. mett's what view

can only be satisfactorily an analog of Dummett's

inmathematics

purpose

if we have This

cannot abilities confirm we know what would practical vast problems. But to state his case, the strict compute impracticably we can't actually to suppose that we grasp the truth conditions of sentences or the analogy with Dum must be practically demonstrable, comprehension could

successfully that our actual

will argument in this paper is to refute this objection, which I argue to a view of do. I link the coherence of strict finitism

by intuitionism. intuitionism collapse

fails. So, our actual abilities must be capable of confirming that we know were sentences the case if actually undecidable the true, contradicting of confirmation. I end by considering objections. argument would be

DUMMETT'S

CASE

AGAINST

REALISM

One reading of Michael matics can be sketched (1) (2)

Dummett's like this:2

argument

against

realism

in mathe

What our language means is an empirical matter. The sole evidential basis for empirical hypotheses about what humans mean by their words are the uses to which we can put language on different occasions. Hence, possibility about meaning any hypothesis of testing by human usage that goes beyond is illegitimate. the

(3)

as (3) should be interpreted reading of Dummett, point no in that distinction the of is claiming meanings expressions possible in usage by humans is possible, and correlatively, where no difference the claim that a sentence is meaningful is empty if there is no possible use to which can be put by humans. The view requires the sentence On my that some specific kind of test be possible of comprehension of every sentence of the language. This condition is violated theories of by no in test like Donald which of Davidson's, compre meaning specific hension of an individual sentence is required, short of understanding the whole language.3
90: 437-458, 1992. Synthese 1992 Kluwer Academic Publishers.

Printed

in the Netherlands.

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

438
(4) Two

SAMUEL

WILLIAM

MITCHELL

kinds of evidence

of comprehension

are possible

from

(5)

of a subject's verbal ability to use the sentence for example, by explaining what itmeans. appropriately, a of subject's practical ability to use the sen (b) Evidence in acting in appropriate ways upon tence, for example, its truth. recognizing is non-circular. For linguistic the practical evidence Only evidence clearly requires that the speaker be credited with the language used in displaying antecedently understanding But this can only be shown by some further comprehension. test that is not question-begging.

usage: (a) Evidence

a practical test may be very I have a very weak reading of Dummett, effort and ingenuity to de difficult to obtain, or require considerable As Dummett vise. But some test must be possible. puts it, "there must be an observable between the behavior or capacities of difference someone who is said to have that knowledge and someone who is said must be a genuine to lack it".4 It is particularly it that test; important that anyone it cannot be ruled out by what we know about humans as having passed it. It cannot, for example, could ever be recognized require an infinite amount of time. is that its insistence The case against classical logic and mathematics is either true or false leaves much of its language that every sentence test of comprehension. What we test, when we bereft of any practical test comprehension, is grasp of truth conditions and, since according to classical logic every sentence is either true or false, this must be the the sentence as having one or the other truth value. ability to recognize (6) of a decid In finite domains, practical tests of the application to each object in the domain are always pos able predicate to practically test an sible. It is therefore always possible of individual's (for example) "Every even understanding is the sum of two primes". Simply number less than a million set the individual a task in which he or she can do something practical to discover the truth value. A rather cheap example to perform of such an ability is the ability to create a machine a the calculation computer. by programming the domain the case where is infinite. this with Contrast

(7)

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

DUMMETT'S

INTUITIONISM

IS NOT

STRICT

FINITISM

439

of truth conditions is a Obviously, knowledge displaying matter of showing knowledge of what is the case if the sentence is true, and what is the case if the sentence is false. To display such a thing practically, the subject must do something which, putting the matter very loosely, is at least case relevant to the truth value of the sentence. Dummett's case in is that it is ruled out, the of undecidable sentences, that anyone could do anything, under any possible circum stance, that is relevant to the truth value of the sentence. cannot serve So, any theory of truth that retains bivalence as an adequate theory of meaning. I hope, however, This reading of Dummett may be very idiosyncratic. nor wildly that it is neither uncharitable inaccurate. It is a familiar state of affairs to find Dummett for those who read him. I confusing in should perhaps the I have that I say, defending reading given, or not it is actually think this argument is actually compelling, whether Dummett's argument.

REALIST

REPLY

to focus on just one reply to Dummett's argument. The claim in of a decidable (6), that we can always check the application predicate to every element of a finite domain, is obviously false. For sufficiently or simply calculations large domains, large num involving sufficiently our access to practical tests of comprehension is as unobtainable bers, as that which the realist postulates for infinite domains in (7). It is no we could discover whether every even good replying that in principle sum two number less than, for example, of lO10000 is the for, primes, in the case of sufficiently it is either highly probable or large numbers, even certain that we shall always lack the practical capacity to carry I wish Gandy points out that spacetime may can be carried out and, even if computations in the there are only a finite number of particles put into only a finite number of discriminable can in practice only represent a finite number of numbers. And, it follows that there may be finite calculations hence, which the intuitionist for which we regards as having a truth value, the computation. collapse before certain that does not happen, universe which can be we states.5 Therefore, Robin out

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

440
can only

SAMUEL

WILLIAM

MITCHELL

affirm our grasp of their truth conditions, exactly verbally to the charge against the realist. analogously limitations upon our ability to that holds that practical The position must in be taken into account represent numbers and perform proofs is called strict finitism by Dummett.6 What the claims of mathematics concerns me is the use of strict finitism by realists against intuitionists. and that the is internally consistent My aim is to show that intuitionism claim by realists that the intuitionist against them can be complaints is false.7 by the strict finitist against the intuitionist exactly duplicated I will argue that in the sense of 'practical demonstration' required by all that we understand demonstrate the intuitionist, we can practically the intuitionist claims we can. and only those sentences I want to look at Dummett's Before turning to this, however, reply to this challenge, which I think is inadequate.
DUMMETT'S OWN REPLY

own

Dummett
vague

argues

that,
such

to make
as, "...

the case,
is in

the strict finitist must


countable", or "

rely on
. . . can

predicates

practice

a variety of ingenious rea in practice be proved". Dummett provides cannot be given an accept sons for supposing that any vague predicate the case the strict finitist cannot make able semantics. Therefore, no in which for there is vocabulary acceptable against the intuitionist, to do so. In the first place, it relies upon a I find the argument unsatisfactory. on the part of the opponent. lack of ingenuity certain by Perhaps, a semantics should meet, an oppon our what conditions view of revising For example, in giving ent may give a semantics for vague predicates.8 considers Dummett his counterarguments, only briefly the possibility It might be difficult to spell this idea that truth itself may be vague. for example, out; one would need some notion of partial entailment In the second but I would not be willing to suppose that it is impossible. are very useful, and Dummett's ap argument place, vague predicates if we have to give up that we give them up. But, pears to demand for then much of the interest of semantics evaporates, vague predicates, concern vague predicates. and scientific problems many philosophical it seems to me there is a because I will not develop these objections Strict finitists are a rare breed. The strict far cleaner reply to Dummett. is usually presented finitism objection by realists who want to know

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

DUMMETT'S

INTUITIONISM

IS NOT

STRICT

FINITISM

441

why intuitionism does not collapse into strict finitism. Such an individual can present the case by using no vague predicates. to be the following of PM I define The Chinese superfactorial
number:

Take
matica,

the factorial

of the G?del

number

of Principia

Mathe

superexponentiate Chinese arithmetic


pi-10

it to its own power,9 the result with the decimal

expansion

of

A Big

to be any number greater than the Chinese is defined PM. of superfactorial so Dummett This is clearly an intuitionistically definition, acceptable as must claim that Big numbers exist. But we are almost certainly as are we these of of check actually going through operations incapable whether every even number is the sum of two primes. ing exhaustively is How can the intuitionist claim both that grasp of these operations our we that of and the abilities grasp practical application displayed by number to such huge numbers? Patently, we do not have such these operations abilities. practical For this reason, I do not think Dummett's In suggest reply succeeds. own I will that finitism An first strict is incoherent. argue ing my reply, adequate requires theory of the practical abilities we actually possess that we be accorded possession of practical abilities we cannot actually of these abilities has been justified in a exercise. Once our possession manner I think Dummett I will argue that they are would accept, that we could discover sufficient to establish the truth values of those as sentences the intuitionists truth values, while they are regard having not sufficient to provide us with the ability to perform infinite tasks.

WHY

I AM

NOT

STRICT

FINITIST

I argue that strict finitism In this section, is incoherent. In order to state their position, the strict finitists must hold a theory of confirmation that they deny to the intuitionists. The fundamental distinction between the intuitionist and strict finitist is that the intuitionist holds:

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

442
A

SAMUEL

WILLIAM

MITCHELL

sentence

principle and the strict finitist A sentence

if and only has a truth value discover which truth value it has, holds: has a truth value truth value if and only it has.

if we

could

in

if we can in practice

discover The strict finitist's

which

is that since we are prevented from ever argument some at sentences for the truth values very actually arriving concerning large numbers, we have no more practical abilities associated with the the intuitionist discovery of such truth values than we do for sentences

allows may be undecidable. Of particular interest, for the argument I want to make, is the position sentences that the intuitionist thinks are of the strict finitist concerning concern are too to but numbers in which that decide decidable, big + or An 1 Dummett is is either "1010a1? gives example practice. prime for brevity, refer to this sentence by V.11 An intuitionist composite", is either true or false, in virtue of the fact that claims that the sentence it. It is irrelevant that the physical could decide limitations of the universe may prevent us from being able to do so in practice, the only are to and mathematical relevant truth falsehood mathematical thing means and refuting sentences.12 By contrast, of establishing the strict finitist points out that the argument against the realist rested upon our to understand and we have no abilities the truth predicate, practical as applying to a sentence practical ability to recognize a truth predicate we cannot in practice verify or falsify. So, a is in the same boat as "Goldbach's is either true or false"; since we cannot discover conjecture which alternative lack holds, we must allow that the disjunction might a truth value, just as the intuitionist allows in the case of Goldbach's conjecture. The point I am emphasizing here is this: strict finitists hold sentence we

(A):
(A) is confirmable if the antecedent No counterfactual cannot practically counterfactual be brought about, the physical constitution of the actual universe. of due that to

Both must

the intuitionist and the strict finitist require that understanding if it is to be genuine. The be confirmed by our practical abilities stems from the intuitionist bite of the strict finitist arguments against

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

dummett's

intuitionism

is not

STRICT

FINITISM

443

to claim that we can confirm our the fact that the intuitionist wants sentences in such a way that sentences of mathematical understanding like a must have a truth value even when we cannot in practice deter mine which truth value they possess. The strict finitists deny this claim to be actually capable of abilities because they require all practical being implemented; we cannot confirm claims about what would happen in circumstances that cannot actually be brought about, so we cannot we would arrive at a truth value for a. Sentence confirm that (A) is a it mean, bit vague. Does "Given all the actual facts about the physical . . .", or "Given the amount of matter in the actual uni universe verse . . .", or what? I will shortly argue that no strict finitist can hold any version of (A), so the reading given to it is irrelevant. is not alone in holding that we cannot confirm The strict finitist if that confir in this case our comprehension of a sentence, something, mation Mach, is ruled out by the physical contingencies of the universe. Ernst is not entitled to (putatively) charitably read, argues that Newton from relative by an experiment we cannot absolute motion distinguish in practice carry out.13 Similarly, Bas van Fraassen holds that practically cannot affect the choice but never performed, experiments possible, between theories.14 We should, then, be at least agnostic about coun On a closely terfactuals involving practically impossible experiments. related

topic, the same author points to the irrelevance of practically van to the practice of science.15 Admittedly, impossible considerations Fraassen is not the exact corollary of a strict finitist, for he is simply about the truth values of practically untestable sentences, agnostic rather than denying them a truth value or granting them some truth value apart confirmation that cannot truth and falsehood. My point is that his theory of in claiming that sentences agrees with the strict finitist's cannot be in practice be confirmed through observation at all. Though strict finitism is a prodigy confirmed in the philosophy of mathematics, it is a familiar beast in the philosophy its of science; name is empiricism. note the following: the strict finitist case argument, at least understand some sentences we are physically as true or false. (The disjuncts of a, above, recognizing from

such sentences.) The strict finitist cannot state his or we case without understand such sentences. Further supposing our ability to identify such sentences counts as a practical test more, of our comprehension, for the physical limitations of our universe are her

To begin the requires that we from prevented are examples of

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

444

SAMUEL

WILLIAM

MITCHELL

one we discovered an empirical matter, by doing practical things, like and observing the red shift of galaxies. building telescopes if the grasp of truth conditions Now consider what must be established on both the intuitionist and strict of any sentence is to be established, It is that if the sentence in question has a truth value, finitist accounts. we then would be able to recognize that truth value, if we were pre a sented with proof. So, what the strict finitist must hold, in order to hold that we grasp the truth conditions of a practically unverifiable sentence the intuitionist calls decidable, is that we can demonstrate that we could recognize the truth value of such a sentence if the universe were physically different. But, clearly, this violates condition (A). Let me spell this out. Let sentence s be any intuitionistically-decidable sentence that the strict finitist claims is undecidable sons. To understand s, our behavior must confirm: (B) We could recognize different. for practical rea

the truth value of s, if the universe

were

physically

Why does a strict finitist have to hold (B)? We must be able to under stand s for the strict finitist to be able to state his or her argument. in the ability to recognize Our understanding consists the truth value of 5, if it has one. But, s would have a truth value if and only if we s were true if the could calculate it. And, we could calculate whether different than it is. So, the strict finitist must universe were physically s would have a truth hold that if the world were physically different, because we understand value which we could calculate. And, s, we would then recognize that truth value. the antecedent But (B) is a counterfactual of which we are quite evidently unable to bring about. Strict finitists cannot state their position without identifying a sentence like s. So, they must claim we understand s. As understanding we must be able to confirm must be confirmable, s. But, any theory that they do understand by their practical abilities of confirmation that allows this also allows the confirmation that in such a universe we would arrive at a truth value for s. And this the strict finitist must As already to predictions Fraassen analogous available so strict finitism is incoherent. deny can be confirmed, that induction should be limited noted, Mach suggested about those cases we are physically able to test, and van I think an the practicable. is agnostic about induction beyond to that will show that the stated argument just only option is to believe the issuances of the theories we hold for cases

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

dummett's

intuitionism

is not

strict

finitism

445

the physical circumstances in which we find ourselves. For are in which we find ourselves limitations of the universe to discover whether If a physicist wishes discoveries. the empirical to future is open and infinite in time, or closed and doomed eventually in on itself, he or she must do something and this collapse practical, is something physics and activity, being part of the physical universe, to the other sciences are required to explain. Suppose one experiment our practical discover limitations is that of measuring the relative that outrun the physical shift of light from approaching and receding arms of galaxies, Doppler and suppose as a result of this experiment we conclude the universe is
closed.

In order experiment,

to explain the activities of the physicist we have to suppose he or she understood

if the universe had been open, otherwise, happened But understanding why this conclusion was drawn from this experiment. this involves knowing that a different have been conclusion would if the results of the had been different. And appropriate experiment and the rest of us if we that the experimenter, that, in turn, means of the experimental result, must hold that the accept the significance about what would have happened if the uni theory makes predictions verse had been different. So, the option is not open to us of restricting the implications of theories we accept to those we could in fact encoun the range of observable ter, given the universe we inhabit. Once phe nomena for which science must account is extended to include ourselves and the phenomena of language use, there are good empiricist reasons for recognizing the truth of counterfactuals which our theories predict even when those counterfactuals can only be indirectly established by the success of the theory in other cases. to my main concern. For which abilities This argument is relevant humans possess is an empirical matter, and a theory of such abilities will issue in predictions for circumstances other than those we actually encounter. These include circumstances in which we are furnished with theorems than are actually greater resources for proving mathematical
the case.

in performing the what would have we cannot explain

view I hold is that we know the Chinese of PM is superfactorial either prime or composite because we can show that we could calculate which it was. We can actually calculate whether quite small numbers are prime or composite, and that demonstrates that we could decide the matter for Big numbers if time and resources allowed. The strict The

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

446

SAMUEL

WILLIAM

MITCHELL

is to deny induction from actual to impractical finitist reply, of course, cases. But, as I have shown, this reply is defective; it depends on an can view of what actual experiments confirm. untenable

WHAT

'COULD

IN

PRINCIPLE'

MEANS

Because

I have shown, I think, that strict finitism is incoherent, I have from 'the frying pan' of having to say why I do not delivered myself unless I can show that the intuitionist hold that position. But, obviously, I shall simply fall 'into the fire' of has a distinct consistent position, to absurdity. In this section Iwant to say what reducing my own position

for the intuitionist in the sentence "A exactly 'could in principle' means sentence has a truth value if and only if we could in principle discover which truth value it has". Having said that, I will go on to argue that are sen to be decidable the intuitionist believes just those sentences tences for which we could in principle discover the truth value. One common reading of the phrase is that "could in principle" means if our practical abilities were extended "could by a finite amount". for example, takes this to be the correct Crispin Wright, that the important don't think this can be sustained. Recall the intuitionist criticism of the realist was that we must of expressions confirm our understanding by our practical we give up the claim that these practical abilities must reading.16 I feature of be able to abilities. If be actually to realism by us, then the central premise of the challenge possessed anything by exercising disappears. We can't confirm that we understand because we can't exercise practical abilities we don't actually possess, such abilities. leaves open the issue of what it is that Such a concession, however, the abilities we actually possess show we could do if other factors were

that kind of performance varied. An ability is a particular capability need But those circumstances will be evinced in the right circumstances. reasons for supposing not actually be present for us to have excellent the performance take place. Could would that, if they were present, Duns Scotus have understood Of course he could. He Esperanto? It is irrelevant understood Latin, and that is a much harder language. was not invented until centuries If after his death. that Esperanto not not but then it is time travel is physically actual, only impossible, to the right that he could have been exposed physically impossible,

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

DUMMETT'S

INTUITIONISM

IS NOT

STRICT

FINITISM

447

conditions

to demonstrate

such

a skill,

yet

that does

not

call

into

its ascription. question can be liberal about what We is demonstrated by the abilities we we must vary our own the suggestion that actually do possess. Unlike this does not undermine the intuitionist abilities, argument against realism. Furthermore, the argument against strict finitism shows that we simply must be liberal about what our actual abilities demonstrate we could do in other circumstances. For it can hardly be denied that we recognize cannot be solved in that certain mathematical problems these practice due to physical facts about the universe. To understand is to know their truth conditions, which in this context means problems we would be able to recognize a proof of the result if one were presented to us. But, for many of these problems, the only proof available will too long for us to comprehend. be one that is in practice So, some of our present practical abilities, those of discovering the limits of calcu we could recog lation our physical universe sets upon us, demonstrate nize proofs that in practice we cannot. So we have some practical abilities skills that cannot be evinced except in that show we possess situations. counterfactual is this: "A sentence has a truth value if Therefore, my suggestion and only if we are in principle able to discover which truth value it has" is to be parsed: "A sentence has a truth value if and only if our actual abilities would be sufficient for us to discover which truth value it has, were the appropriate triggering circumstances present". What kind of abilities am I talking about here? In the mathematical case, these certainly include things like the ability to count, add, multi ply and divide. They also include the ability to recognize proofs, and differentiate and integrate formulae. They also include abilities associ ated with the communicability of mathematics. One of the major differ ences between Dummett and earlier intuitionists is his explicit recog nition of the public nature of mathematics and language. Thus, the our grasp of the in demonstrating practical abilities we may exemplify are not those of an incommunicable truth predicate mental conception of proof, but include things like the ability to communicate with one another and keep records. We possess these abilities because are. But, the abilities can nonetheless the physical details of their causes familiar one from the philosophy of the kinds of physical objects we be characterized of independently and instantiations. The point is a of mind and computer science. We

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

448

SAMUEL

WILLIAM

MITCHELL

can specify what something must do to approximate square roots using method. Newton's that actually does so is a physical object, Anything on the but this way of describing what the thing does is supervenient are all minds with which I am acquainted physical details. Similarly, but a statement that one of them is in a mental objects, physical state is not a statement that uses a physical vocabulary. We are slow, but can keep inefficient, calculating objects. We have small memories, as a community. With records or solve problems of the the exception nature of mathematics, inclusion of the communal these kinds of abili ties are exactly those the intuitionists used to found their mathematics, in a way that is independent of virtually every physical time. except predicate It is because of this fact that only mathematical barriers to proof are relevant to mathematical truth.17 The abilities we have to prove things are independent of the physical contingencies of the universe, and it is a proof that is relevant to mathematical the ability to provide truth. and are described For assert that were the same reason, the intuitionist may legitimately or even circumstances itself these different, different, physical physics issue in decisions for problems we are in practice unable abilities would to decide and, hence, that sentences stating such puzzles have a truth that truth value is. And, value, though we shall never know what our mathematical because the exercise of such abilities constitutes prac for the realist to answer intuitionist objections tice, it is illegitimate by that we could speak with a community of angels who tell us proposing that they have verified that by exhaustively checking the even numbers each is the sum of two primes. Even of the abilities and if, convinced a this constitutes honesty of the angels, we would accept their word, radical change in our mathematical what counts as a practice. Although to characterize proof may be subject to evolution and, hence, impossible this constitutes of our current any semblance abandoning precisely,
notion.

need have truth values, On Dummett's view, not all counterfactuals so it remains a question whether "If the universe were physically differ ent we would conclude that '101Oa1? is prime' was either true or false" has a truth value. As an example that probably of a counterfactual lacks a truth value, Dummett Jones, gives that of the now-deceased we out in his life. Dummett who has never faced danger that points to think that "if Jones had faced danger he would have little inclination either have acted bravely or acted like a coward" must be either true

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

DUMMETT'S

INTUITIONISM

IS NOT

STRICT

FINITISM

449

or false.18 But as Dummett himself points out, our lack of inclination a a to accord such sentence truth value stems from the fact that we at lack any evidence all linking Jones to bravery or cowardice. If we on toast had overwhelming that all those evidence who enjoy anchovies are invariably courageous, on toast, and that Jones enjoyed anchovies we have the truth value of the if they are predicted a true theory. In the circumstances by reason to think that there is a true abilities that does bestow a truth value shall know truth values counterfactual. Counterfactuals to be true in the counterfactual case in question we have every theory of human calculational to the relevant counterfactuals.

DISCUSSION

I will conclude have outlined.

by considering

several

realist

challenges

to the view

that human abilities are too unreliable First, there is the objection to support the kinds of conclusions I have rested upon them. If these are our actual abilities that are being considered, then we don't have to calculate whether the ability, even if physics were vastly different, some vast number was odd or even because, using our abilities, we'd be virtually certain to make an error in the calculation. can draw upon a competence/performance To reply, intuitionists distinction. A practical ability can be meaningfully ascribed to someone even if they sometimes fail to perform They have other accurately. abilities associated with the correction and recognition of such errors that show that their intention was infelicitously executed. A sentence has a truth value if we could discover which truth value it has, and we sentences even when could always discover the truth value of decidable we err in executing we can the algorithm, for always discover the errors. we are sometimes mistaken in thinking something was So, the fact that a proof does not show we can't recognize proofs. An ability may be as one of being able to add one to any arbitrary number, characterized even if in repeat and the characterization will be perfectly acceptable, one loses count this ability upon the stones inWales, edly exercising inMid Glamorgan.19 statements counterfactuals Perhaps while concerning mathematical that their antecedents be practically real require for their confirmation sometimes izable, statements about empirical matters do not. Dummett mathematical from physical so, perhaps, distinguishes subject matters somewhere

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

450

SAMUEL

WILLIAM

MITCHELL

In which case, the strict finitists do the strict finitist can do likewise. not contradict in accepting (A) for mathematical themselves statements, but denying (B) for empirical ones.21 and in the argument In other parts of Dummett's work, however, at the beginning of this paper, Dummett summarized clearly points out to both mathematics and empirical that his argument applies equally to not collapse is this does The say argument why challenge topics.22 into strict finitism and, since this argument applies equally to mathemat to introduce this distinction it is illegitimate ics and empirical matters, in an attempt to rebut it. underlies the Another challenge: What theory of counterfactuals to tie my account to a specific claims made here? It would be unwise or possibility, but I do want to point out that theory of counterfactuals there is at least one anti-realist theory which is friendly to it. Recently, van Fraassen has suggested an anti-realist account of possible worlds.23 allowed by a theory, this view, a possible world, Under is simply a a this to the view I am proposing, model of that theory. Applying sentence of mathematics has a truth value if and only if there is some abilities in which we of a true theory of our actual practical model That makes the truths of discover the truth value of that sentence. but mathematics upon the facts about human capabilities, dependent was the whole point of the intuitionist critique of classical mathematics is a human activity makes our capabilities that the fact that mathematics the abilities we possess relevant. As I pointed out in the last section, we have in virtue of our physical structure, but those abilities may be It therefore does not follow, inmany other physical models. instantiated from the fact that the universe would collapse before we could calculate that the sentence might lack a truth value, the truth value of a sentence, or otherwise fail to be either true or false. In other physical models, our its truth value. with abilities to calculate do discover beings Let me show how such a theory might be used to reply to another that our actual practical abilities may be I have emphasized objection. the challenge in different physical instantiated systems. This suggests are just Turing machines, sentences will then many that if humans is that are not intuitionistically be provable provable.24 One example term will 'M': which I Markov's principle, (Vx(Ax The sentence v ~iAx) & -i Vx~iAx) if there ?? 3xAx of creatures with human

is provable

is a model

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

dummett's

intuitionism

is not

STRICT

FINITISM

451

is that there will always abilities who actually prove it. The objection for this principle, since there is an algorithm for be such a model an instance of the consequent from the antecedent. The left generating that creatures with our abilities guarantees conjunct of the antecedent can tell of an arbitrary number whether it satisfies the predicate or not; the right conjunct shows they can't all fail to do so. So, simply take a in which time and resources do not place an upper bound on universe how large a finite number we can check, and go through the numbers is found. But, of course, the right until an instance of the consequent conjunct doesn't show what we need; all it shows is that we can reduce that the predicate holds of an arbitrary finite number demonstrations so to absurdity, intuitionistically speaking, even a model without a finite the needed instance. for the upper bound cannot guarantee (Granted, sake of argument, the actual abilities we possess could show we can of such a model.) to turn back the objection. After this is not enough But, perhaps, of the quantifiers in all, the only case for taking our understanding rests upon what we could do. In contrast to the real tuitionistically to me to refuse to affirm that it is not available intuitionists, perhaps an instance of the consequent will eventually be found. Consider, for is decidable if and only instance, the following: on my view a sentence if there is a model of creatures with our practical abilities who actually decide it.Well, then, cannot the above argument use a model of human in which to abilities there is no upper bound? And, if I am allowed use such models for impractical calculations, why can the Markovian sense of their own not press it into the service of making constructivists conceive quantifiers? a further restriction on our use of models: we There is, however, our must be able to confirm by actual demonstrable abilities that we in question.25 In order to genuinely in grasp the models grasp a model to is calculated by beings like us, it is first necessary which M, above, inM. But, to reiterate Dummett's of the quantifiers grasp the meaning there is nothing we can do, save simply to claim we under argument, stand the quantifiers in the constructivist way and, so, no way to confirm this claim. Dummett's of constructivist argue against the application might to this case. In the constructivist case, unlike the argument can only be proved realist one, an existential if an in generalization as case. seems as stance of it is proved, in And intuitionist the it just A Markovian

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

452

SAMUEL

WILLIAM

MITCHELL

for in cases in if an instance of the consequent always can be proved, to be satisfied, we can actually of M is known which the antecedent in a direct way and so demonstrate find an instance of the consequent, our understanding of the quantifier as it appears inM. that in the actual cases we have no guarantee But can we? We In cases in which the instance of the consequent. actually shall discover we lose interest, or funding, for the project before its completion, the Markovian is reduced to the claim that, had we only gone on, we would to me it seems that the the instance. But, have discovered again, in this claim is of the quantifier embodied understanding putatively illusory and that even the myriad cases in which we actually do find an do not in fact confirm what the constructivist instance of the consequent come up wants. is that even when we do not actually The reason still claims the the Markovian with an instance of the consequent, can confirm anything un is not refuted. No experiment understanding be that it could fail, and the understanding less there is the possibility to will allow no failure of the experiment refuted. The constructivist of the quantifiers. So that under refute the proposed understanding and the process of going through the standing cannot be confirmed, even when in finding an instance does not calculations they succeed claimed. confirm the understanding for To make the point vivid, contrast it with the case of a sentence sense that in the intuitionistic which we possess a decision procedure are in practice unable to carry to completion because of contingent some very facts about the universe. (For example, factoring physical In this case we can use the facts we know about how large number.) quickly humans can calculate in order to establish what kind of universe to be completed, how old it for example, would allow the algorithm involves ex of such a universe would have to be. The specification to that we can actually use, for example and knowledge pressions occupy someone for a fixed period of time by asking him or her to factor so there is some genuine some number, check on the comprehension we from the contrast, we cannot guarantee, that we shall be of the above schema, in which the consequent is satisfied in a way able to specify any model that brings such practical abilities in its train. We can't say simply, "the to actually find far enough in which the computation model proceeds cited the number", since that answer falls to Dummett's objection, our at the beginning of this paper; there is no way to demonstrate of the expressions used. By satisfaction of the antecedent

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

DUMMETT'S

INTUITIONISM

IS NOT

STRICT

FINITISM

453

that is not question-begging. of that answer And, comprehension in which creatures with our abilities clearly, we cannot say, "the model for there is no such model. check the complete totality of numbers", to force in This raises another controversy: Could realists attempt our to finitism that tuitionists back strict by challenging practical abilities are independent of physics? I cannot claim to have a decisive argument that this cannot be done, but it seems to me to be very difficult. In the even the empiricists, out that everyone, must last section I pointed we some admit that have practical abilities the physical world prevents us from exercising, out the ability to recognize namely, experimental comes that would demonstrate that the physical world is otherwise than cannot take the form of claiming it is. So, the realist argument that intuitionists must only talk about physically actual circumstances. The most promising line of attack seems to be that of denying the separ abilities by attacking the claim that ability of physical and mathematical on physical ones. mathematical abilities are supervenient But consider how little the intuitionists need. Mathematics is founded abilities; we can count one syllable num upon familiar mathematical bers at the rate of about one a second, we can add two one-digit and so on. These statements of ability don't on the face of it numbers, for their realization. The reason require a specific piece of physics are so useful is just that they do what we intend to do at computers much higher speed in a different physical way. All the intuitionists need to new of our abilities is to extend our present mathematical practice cases. We might be a blind, paralyzed, in non-actual heights ageless community of mathematicians living in a vast cave, organizing ourselves into huge human computers, each of us entrusted with remembering a deals, just part of the immense numbers with which the community or calculating in the Promethean just a single operation proofs we It is a short step from that to being completely disembodied perform. in which case we can dispense with the physics altogether (al spirits, though it is hard to make sense of practical abilities deriving from grasp of mathematical concepts by a non-physical being). two senses of the sentence Alexander schema George distinguishes "X can in principle construct a proof that P".26 Under the first reading, called the narrow scope by George, the sentence means that X knows a strategy that, if it were carried out, would establish the truth value of P. For example, knowing how to generate prime numbers and divide a large number by them successively might be an in-principle-proof

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

454
that the Chinese

SAMUEL

WILLIAM

MITCHELL

is either prime or composite. The wide superfactorial here, is that X could construct a canonical proof reading, emphasized that P under certain circumstances. I have several interconnected about the narrow scope. misgivings the realist from having an The first is that I cannot see how to prevent or its negation of Goldbach's conjecture by saying in-principle-proof we could check whether is the sum of two primes. every even number as George cannot by himself cogently Second, argues, the intuitionist this means indicate what he or she means by 'finite' or 'effective proced ure' in an acceptable that the way, and the narrow reading requires rather than the other abilities be defined in terms of these expressions, the way around, as I have suggested here. Third, and most important, that truth must be accessible is based upon the must be non-circularly that understanding demonstrable the narrow reading, our recognition of an by practical abilities. Under does not require that we practically be able to cash in-principle-proof out the activity of proving in a practical way, even in counterfactual intuitionist demand requirement the wide reading, by contrast, our comprehension circumstances. Under of a sentence, and its possession of a truth value are always explained by the abilities we actually possess. There could always be a discoverable that hangs on the truth value, of a sentence, al practical difference though our own world may be too sparse for us to be able to take this anti-realist theory of truth advantage of it. So, the central question on the narrow reading. to address goes unanswered is supposed on Rules and Private Language Saul Kripke's Wittgenstein addresses a closely related but distinct question to that raised here. Kripke asks from actually practically tested uses what it is that allows the projection a uses. to future actually practicable mathematical function I have of been addressing the issue of how our present restricted abilities could serve as a public test of what we would do in other circumstances. the divergence between Kripke's purposes and my own, he, at Despite one point, appears to contest the view that the construction of a ma a function demonstrates chine to calculate that an individual knows which function she or he intends.27 These arguments might perhaps be extended to untested to vitiate cases. the extension of other demonstrations of abilities

Kripke has three arguments.28 His first points out that a computer so there will still be untested to calculate, only has a finite capacity cases in which the skeptic can doubt the algorithm. Clearly this is true,

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

dummett's

intuitionism

is not

strict

finitism

455

with values but the task is only to show practical abilities connected to have beyond the function is supposed those so far explicitly calcu shows this even if it only has lated. The ability to construct a machine a finite capacity. If a doubt arises beyond this capacity one need only a build bigger machine. cases for reply that there will still be untested Kripke will perhaps the larger machine. is But that entirely beside the point: the ability to construct a machine is a practical demonstration of what is intended can calculate, and the ability to construct that a machine for arguments shows what values are intended for larger larger and larger machines this for indefinitely-large and larger arguments. Doing finite arguments is identical to showing that we understand the function. We do not have, nor do we need, a practical ability to show we know what all the values of the function will be for the entire totality of natural numbers, retreat to for that totality does not exist. I think the objector must numbers so large that the physical universe prevents us from calculating the values, and I have already addressed such a position. so that a that machines malfunction, Kripke points out, secondly, I intend. But this over the procedure may fail to instantiate to detect malfunctions. looks the fact that we have practical abilities or use the output We can run the algorithm on two different machines, as input to another program that should give easily surveyable and and so on. I can show that I know what it would outputs, predictable be for a machine to correctly instantiate my intent, often through my machine attempts automatic
about.

to prevent calculator

errors. So, once again, my ability to build an that I know what I'm talking gives some evidence

the embodies Kripke argues, thirdly, that if we say that the machine as an embodiment function we intend, we must explain its workings of the algorithm. To do this we have to use language, and the strict finitist can ask whether we really grasp the meanings of the words we use when we refer to cases outside those we have explicitly calculated. Now here, I think, the divergent purposes of myself and Kripke come to the fore. For my reply is simply that I do not have to explain the as an embodiment machine of the algorithm at all. The construction of the machine is, for me, a means by which a third party can test my which does not depend on what I say. If those who comprehension doubt my comprehension which will say to me, "Build us a machine mean by 'plus'", and a what mathematicians compute automatically

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

456

SAMUEL

WILLIAM

MITCHELL

is produced, machine then as far as they they judge to be appropriate are concerned that is very good evidence matters that I understand to their lights without my saying anything else. Even if the according machine computes only a finite initial segment of the plus function, to the is a ready explanation for this in the materials available which can be tested by running the experiment subject, an explanation resources. There is the further issue, of again with greater available both they and I intend the 'right' function by course, about whether 'plus', but that is not my concern in this paper. The point I've been arguing might be put this way: the practical activities of empirical formation, hypothesis testing and confirmation entail that the only available hypotheses about how we would behave there inaccessible circumstances in the way we intuit generalize ively accept as correct from those we can actually test. In this light, the strict finitist challenge becomes: say what is wrong with a practice of etc., which predicts the intuitively accept smonfirmation, sminduction, we can practically able result for hypotheses test, but which makes no predictions, or different predictions, about practically inaccessible in practically
cases.

As I've argued, one must make some predictions about the inaccess become ible cases, or certain actual pieces of behavior unintelligible. to And as soon as one tries to give an example of a possible exception an intuitively about how we would go on generalization acceptable one has to presuppose that one understands the example. calculating, a an As this is in practically inaccessible realm, this attribution example of comprehension requires that the practical abilities in which compre case. But, of to the inaccessible hension consists can be generalized a is refutation of the challenge that is supposed course, that presumption to be presented. Dummett argued in his review of Wittgenstein's philos that it was the essence of language that we should ophy of mathematics sentences in that understanding and, hence, acquired to new cases.29 should be generalized range of examples even beyond My argument has been that we are forced to generalize cases we shall ever encounter and, so, we would carry on calculating in the same way as we now do if the universe were physi and proving understand a restricted cally richer than it is. I think these are the most pointed objections to the position I have a I at I While in retain whether have disquiet correctly developed. resources to I have think his does the Dummett, philosophy terpreted new

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

DUMMETT'S

INTUITIONISM

IS NOT

STRICT

FINITISM

457

into strict finitism, which reply to the charge that it collapses the most common and robust charge leveled against it.

I think

is

NOTES want to thank Alex George I especially I'd also like to thank of the members Braaten, to Peter Jay Garfield, Godfrey-Smith 1 and Philip Kitcher for their help on this paper. the Propositional Task Force, Jane Attitudes Lee Bowie, Murray and Tom Tymoczko. thanks also My Kitely one of whom was and the anonymous reviewers of Synthese,

particularly helpful. 2 is clearest. Dummett See also Dummett (1978, pp. 216-20) (1976, pp. 80-101). 3 See Davidson 1984, pp. 22, 133. 4 Dummett 1978, p. 217. 5 1982, p. 131. Gandy 6 Dummett 1978, pp. 248-68. 7 P. Bernays, in Mathematics' for example, 'On Platonism and See, (in Benacerraf 1964 (1983), p. 265). Putnam, 8 in Wright such a semantics, which 1987, pp. 167-75) (1982, Wright reprinted gives "... is actually in the metalanguage. contains the vague predicate verifiable" 9 m superexponentiated to the n is m raised to its own power n times. 10 is the operation of summation if one of the Chinese arithmetic without carrying; so is the value. is unpatterned, arguments 11 If you are not Dummett '1010' as an exponent. 1978, p. 239. I use '10*10' to write some much bigger number this example, is used. just imagine happy with 12 Dummett 1977, p. 19. 13 Mach 1893 (1960), pp. 281, 284, 340-41. 14 Van Fraassen 1980, p. 60. 15 Churchland and Hooker 1985, p. 255. (eds.), 16 1987, pp. 107-75. Wright 17 Dummett 1977, p. 19. 18 Dummett 1978, pp. 15-16. 19 Adams 1988, p. 220. 20 For example, in 1977, pp. 55-56. 21 was suggested This argument reviewer for Synthese, whom I should like by a helpful to thank. 22 For example, Dummett 1978, pp. xxix, 16-17, 157, 227; Dummett 1982, pp. 55-56. 23 Van Fraassen 1989, pp. 92-93. 24 was suggested This useful objection to whom I am grateful. reviewer, by an anonymous 25 This is the burden of The of G?del's Theorem' Philosophical Significance (Dummett 1978, pp. 186-201). 26 1988, p. George 27 1982. Cf. Kripke 28 1982, p. Kripke 29 Dummett 1978, 152. Dummett 32ff. p. 177. 1978, pp. 171-72.

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

458

SAMUEL

WILLIAM

MITCHELL

REFERENCES Adams, D.: Tea Time 1988, The Long Dark H.: 1964 (1983), P. and Putnam, the Soul, Pocket Books, New York. Read

of

Benacerraf,

Philosophy

of Mathematics,

Selected

Press, Cambridge. University ings, Cambridge P. and Hooker, C. A. Churchland, 1985, Images of Science: (eds.): of Chicago, and Empiricism, University Chicago. D.: into Truth and Interpretation, Clarendon 1984, Inquiries Davison, M. A. E.: 1973 (1981), Frege, the Philosophy Dummett, of Language, MA. Press, Cambridge M. Dummett, J. McDowell pp. 67-137. M. M. A. A. E.: E.: A. E.: (eds.), 1976, Truth 'What is a Theory Essays and Meaning: of Meaning? in Semantics,

Essays Press, Harvard

on Realism Oxford. University

and IF, in M. G. J. Evans Clarendon Press, Oxford,

Dummett, Dummett,

Oxford 1977, Elements Press, Oxford. of Intuitionism, University Harvard 1978, Truth and Other Enigmas, Press, Cam University in D. van Dalen, D. Knowledge', Com '80, North-Holland Publishing

bridge MA. to Mathematical R. O.: 'Limitations 1982, Gandy, Lascar and J. Smiley (eds.), Logic Colloquium pany, Dordrecht, pp. 129-46. George, Alexander: 1988, 'The Conveyability Journal

an Essay on Mathematical of Intuitionism, 133-56. YJ, Logic of Philosophical Cognition', on Rules Saul: and Private Harvard 1982, Wittgenstein Language, University Kripke, MA. Press, Cambridge Ernst: 1893 (1960), The Science La Salle IL. Mach, of Mechanics, Open Court Classics, Newton, Isaac: D.: 1729, Principia, trans. Andrew Motte, University of California Press, Berkeley. Van Dantzig, Van

'Is 1010'10 a Finite Number?', Dial?ctica 10, 273-77. The Scientific Press, Oxford. Image, Clarendon Van Fraassen, B.: 1989, Laws and Symmetries, Clarendon Press, Oxford. and Truth, Basil Blackwell C: 1987, Realism, Ltd., Oxford. Meaning Wright, 1956, Fraassen, B.: 1980, Dept. Mount of Philosophy Holyoke College

South Hadley, MA 01075


U.S.A.

This content downloaded on Mon, 7 Jan 2013 22:05:43 PM All use subject to JSTOR Terms and Conditions

You might also like