Phil 2511: Paradoxes

Lecture 5:  Sorites I: The Problem and Some Standard Solutions

 

First – here’s the paradox of MINIAC that I should have included in the notes to Lecture 4:

 

MINIAC

Take a coin and designate one side `Yes' and the other `No'.  Think of a question for which you would very much like an answer, e.g. `Will I find the girl of my dreams?'.  Toss the coin and note the answer it gives.  But how can we tell whether this answer is true or false?  Easy.  Ask the question `Will your present answer have the same truth value as your previous answer?', flip the coin and note the response (`Yes' or `No').  If the second response is `Yes' then MINIAC's answer to the first question was true; if the second response is `No' then the answer to the first question was false, so now you know for sure whether you will find the girl of your dreams.  Proof: (I'll just do the proof for the case where the second answer is `No.')  Suppose the second answer is `No'.  This answer must be true or false.  If it's true then the answer to the first question is false.  But, if the answer (`No') to the second question is false, then the truth-value of the second question must be the same as that of the first, so, again the answer to the first question is false.  Therefore, if the second answer is `No' we have proved that, whether or not that answer is true, the answer given by MINIAC to the first question must be false.  By similar reasoning we can prove that, if MINIAC answers `Yes' to the second question its answer to the first question must have been true.  What's paradoxical here, of course, is that one cannot guarantee to get correct answers to momentous questions merely from two flips of a coin.

(Reference: T. Storer, `MINIAC: World's Smallest Electronic Brain', Analysis 22 (1961-2), pp.151-152.)

 

 

We now move to a section of the course devoted to the Sorites paradox.  Last time, I suggested some preliminary reading for this section of the course, and I hope that you have now read at least some of it.  If you want to go deeper into the Sorites problem, I would recommend a collection of readings: Rosanna Keefe and Peter Smith (eds), Vagueness: A Reader (Cambridge MA, MIT Press, 1996), which includes a useful long introduction by the editors, and also Rosanna Keefe, Theories of Vagueness (Cambridge University Press, 2000).  Her Chapter 1 covers more thoroughly than does this lecture questions concerning what vagueness is and what types of solution to the Sorites have been attempted.  Timothy Williamson’s Vagueness (London, Routledge, 1994) is very tough, but very rewarding.

 

1.     The form of the Sorites paradox may be expressed as follows:  X1 is F; If Xi is F then Xi+1 is also F, hence Xn is F, where `F' is a vague predicate such that, for any pair of adjacent Xs, if one member is F then so is the other.  This form of argument is paradoxical because we can choose premises that are evidently true, the reasoning consists only of a simple mathematical induction or, alternatively, of multiple applications of modus ponens, yet, for suitable n, the conclusion (e.g. that 0 is a large number, or that one grain of sand is a heap) is false.  There are many instances of the Sorites.  See, for example, the tadpole/frog case devised by Cargile (Keefe and Smith, p.89).   Vague predicates are tolerant.  As Sainsbury puts it in his text (p.28): `Sorites reasoning depends on the supposition that vague expressions are “tolerant”: small changes don’t affect the applicability of the word.  If someone is tall, so is a person a millimeter shorter; if a collection is a heap, so is one otherwise similar but with just one grain less’.

2.     Suppose that I am leafing, right to left, through a thick book containing 2001 of the flimsiest sheets.  Near the beginning of the book I should correctly judge that the right hand section (RH) is fatter.  And turning one flimsy page will not alter that judgment.  Now, in this case, the principle `If Fatter(RH, LH) then Fatter(RH-1 page, LH+1 page) fails when LH = 1000.  In order to get a Sorites-type paradox, we need to substitute `feels fatter' for `is fatter' in the above principle.  Such paradoxes depend crucially on the employment of what have come to be called `observational' terms - those the application of which does not wholly rest on testable matters of fact.  Thus, while we know from chemistry that a water molecule minus one atom is not water, chemistry cannot tell us whether Jones minus one atom is Jones, or is a person.  Neither the name `Jones' nor the sortal `person' is a term of chemistry.  The predicate `is a person' is applied on the basis of discretionary judgments - and the making of such a judgment may be difficult in the case of a Jones who borders on the state of a vegetable.

3.     The Sorites is extremely important, not because it is an intriguing puzzle, but because it brings into question some of our most fundamental views about language.  As Crispin Wright puts it (1987, p.231 = Keefe and Smith, p.208) `the most fundamental task in the philosophy of language is the achievement, in the most general terms, of an understanding of the nature of language mastery.  It is here, if I am right, that the paradox has most to teach us'.  Likewise, a paper of my own is called `The Sorites as a Lesson in Semantics’, Mind 97 (1988), pp.447-455.

4.     Obviously, most `solutions' to the paradox consist of rejecting either a premise or a principle of inference used in the argument.  Some proponents of the second course say that mathematical induction is invalid.  This does not seem a promising solution because (i) nothing independent of the Sorites would tempt us to think that mathematical induction is invalid; (ii) the Sorites argument need not be viewed as a mathematical induction but as a chain of modus ponens steps.

5.     Few people would say that modus ponens is invalid, but some people have argued that a chain of applications of modus ponens is invalid.  This suggestion is as hopeless as it looks -- modus ponens doesn't get worn out with repeated use.  Several attempted types of solution are explained and criticized by Roy Sorensen, `Vagueness, Measurement and Blurriness', Synthese 75 (1988), pp.45-82, as well as in the general readings I have suggested.

6.     Sainsbury spends most of his time considering two types of approach.  I have found that students often misunderstand this approach so I am distributing the 5 pages of his book where he explains it.  The `supervaluational' account begins with the thought that there are penumbral objects that are not comfortably called either V or not-V.  For example, there are small mounds to which we wouldn't confidently assign or deny the title `heap'.  But we could invent a new sharper predicate `heap*' such that there is a sharp dividing line between those objects that are heap*s and those that are not (see Sainsbury's diagram, p.35).  Call `heap*' a sharpening of `heap'.  There are many such sharpenings, depending on where we divide the penumbral region.  Stipulate that, for some object X, `X is a heap' is true if and only if the sentence is true for all sharpenings of `heap', and false if and only if it is false for all sharpenings. It follows immediately that for an object within the penumbral region for the application of a predicate F, it is, on the supervaluationist scheme neither true that that object is F, not false that that object is F.  In other words, the statement `That object is F’ is neither true nor false.  In classical semantics, every statement is either true or false, so if we accept supervaluationism, we must abandon classical semantics.  Also, on the supervaluational account, the disjunctive statement `The book is green or it is not green’, said of a book borderline between green and blue’ receives the evaluation `true’ on the supervaluational account, even though neither disjunct is true.  Thus, the classical semantics for `or’ is also rejected.  We can show that a statement like `b is F or it is not the case that b is F’ comes out true on all precisifications.  So the Law of Excluded Middle, and, in fact, other laws of classical logic are preserved by supervaluationism – but note the reservations expressed by Keefe and Smith (pp.29-32) regarding the claim that supervaluationism retains classical logic in toto.

7.     Now take the conditional premise of the Sorites: If Xi is F then Xi+1 is F.  On the supervaluational account, this is false, for there will be some sharpening F* of F for which, given an Xi within the penumbral region, the antecedent is true and the consequent false.  Sainsbury's main criticism of this approach is that it has as a consequence that there is some number n such that an n-grained collection is a heap, while an n-1 grained collection is not; and this simply defies our intuitions which give rise to the paradox in the first place.  For further criticisms of supervaluationism, see Keefe and Smith, pp.32 –35, and note that `D’ is short for `definitely true’, and is defined by Kit Fine is his seminal paper reprinted in Keefe and Smith, pp.119-150 (but do not attempt this paper unless you are very confident of your ability in logic).  Note also that Keefe defends a version of supervaluationism in Chaps. 7 and 8 of her excellent book.

8.     Before we meet this Thursday, try to fully understand the supervaluational theory.  A quite different solution – the `degrees of truth’ theory is contained in the next section (2.6) of Sainsbury and, if you have time, read that.  In brief, the `degrees of truth' solution is as follows: where a and ß are adjacent items and a is more V than ß, then the degree of truth of `a is V' is higher than that of `ß is V', and the degree of truth of the conditional `If a is V, then ß is V' is close to, but less than 1.  Since a Sorites argument is a series of modus ponens, the cumulative effect is non-validity.  For a sophisticated defence of a version of the `degrees’ theory, see Dorothy Edgington in Keefe and Smith, pp.294-316.