Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

BETWEEN SEMANTICS AND PRAGMATICS: IN SEARCH OF A LOGIC FOR VAGUENESS

Bart VAN KERKHOVE*

Abstract This paper offers a survey of current logical investigation in the field of vagueness. Its ambitions are, first, to reach every interested philosopher, remaining as general and non-technical as possible, without however being incomplete or inaccurate, and, second, efficiently to refer the readers keen on more to further literature. The article begins by spelling out the problems raised by sorites-paradoxes both formally and informally. Further paragraphes then focus on traditional, strictly semantical treatments for these paradoxes (such as Epistemicism, Supervaluation and Fuzzy Logic), followed by more recent attempts to expand the field of investigation to pragmatics. Finally, the heavy metaphysical debate concerning the origin of vagueness (word or object) is highlighted.

1. Introduction

Despite their antiquity and numerous ambitious attempts at their resolution, Sorites Paradoxes remain of the most astonishing philosophical knots. The famous question being how to turn a non-heap of wheat into a heap of it (i.e. the original Sorites),1 or a non-bald person into a bald one (i.e. the so-called Falakros), when adding grains or removing hairs only one at a time, indeed things soon appear unintelligible. For how could we reasonably expect anyone ever to become bald merely by losing one single hair? However, if not so, we are driven into an equally unpleasant regressus: no hairy person can ever be turned into a bald one, provided the hairs (even all of them!) are plucked from his scull one by one. Clearly, from the moment we involve in slippery-slope or little-by-little arguments like these, something inevitably goes wrong. 2 Before turning to some proposals for unraveling the mystery, I will begin my exposition, in section 2, by carefully formulating the modern version(s) of this puzzle, and making some introductory, as objective as possible, remarks on it. As I will further focus on modern theorizing on the topic, the reader particularly interested in historical aspects, should be recommended to consult other material, e.g. Burnyeat [1983], or the first chapter of Williamson [1994]. During the last twenty-five years, various claims to a logically coherent solution for similar paradoxes have been made. Probably the most simple, yet highly contra-intuitive one is the semantically realist Epistemic View, as heavily defended by Roy Sorensen and Timothy Williamson. This account allows classical propositional or predicate logic (CL and PL respectively) to be entirely retained in the face of vagueness.3 However, to date, the more
*

I am highly indebted to my mentor Jean Paul Van Bendegem for his encouragement and constructive criticism throughout my work on this topic. Also a big thanks to Brick de Bois, for his valuable comments on an earlier version of the article.
1 2

As probably conceived by Eubulides, a Megarian contemporary of Aristotle.

Another popular version of the paradox involves a colour spectrum varying from, e.g., red to orange, in a smooth way, i.e. non-observational with the bare eye.
3

See their [1988a] and [1994] respectively.

popular coherent remedies remain the many-valued approaches known as Supervaluation and Fuzzy Logic. The supervaluational account, which combines classical models with the possibility of higher-order indeterminacy, was initiated by Bas van Fraassen, and applied to vagueness by Fine [1975]. Fuzzy logicians, who surely paved the way for substantial practical, notably technological successes, offer a diagnosis for vagueness by generalizing two-valued truth-functionality ad infinitum. I elaborate on these three strategies in section 3. Less debated coherent solutions, which I will by-pass here, involve the Intuitionistic one put forward by Hilary Putnam in his [1983] and some three-valued paths using Kleene Tables (be it the strong or weak variant).4 Contrasting sharply with these fairly optimistic approaches are some inconstructive ones, such as the logicistic lets get rid of vague concepts (as Frege decidedly advises us) and the nihilistic theres no use in trying to get a firm grasp on vague concepts (most elegantly defended in Dummett [1975]). In section 4, I elaborate on these views and some of their variants. Increasing dissatisfaction with the aforementioned approaches, the constructive as well as the non-constructive ones, has led a number of logicians to consider fundamentally other, notably pragmatical strategies to treat sorites-infected paradoxes. In section 5, Ill be dealing with two of them in some detail: that of Diana Raffman and that of Ruth Manor.5 Their inspiration is, in my view, beautifully exemplified by Mark Sainsbury in his [1990]. The too strict and therefore alledgedly contra-productive adherence to semantics the former treatments exhibit drives him to promote a radically different view on sorites-related issues, and particularly on our classifying ability. He says: What I suggest is that almost all concepts lack boundaries, so that the classical picture is of very little use to us ([1990], p.7, my emphasis). Classical should here be interpreted in a much broader sense than CL/PL-wise, as to include at least all solutions gathered in section 3, and probably more. The most important message seems to me that we should not at any moment forget it is elusive natural language we are dealing with after all. Before reaching a conclusion, a final regular section (6) will be dedicated to the metaphysical dimension of vagueness. It questions what we should impute vagueness to: objective characteristics of the world we live in, or rather language and the way we use it. In other words: is the world itself a fuzzy place (the ontological view) or are we just bound to perceive it being such (the semanto-pragmatical view)? The latter view is to date the more popular. I will particularly highlight the logical debate triggered by Evans [1978], containing a formal proof contra ontological vagueness. A final introductory remark. This article has been written in order to provide with a concise (yet as complete as possible) overview of a vast field of inquiry, and at the same time, it is aimed at a wide audience of philosophers with a logical interest. For both these reasons, I restricted the technicalities to an absolute minimum. As a consequence, the technically more hungry readers shouldnt feel underestimated at any point, but are encouraged to explore further literature, as mentioned in the course of every section.

2. Sorites and Vagueness

4 5

For the latter, see, e.g., Williamson [1994] ( 4.4 and 4.5). These and other pragmatical accounts of vagueness are addressed to more thoroughly in my [200+].

The paradoxical argument called sorites can be formulated in several ways. I distinguish three of them. A first one uses mathematical induction (MI) and universal instantiation (UI) on an ordered domain of objects. The induction can be carried out on the conditions that a predicate P is at least applicable to one member of the domain (call such an object x0), and that, if P is applicable to an xi, then this is also the case for its successor xi+1. Formally, we get: 1. 2. 3. 4. Px0 xi (Pxi Pxi+1) xi (Pxi) Pxn basic premiss (minor) quantified premiss (major) 1,2; MI 3; UI

Hereby, xn stands for any x, so that the paradox is established. In the second form, mathematical induction as an inference pattern is replaced by modus ponens (MP), viz. applied to the first premiss and the instantiation of the second (and so on until xn is reached). In most cases, the formal proof takes some more writing, but it also shows that MI is in fact not needed: 1. 2. 3. 4. 5. 6. 7. Px0 xi (Pxi Pxi+1) Px0 Px1 Px1 Px1 Px2 Px2 Pxn-1 Pxn Pxn basic premiss (minor) quantified premiss (major) 2; UI 1,3; MP 2; UI 4,5; MP .,.; UI .,.; MP

A third and final presentation of the sorites is most straightforward. It opposes two incompatible observations, e.g. that a man with zero hairs is definitely bald (Px0) and a man with a hundred thousand hairs is clearly not (Px100.000). It is not difficult then to conclude that there must be a number i between 0 and 100.000, so that a man with i hairs is not bald, whereas a man with i+1 hairs is: 1. Px0 2. Pxn 3. (xi)(Pxi & Pxi+1) basic premiss final premiss quantified conclusion

How beautiful! We have identified an obvious paradox within the borders of classical logic. But now we may start wondering whether this carefully formulated paradox has anything to do with what ordinary people as well as philosophers usually call vagueness. Lets image a reallife situation to try and find out. Youre in your favourite pub together with a bunch of friends, and being in an experimental mood, as youve all had a few drinks, you agree to mark off the set of bald men in the pub. Unlikely that there will be a consensus among you (problem 1), unlikely that when repeating the task half an hour later (meanwhile some people have left, some have entered, there might be more, or less, customers present, and you had a few more drinks) the pattern of your demarcating behaviour will have developed similarly (problem 2). I will come back to these remarks in a further section. For present purposes I only want to illustrate one simple point, namely that it is very unclear (to me) whether this fictitious experiment has got anything to do with the kind of paradoxes spelled out in the first part of this section. Contrast the pub scene (form 2) with the following one. Youre all by yourself in a

clean lab, observing a single, isolated test-person. This man is being mechanically robbed of his carefully counted hairs one at a time, after each removal of which youre expected to answer the question Is this person bald? by a simple yes or no (form 1). Ill gladly admit there being a certain amount of indeed vagueness between the two forms,6,7 but that could do nothing but strenghten my case, namely that it is very hard, and virtually impossible, to pin down the phenomenon of vagueness. In this respect, it is furthermore peculiar to observe, first, that the literature most often concentrates on predicate expressions, as the central or exemplary cases, but, second and most telling, that it does so by means of examples of an extraordinary variety, making it tough to specify what, if anything, they have in common. 8 And thats exactly the reason why I didnt introduce the topic from the ordinary point of view (form 2), but, on the contrary, chose to go for the most rigorous, i.e. strictly logical, way first (form 1),9 be it that, when done carefully, in a lot of cases the two forms can be (more or less) reduced to one another. The trap I wanted to keep the reader from stepping in, is that, when confronted with real-life examples, one is tempted to start by defining a vague concept (even implicitly) as one lacking sharp boundaries, as clearly there is a grey zone or no mans land between the areas of certain application and non-application respectively. The actual danger is that of taking a theoretical move too soon, e.g. by immediately valuating the law of excluded middle (LEM) and/or the principle of bivalence (PB) as invalid. 10,11 On the other hand, I find this second, non-formal, expression of vagueness far more instructive, and indeed essential, as in the end, it is real-life situations that matter to (most of) us. To catch my point, simply think about the application of age limits in criminal law. There is however a deeper consideration to be made here. After all, one could say, isnt CL, the language in which we discovered the sorites paradox, a logical theory as good (or bad) as any of its alternatives? Well, yes, and no. There are, admittedly, strong reasons to prefer good old two-valued CL. As one simply has to employ a logic, it might as well be kept as coherent and consistent as possible under the circumstances. So why not stick to powerful CL, as long as it works? We start out dividing (well formalized) sentences, within particular theories, in true and false ones. Whenever we run into trouble, theres room for argument whether to change our logical instrument or not, and in the latter case e.g. to add extra-logical, c.q. pragmatical, constraints to our account. It should be noted that the actual dominance of CL has allowed paradoxes like the sorites to pop up in the first place, dragging along long discarded fundamental questions such as Whats validity after all? It is the major, and seemingly endless (though significantly so) task of the discipline called Philosophical Logic to sort out essential matters like these. It therefore should and does investigate the foundations of logic(s), in their relations to other philosophical branches, such as ontology, epistemology, or ethics. In particular, this task is fulfilled in designing and promoting proper formal tools to handle philosophical puzzles such as the semantic paradoxes. There obviously is far from a consensus in doing so, as in fact this entire article may testify. But this, I think, is good news. At the very
6 7

The two problems I just spelled out will have fundamentally transformed, but in no way have disappeared.

More on a most unpleasant manifestation of the vagueness of vague we touch upon here, as causing particular problems for the design of logical tools trying to cope with it, is to be found in 3.4.
8 9

Susan Haack pointed this out, in her [1974] (p.110). Which is anyhow rarely considered necessary across the hundreds of pages of specialized logical literature.

10

Im not acting paranoid here. This tendency is at least reported by Sorensen [1985] (p.134-135), Horwich [1997] (p.929-930) and Schiffer [1998] (p.199).
11

LEM states that the proposition P P is a theorem, PB that for every proposition P: P is true or P is false.

least, it secures us from an unfruitful, and in a sense even dangerous, monopoly of CL. Not any logic should ever be taken for granted (let alone venerated), but its very basics should be constantly open to scrutiny. As Read [1995] in a perhaps somewhat dramatic fashion proclaims: Teaching in philosophy departments across the world exhibits this schizophrenia, in which the dogmatic approach to logic sits uncomfortably side by side with the ceaseless critical examination which is encouraged and demanded in philosophy (p.2). Finally, and closing down this section, I can hardly left unmentioned the warning, echoed by some eminent philosophers,12 not to reduce vagueness to the apparently related phenomena of generality or ambiguity. As far as Im concerned, most (if not all) of the subsequent discussion can be brought back to a terminological matter, so that, provided we define the different terms in play properly, no problem whatsoever should arise. One could for instance use Max Blacks clearcut distinction, made in his [1949]: The former is constituted by the application of a symbol to a multiplicity of objects in the field of reference, the latter by the association of a finite number of alternative meanings having the same phonetic form (p.29).13 Nevertheless, various philosophers have found it necessary to relate or even reduce vagueness to one of these phenomena. Bertrand Russell, for instance, is reputed to have unwillingly confused vagueness with generality.14 For some outstanding further work on this topic, see Sorensen [1989] and [1998].

3. Coherentism

3.1. Epistemic View

No doubt the most conservative reaction to our sorites paradox is to say that, yes, there is a fact of the matter whether a certain, problematic, amount of wheat is a heap or not, but that, alas, we humans are in no position (nor could ever be so) to find out. More specifically, it is due to a fundamental epistemic shortcoming that human access is denied to some crucial missing knowledge about the borderline case. In other words, the Epistemic View (EV) is about ignorance as an essential feature of borderline cases (Williamson [1994], p.201). And once hidden lines are admitted, the argument runs, why should a line between truth and falsity not be one of them (ditto), so that bivalence holds, be it that we cant find out which of both truthvalues applies. Short: CL shouldnt be abandoned when confronted with vagueness, and this is used as a major trump against virtually every rival account. As might be expected (see the previous section), I have my doubts about this policy. Nevertheless, next to this alledgedly comparative advantage, in fact reducing to simplicity, EV-theorists have formulated independent reasons to prefer their solution. I will concentrate on the case for EV as presented in Williamson [1994], canonical as it seems to have become in recent years.15

12 13

See, e.g., Black [1949] (p.29), Quine [1960] (27), Alston [1964] (p.85).

As a simple illustratation, take the word bank. Its ambiguous because, depending on its use, it denotes a piece of furniture, a financial institution or a strip of land alongside a river. Its general because it captures different kinds of furniture and institutions, which can be broken down further. And finally its vague, because in some cases, its unclear whether to consider a particular object as a bank or not, e.g. as opposed to a chair.
14

See his [1988], dating from 1923, and, for some interesting comments, Williamson [1994] (2.4), attacking Russells account, and Hyde [1992], defending it. Ill come back to Russell in section 4. An alternative account is offered in Sorensen [1988a].

15

To start with, let me introduce EVs view on the nature of vague facts, which amounts to a supervenience thesis. It consists, I think, of two crucial components. First, if two possible situations are identical in all precisely specified respects, then they are identical in all vaguely specified respects too. [] There could not be two situations differing vaguely but not precisely (p.202-203). Second, if nature does not draw a line for us, then a line is drawn only if we draw it ourselves, by our use. [] To say that use determines meaning is just to say that meaning supervenes on use (p.206). The central part of the account then, is to demonstrate the failure of the so-called KK-principle, which states that if I know something, then I know that I know it. Williamson uses the notion margin for error to invalidate KK. In the case of the sorites paradox, it involves something like If we know that n grains make a heap, then n-1 grains make a heap ()(p.232). The idea is that if we know, i.e. are sure and aware, that (it is true that) theres a heap of n grains in front of us, then (it is true that) 1 grain less could never magic it away. This margin for error-principle gives rise to inexact knowledge, i.e. the failure of KK, in the following way. In order to produce the sorites argument, we need a sequence of statements of the form (), one containing the antecedent of () for an n, to start up the argument, and MP as an inference rule. But now watch closely. In order to know that n-1 grains make a heap, we should have to know the premises from which we deduced n-1 grains make a heap. [] Thus we should have to know that we know that n grains make a heap. But the previous stage of the argument showed only that we know that n grains make a heap. With the KK-principle, a sorites paradox would indeed be forthcoming. Without it, one iteration of knowledge is lost at each stage of the argument (p.233, my emphasises). As a result, the sorites paradox melts away before our very eyes. The source of this inexactness of knowledge is to be found precisely in the supervenience of meaning on use, introduced at the beginning of this paragraph. Small changes in the use of words not ruled by precise divisions (such as vague ones) can imperceptibly shift their meanings, so at the same time the truth-value of statements containing them. The sharp boundary between rival vague concepts entirely depends on the unstable, non-graspable dispositions of speakers. In other words: a man might be (borderline) bald, but that very same man might as well have been (borderline) non-bald. This account is often considered as not touching the very core of vagueness. According to Horwich [1997], it is far-fetched to label the obvious(?) knowledge problem as an external failure, viz. to realize, in common language, compatibility with the alledged objective meaning of vague terms. He himself sticks to an internal paralysis of judgement for a diagnosis.16 EV explains our problematic phenomenon in a most simple, i.e. economic, fashion. True, but that does not warrant the explaining away of it, making use of supervenience.17 Both Schiffer [1997] and Burgess [1998] scrutinize this point. Briefly, their message is one of disbelief in the essence of ignorance as presented by Williamson. Why should the lack of an epistemic procedure to discover the connection between semantic and underlying fact, exclude this knowledge in principle?

3.2. Supervaluation

The logical technique called Supervaluation (SV) tries to reconcile two apparently conflicting opinions when confronted with a typical sorites series as introduced in section 2: that there admittedly is no one transition point which is valuable for each and every observer (1), but that at the same time, for each and every observer there surely is a transition point (2). In
16

The constructive part of his account, which somewhat disappointingly reduces to a version of Max Blacks (see below 3.3), I will leave unmentioned.
17

I refer to my warning concerning CL, somewhere near the end of the previous section.

reconciling these positions, SV will preserve CL on the object-level (accomodating 2), while allowing dissensus on an extra meta-level (accomodating 1). The following simple technical presentation will prove SV to be a transparent and user-friendly way to deal with the soritesparadox. We start out with a CL-model M for all well-formed formulas A of a language L, so that vM(A)=0 or vM(A)=1, A being false or true respectively. Taking M to be a set of models like M, a SV can be defined as follows: - SVM(A)=1 iff for all M M: vM(A)=1 - SVM(A)=0 iff for all M M: vM(A)=0 - SVM(A)=n in all other cases, n standing for neutral The idea behind all this should be intuitively clear. When confronted with a vague expression, different speakers judge it to be either true or false (they are restricted to the level of their particular two-valued M). The supervaluationist gathers their models (called sharpenings or precisifications of the vague expression) in an M. In case there appears to be a consensus among these speakers, the expression at hand is called supertrue or superfalse on this SVlevel, depending they all have judged it to be true or false. In case of a dissensus, the expression is awarded the status undecided or neutral SV-wise. Note that every participant to the natural (object) language is forced to allocate a problematic (i.e. vague) expression one of two truth-values, notably these of CL. The ultimate truth or falsity as it were, of expressions like these, can only be decided (or left undecided) at a meta-level unattainable for these ordinary speakers. What to think of all this? To be honest, from my first confrontation with SV, it has made me feel uncomfortable even more than EV does. You see, it seems to amount to a neat trick, trying to handle vagueness in a surveyable way, to add a level of judgement which is completely foreign to common talk. In doing so, undoubtedly, SV clusters the advantages of a number of accounts. Thats nice, were it not that it also shares the drawbacks with them. Let me elaborate. In SV, one is obliged to determine a boundary between the vague concept and its equally vague opposite. On the basic level, that is, the only area an ordinary speaker is allowed to hang about. Clearly, this meets EVs wishes, as well as those of any others wanting to retain CL. To know: all inference rules remain unaffected,18 no words (such as definitely) are added to the object language, and no theorem whatsoever is banished. This is possible only because a level on top of the usual meta-language is introduced, the SV-level, where PB no longer holds. The logic turns three-valued on this level: a statement is supertrue, superfalse, or else remains undecided. This stage of the analysis is supposed to satisfy another group of theorists, viz. those prefering logic with several thruth-values, from three up to an infinite amount of them (see next paragraph), as well as those defending fundamental under- or overdetermination of formal models compared to the natural language theyre trying to capture (see further sections). If youre wondering why all this is so, just note that the value undecided is not interpreted here. But now for the drawbacks, throwing a bit more light on my resentment with SV. First of all, there is the somewhat strange role of the quantifier , as compared to the phrase there exists. It is unclear what its truth-conditions are, if, on the one hand, (xi)(Pxi Pxi+1) holds, but, on the other, (xi)(Pxi & Pxi+1) doesnt.19 This can be reformulated as follows: SV creates itself an extra metalanguage, in which LEM is retained, while PB is rejected.11 A severe difficulty arises when asked to justify this move independently, i.e. other then as an ad hoc one. Second,
18 19

As theyre to operate only on the basic level, in the end they all reduce to the classical ones. This point was made in Hyde [1997].

viewed the fact that PB is not done away with in the different models M, it is still assumed that a (every) vague concept is, in principle, sharpenable, i.e. eliminable. The question now amounts to what an admittable sharpening is. Yet thats nothing but the initial problem all over again, so what has been gained? As in the case of EV, from the perspective of daily practice, the nature of vagueness (whatever it may be, and without doubt thats itself a vague matter) seems thoroughly violated whenever a cut-off point is accepted, be it on an underlying level. More on all this, as concerning the very essence of vagueness, quite naturally is to be found in the sections to follow. In my view, the next few lines from Roy Sorensen sum up this discussion, and at the same time take it to another level: Deviant logic was recruited because it is supposed to be better suited to genuinely vague predicates. But now we find that these purported predicates have been expurgated by the meta-language used by the deviant logician. The very process of setting up the apparatus needed to recognize genuine vagueness precludes it (Sorensen [1996], p.207). The way I understand it, in the models M, vagueness is eliminated, while in the model M, it is reintroduced as a kind of disagreement concerning application of the concepts in question. But difference of opinion can not be what vagueness is all about. It is just as much an individual problem, over time and place, but even when confronted with a single, isolated sorites-like case. Sorensen goes on with a suggestion sometimes met (which he doesnt support): At this point, a deviant logician [] might agree that any logic of indeterminacy must be a meta-logic of indeterminacy (p.209). It is indeed a vague meta-language that is refered to here. Is this idea intelligible? Or do matters tend to get trivialized, as in this inspiration, nothing less than a natural language seems to be acceptable as an adequate model of natural language, to paraphrase the later Wittgenstein. 20

3.3. Fuzzy Logic

To make clear the idea behind Fuzzy Logic (FL), I will go back to a famous philosopher, who is said to be one of the grandfathers of this alternative logical school. 21 This philosopher was called Max Black, and in his [1949], first published in 1937, he set out on an early attempt to try and capture the phenomenon going under the name of vagueness. Black was looking for a way to map the defective correlation between the ideal world of theoretical objects and empirical reality, to indicate an appropriate symbolism for vagueness by means of which deviations from a standard can be absorbed by a reinterpretation of the same standard (p.27). It was Blacks ambition actually to measure degrees of terms vagueness, compiling so-called consistency profiles of their use, on the basis of linguistic input delivered by a representative group of observers. What this kind of empirical analysis was supposed to circumvent was the need to locate, in a sorites series, one or more cut-off points, in the case of a two-valued logic and that of a three-valued logic (demarcating a region of uncertain application; as, e.g., in SV) respectively. These crude distinctions indeed become unnecessary, once one admits differentiation to develop smoothly. This, in Blacks eyes, could be achieved taking into account that the uncertainty is a matter of degree, varying quantitatively, though not regularly, with the position of an object in the series (p.32). The practical, viz. statistical, problems that accompany an enterprise like this might seem, even at first sight, to be insurmountable.22 Nevertheless, during the 60s and 70s, grandfather
20 21 22

Williamson [1994] (p.84). See, e.g., Goguen [1969] (p.326-327), Williamson [1994] (p.103), and Trillas [1996] (p.91). For a tentative overview, see, e.g., Burns [1991] (pp.34-36).

Black, along with some others, apparently inspired a group of people, mostly (this is still so) mathematicians and computer scientists, to pursue and elaborate his initial idea. The person who grew up to be the father of FL, turned out to be Lofti Zadeh (or so is said). Now what distinguishes the efforts from this era on, is the explicitly formal ambitions. FL aims to extend ordinary deductive methods to situations in which the information available may be only partly or approximately true. In FL, each statement has some numerical degree of truth or falsity (Copeland [1997], p.517, my emphasis). What I shall do in the sequel of this paragraph, is to give a very rudimentary sketch of what FL standardly looks like, show how it can be applied to vagueness, viz. the sorites, and formulate some objections. Doing so, Ill further rely on secondary sources as the one just quoted. Examination of the best known basic litterature, such as Goguen [1969], Zadeh [1975], Machina [1972] and [1976], or Sanford [1975] and [1979], would require too much of the space available, and is kindly left to the reader. I should warn that all of this commendable work is generally quite hermetic (perhaps less so in the case of Machina and Sanford), or at least thats my appreciation of things, as theres very little philosophical preoccupation to be noted.23 In FL, the multitude of truth-values between falsity and truth, i.e. in the interval [0,1], is directed by a number of semantical rules, which, in their most simple format, look something like this: For the four classical connectives: FL1 (p) = 1 (p); FL2 (pq) = max [(p),(q)]; FL3 (p&q) = min [(p),(q)]; FL4 (pq) = min [1,1(p)+(q)]; Once modal quantifiers are added, the following clauses could e.g. be added: FL5 ((x)p) = glbx(v(p)); which stands for the greatest lower bound, viz. of all (p)s only differing in value due to variable x, freely occurring in p FL6 ((x)p) = lubx(v(p)); which stands for the least upper bound, viz. of all (p)s only differing in value due to variable x, freely occurring in p Using this framework, one can actually force any sorites paradox to melt away (Copeland op.cit., p.520). The quantified premiss indentified in section 2 is judged not entirely true, be it true enough in most of our daily practice, where we very rarely run through an entire sorites series. When forced to, what FL does, is to rob the statement under consideration of a tiny fraction of its truth, every time MP is applied. That this subtle mechanism uncovered by FL remains however unnoticed by the ordinary speaker, explains why the latter gets embarrassed in the course of a typical sorites argument. Intuitively acceptable as this all may sound, one runs in serious difficulties trying to justify any attempt in writing out this account formally. To begin with, how to interpret every time MP is applied? In other words, how many truth-values (i.e. how big a single fixed or variable? jump in value) should we take into consideration, and not to forget: why?24 It seems almost inescapable that this will depend on the case at hand. But this opens the route to an endless debate, in which
23

No discussion whatsoever on the nature of philosophy is being elicited here. Roughly, what I mean is that the authors write from a practical, as opposod to theoretical (also in the sense of reflecting), point of view. And that, I think, is the major reason for their tending to overlook the crucial shortcomings in their account so obvious to most philosophers (certainly the professional logicians among them), which are, first, that the possible interpretations of the formal framework are not sufficiently screened or justified, and second, most important, that the philosophical problem raised by the sorites paradox is not addressed after all, but on the contrary aggravated. More details to follow.
24

It is very difficult to argue why certain parameters should be used. [] It is also difficult to argue what a particular membership degree [e.g. to the set of bald men, bvk] for a certain object really means (Hellendoorn [1990], p.19).

the most various deciding factors, which should be processed empirically, are being put forward. Apart from the danger of loosing ourselves in it, in stead of concentrating on a suitable semantics, at this particular point, the account drifts away from the very question at hand. More particularly, the practical-descriptive preoccupation of most fuzzy logicians, which has been very fruitful indeed,25 makes us lose contact with what the sorites paradox, for a philosopher, is or should be all about: how to justify a certain threshold rather than another or a multiplicity of them, and how to justify a threshold at all. In founding a FL, it is looking for easy answers trying to reduce a scale of truth-values to the alledged average of use in a certain population (as Blacks original proposal was). Writes Sainsbury: To what in our actual use of language does this division correspond? [] We do not know, cannot know, and do not need to know these supposed boundaries to use language correctly ([1990], p.12). What is in fact being questioned here (see section 5 for more details), is the point in trying to map linguistic concepts (the propositions in which they appear) and truthvalues in a rigourous way, the sense of determining boundaries after all. This is a problematic affair, for FL as for every alternative constructive account. Indeed, the criticism can as easily be directed towards two- or three-valued proposals, but evidently is to be radicalized in the case of FL, as it does so itself with the notion of truth-value. In this perspective, one could even say FL is nothing but an extreme version of CL.26 According to Michael Tye, in his [1994b], the result is a commitment to precise dividing lines that is not only unbelievable but also thoroughly contrary to [] vagueness [as we conceive of it in connection with everyday examples] (p.11).

3.4. Higher-Order Vagueness

Facing the accounts presented in both the preceding two points, is a problem that can hardly be left unmentioned, cause its particularly nasty, and therefore indeed deserves a (brief) paragraph of its own: the phenomenon named Higher-Order Vagueness (HOV). In the previous subsection, we learned that FL (as any multi-valued logic) runs into difficulties trying to characterize truth-functionality, i.e. determining the proper number of values and the way they should relate to one another. Now, suppose we achieve to do this. Then still, a far more serious obstacle remains: that of the attribution of the different truth-values to statements, i.e. that of linking the vague object-language on the one hand, and the precise meta-language on the other. In any multi-valued logic, one has at least three options when asked to validate a formula, for either it is true, false, or something in between these two, where the something in between gets filled in depending of the number of intermediate values available (from one to an infinite amount, i.e. a continuum, of them). At this point, the objection raised against the CL-account by the sorites paradox itself, can be generalized towards multi-valued ones. In the words of Williamson: As grain is piled on grain, we cannot identify a precise point at which That is a heap switches from false to true. We are equally unable to identify two precise points, one for a switch from false to neutral, the other for a switch from neutral to true. If two values are not enough, three are not enough (op.cit., p.111). And so on: for the very same reason, four, five, six or more truth-values wont suffice. Fuzzy logics might indeed be able, to a large extent, to circumvent or ease out obstacles met in real-life applications, reaching from washing machines to expert systems
25 26

Zadeh/Kacprzyk [1991] offers a nice view of this branch of research and development.

I owe this point to Jean Paul Van Bendegem, who in his turn pointed to the late Leo Apostel as his inspiration. See also the problem of higher-order vagueness set out in 3.4.

10

providing with medical diagnosis. These uncontested practical successes do not, however, answer the philosophical question, which gets duplicated at every threshold: why should we switch from one answer to another at this very point? On the contrary, we are dragged ever more deeper in trouble, as long as more truth-values are being added to the analysis. Eventually, FL leaves us philosophers with an infinity of these arbitrary thresholds to justify. Nothing is gained. The regress into which we are driven only highlights the problem which gave rise to the initial problem (Burgess [1990b], p.418).27

4. Incoherentism

Solutions in the previous section recommended an alternative interpretation of the quantified premiss (3.1), an alternative inference-tool (3.3), or something in between these two (3.2). The current section will concentrate on those who believe the sorites argument as a whole is of an irreducible incoherence, i.e. that it forms a genuine paradox. This means that, in their view, a coherent logical interpretation as in section 3 is in principle impossible. Part of them, viz. a number of rather traditional(ist) logicians, propagate the radical banishment of vague components (i.e. the incoherence) out of natural languages realm, while another group devotes itself to the joyful acceptance of this incoherence. The latter school can be further subdivided. Some think that, matters being incoherent as they are, the quest for a sensible semantics should be abandoned in the face of vagueness. Others (i.e. few) opt for paraconsistent models. Still others aspire to develop approaches much richer and complicated than the logical ones in the strict sense, i.e. having the field of examination contain pragmatical criteria next to the well known semantical ones. The latter tendency will be dealt with extensively in the next section, the others right here. Lets start out with the logicians whom I somewhat scornfully called traditional(ist). They are being exemplified by Gottlob Frege and Bertrand Russell. Where the formers attitude towards vague language will strike the reader as pretty straightforward, the latters is rather dubious. Frege leaves no doubt whether vagueness should be taken seriously or not, as every symbol in his newly developed formal language can have nothing but a stipulative meaning. In this Begriffsschrift-logic of 1879, a complete and consistent set of definitions and recursive rules saveguard the logically perfect language against the creeping in of semantically defective propositions, such as the ones containing vague terms. Every formula can have but one referent or truth-value. Clear enough that CL fails for sorites paradoxes: because vague predicates generate borderline cases, LEM would end up under unbearable pressure. This leaves us, according to Frege, with no choice but to eradicate vagueness. Susan Haack captured the essence of a similar program quite nicely: The recalcitrant sentences, those the assignment of true or false to which is thought to give rise to difficulty, do not make statements, or, do not express propositions ([1974], p.50). And thats it. Why bother precisifying vague concepts for instance, if in doing so, the meaning of the terms involved unavoidably gets altered? Frege never even considered a systematic study of the matter and only mentioned it briefly. 28 Russell, on the contrary, devoted an entire article to it (the reflection of a lecture held at the Jowett Society in 1922). The first of its two major conclusions comes down to Freges: CL, i.c.
27

A lot more could be said here, but I will kindly leave it to the reader to discover. There is, e.g., an interesting discussion brought about by Sorensen [1985] (viz. Deas [1989], Burgess [1990b], Hyde [1994], Tye [1994a]), during the study of which I couldnt help wondering whether HOV really reduces to vagueness of vague (this remains an open question to me). Also notice that section 5 will open an alternative route for dissolving HOV.
28

For instance, see indeed his famous Begriffsschrift (e.g., in Frege [1971], p.62).

11

LEM, doesnt hold for vague language. The other part of the message brought by Russell however, is not that vague terms should be banned, for that would leave us with empty hands. The fact is that all words are attributable without doubt over a certain area, but become questionable within a penumbra, outside which they are again certainly not attributable (Russell [1988], p.149, my emphasis). When saying all words, Russell means all words, including the logical ones, such as connectives and truth-values.29 The result of this is even more drastic than in Freges case: All traditional logic habitually assumes that precise symbols are being employed. It is therefore not applicable to this terrestrial life, but only to an imagined celestial existence (op. cit., p.151). Thats the story of earthly life versus logical heaven in a nutshell. Williamson comments that, as Frege, Russell never meant to seek some non-traditional system of logic better adapted to vagueness, for no logic worth the name is reliable when applied to vague symbols ([1994], p.59). Nevertheless, both Freges and Russells solution come down to good old CL, getting rid of vague concepts all together (see, in fact, 3.1). The argument is that logic is an ideal realm, aimed at, yet finally unattainable for poor humans, who as a consequence have but one option: to live their miserable lives infected with such terrible diseases as vagueness. Once logic gets in, vagueness gets thrown out. It is, however, left extremely mysterious how we should be able to conceive a perfectly precise CL-like language, when at the same time, all of our words are vague. In the (substantial) rest of this section, I shall focus attention on the theorists ready to embrace incoherent vague language. They are, perhaps somewhat surprisingly, called nihilists, for holding that inconsistency is a prevalent feature of natural language (Burns [1991], p.28). How is that? Well, if the paradox, i.e. the reasoning involved, is accepted as genuine, clearly no cut-off point can ever be admitted. Easy enough to see, that either both or none of basic premiss and conclusion then are to be accepted. Any which of these two options we take, we are left with contradiction. The nihilistic attitude consists, of course, in embracing the latter, rather than withdrawing in the face of it. Formally, this can be done using, or even devising, a paraconsistent logic. In fact, this option is all too often brushed aside without any serious consideration. Reviewing Williamsons [1994], Dominic Hyde notices: The broad scope of the historical survey, extending to a comprehensive bibliography, is marred only by the omission of dialectical or paraconsistent analyses ([1995], p.925). This is particularly strange, he adds, for our daily practice more often seems to amount to overdetermination rather than underdetermination.30 Just imagine, or rather realize, how many times questions like Is it raining?, Is he bald? or Are you tired? tend to get answered by a Yes and no. 31 According to Sorensen [1985], the most straightforward acceptance lf the incoherence thesis appears in the work of Peter Unger anj Samuel Wheeler (p.135). Indeed, after inspection of their [1979] and [1875] respectively, both his and Williamsons [1994] characterisation of these authors, viz. as global nihilists, prove to be incontestable. The standpoint they promote is most radical: it doesnt allow any sorites-like argumeot ever to start up, for it denies the legitimacy of its basic premiss. Let me illustrate. Assuming ths paradox to be a genuine one (A1), we start out with a definite heap of grains (A2).32 Now we begin removing the grains one by one. For it has been rgreed that the argument is acceptable (A1), , three, two, one, and even zero grains will make a heap. The latter assertion clearly amounts to absurdity, but rather than rejecting (A1), as is
29 30 31

Some criticism on this is to be found, e.g., in Sorensen [1988b] (p.268) and Williamson [1994] (p.55-56). The latter of which is remedied, for instance, by the popular SV-method explained in 3.2.

For some introductory literature on paraconsistency, see Priest et al. [1989]. For an interesting paraconsistent variant of paracomplete SV (3.2), called subvaluation, see Hyde [1997], and, particularly, Varzi [199+].
32

Say a million of them. If youre not satisfied, add more grains until you agree the configuration forms a heap.

12

done elsewhere, global nihilists reject assumption (A2), and hold that a million, or whatever numser of, grains do not form a heap. Not unsurprisingly, the title of Ungers [1979] said there are no ordinary things.33 Any other response, he notes, either directly forces us to accupt a metaphysical and/or conceptual miracle, as the existence of a principal cut-off pnint in a gradual sorites-process is called, or indirectly does so, as in the case of multi-valued logics, no rational counter is provided with, for the miracles just mentioned remain necessary, be it on a higher level (see 3.4). Undoubtedly, with his [1975], Michael Dummett has delivered one of the most impressive philosophical writings on our topic. Nevertheless, it is primarily concerned with something else, viz. strict finitism as a philosophy of mathematics, an item which is combined with the status of observational predicates and phenomenal qualities. The articles conclusion is the declaration of both finitisms and observationals incoherence. I will here quite naturally concentrate on the latter. Throughout, Dummett reflects on variants of the sorites, e.g. (Hao) Wangs Paradox (formulated in the domain of natural numbers), 34 or little-by-little arguments exploiting the non-transitivity of indistinguishability (e.g. try the minute hand of your watch). The heart of the paradox, for Dummett, is that either we shall have to say that a contradictory state of affairs may appear to obtain, or we shall have to say that, from It appears to be the case that A and It appears to be the case that B, it is illicit to infer It appears to be the case that A and B ([1975], p.321). Any which way, it should be clear, for him, that a workable logic for vagueness is illusory, since the use of vague language is intrinsically incoherent, and the paradoxes involved are insoluble. This is a confirmation of what Frege said, but Dummett goes on to appreciate vagueness as an essential feature of any natural language that truly functions, instead of wanting to eradicate this problematical quality. Vague fragments nonwithstanding, we speakers usually understand ourselves and one another, at least to a large enough extent so as to accomodate daily life. In other words, most of the times, language works. If we would demand, on top of that, mutual (or indeed even individual) consistency in the meanings of the vocabulary we share, all linguistic efficiency would desperately be lost, because awaiting the impossible, (internal) communication would inevitably come to a hold. A final philosopher of logic and language I cant leave unmentioned here, for having on his curriculum some of the key texts on the topic of my article: Crispin Wright. Moreoer, treating him at the end of this section, is not without significance, for in his writings, an important evolution takes place. It is this movement, viz. towards a non-propositional account, that Ill try to highlight now, stripped of all possible detail. What Wright has been increasingly criticizing, in fact, is the principle of rule-boundedness in linguistics. Even admitted its actually there, he further wondered how we could ever, and why we would ever want to, penetrate to the understanding of any such rule, which speakers implicitly (have to) follow, no matter what. His initial position towards vague concepts is that of Dummett: the linguistic items in question lead to semantical incoherentism, not because of our intellectual laziness (Frege), but inherently so: The utility and point of the classifications expressed by many vague predicates would be frustrated if we supplied them with sharp boundaries. [] It is not generally a matter simply of lacking an instruction where to draw the line; rather the instructions we already have determine that the line is not to be drawn (Wright [1975], p.330). Wright goes on to confirm the large actual success of the language-game, as he calls it (after later Wittgenstein), but his eventual diagnosis, fully elaborated in his [1987], deviates from, or rather
33

Not just traditional sorites-prone concepts as scalps or heaps are subject to this law. As there are no ordinary things whatsoever, every object we name, e.g. a stone, lacks an existential status (i.e. genuine nominalism). In short, the paradox states that every number n is small, since 0 is small, and if n is small, then so is n+1.

34

13

goes beyond, Dummetts. In fact, according to him, the latter does not offer one after all, as he is unable or unwilling to identify a fallacy in our inconsistent, yet functional fabric called natural language. For Wright on the other hand, it is the very idea of incoherently codified practice that is in itself incoherent: If the rules for the use of red really do sanction the paradox, why do we have absolutely no sense of disturbance, no sense that a real case has been made for the inferential ingredient at all? Are we so abjectly irrational that we cannot recognize our confusion even when it is completely explicit? A different account is called for.35 His spinned out [1987] is an attempt to get a grip on such an account, which has to be non-proposional by nature. This is not to say that the notion of meaning becomes useless. The account will concern non-cognitive, yet meaningful knowledge, like the one constituting practical skills, which we just seem to be able to perform, but couldnt ever (want to) describe in full detail. We learn them by doing, in the course of an ever-lasting ostensive training, so that their meaning is always under construction. Note that, together with Wright, we have already stepped with one foot in the next section.36

5. The Pragmatical Turn

So finally now for the radical departure from all traditional semantico-logical theories of language brought together in the previous two sections, theories which consider the linguistic vehicles of a concept as having a meaning which fixes its extension, the set of things of which it is true. Following Mark Sainsbury however, no sharp cut-off to the shadow of vagueness is marked in our linguistic practice, so to attribute it to the predicate is to misdescribe it ([1990], p.11). In the course of section 2, we have been confronted with one particularly tough problem over and over again: that of HOV (2.4). No wonder, says Sainsbury, for you do not improve a bad idea by iterating it (ibidem), and thats exactly what multi-valued semantics try to do. Hence, the argument in favour of pragmatical accounts, concerned with language-users rather than the world they talk about, will be the absolute uselessness of rigid linguistic boundaries, inextricably bound up with the apparent unwillingness and/or inability of speakers to draw them. Doing pragmatics, contextual constraints, whether internal, as in the case of Diana Raffman, or external, as in the case of Ruth Manor, will prove to be extremely important.37 According to Raffman, an adequate treatment of vague predicates and their sorites puzzles must appeal to the character of our judgments about the items in the series ([1994], p.44). Lets regain the red-orange continuum of colour patches to illustrate. It is easily agreed upon that any subject has to (and will) report a shift in colour sometime, be it sooner or later, when running through the series, in order to avoid the paradoxical conclusion. But does this not inevitably mean then, that two particular adjacent patches have to be reported as having a different colour? Raffman thinks not, as neighbouring patches can never be in different categories when judged pairwise, only when judged singly. To be able to explain this, the traditional (cartesian) picture of the integrated mind has to be challenged, for at least two different mental personae are said to be at the basis of any observational decision: one categorizing, i.e. vertically comparing with a (possibly imaginary) model, another discriminating, i.e. horizontally comparing two objects. So, the latter
35 36

Cited from the abridged reprint of Wright [1987] in Keefe/Smith [1996] (p.213).

In fact, Wright also seems to have inspired others to look for a more behaviouristic semantics, as Raffman confides in her [1994] (p.43). See the sequel.
37

Some further pragmatically inspired accounts can be found in the writings of Hans Kamp (viz. his [1981], preceded by [1975]), Linda Burns (viz. her [1991] and [1994]), Graham Priest (viz. his [1998], prepared in [1991]), and Terry Horgan (viz. his [1994] and [1998]). In my [200+], I additionaly elaborated on the latter two.

14

only comes into play from the moment the patches are (to be) considered pairwise. Finding a pair to be marginally different, he will constrain his colleague to categorize them identically: Categorize them as you like, hell say, but categorize them together (op.cit., p.47). A suitable diagnosis for sorites is straightforward: if judgments alongside the series are made pairwise, a shift in colour cannot happen but pairwise. Hence, it becomes possible that two adjacent patches x5 and x6 are said to be red, while just after (or before) that, the patches x6 and x7 jointly receive(d) the label orange. And theres more to it. A phenomenon called judgmental inertia prevents categorical shifts from being local. As long as pairs of patches are judged as red, the polar attraction of the original red reference will remain fairly strong. Thats the reason why, once a shift to orange has occured, it will often be observed that the brand new, orange reference spreads its force way back in the series, so that red pairs now have become orange ones. This does not however, at any moment, imply that couples of patches are seen as red and orange at the same time. Perception is episodic, and it is precisely because of this kind of shifts that no discontinuities whatsoever are perceived. In this respect, also, and finally, notice that the presentation of the colour-spectrum can turn out to be extremely important. Just think about the force of the initial red reference when all of the patches (also the clear orange ones at the extreme of the continuum we are moving towards) are disposed during the entire experiment, as compared to its force when only a couple of patches are uncovered at a time. The account certainly would deserve to be dealt with more thoroughly, for its foundations are far more elaborated than even suggested here, but that would really take me too far, so I have to leave it at that.38 In the interpretation of Ruth Manor, who in her turn wishes to challenge the sharp distinction between semantics and pragmatics, the relevant contexts are not internal or psychological, as Raffmans, but rather external or physical. More precisely, it is the domain of objects to which properties are (not) predicated, that, in her view, plays the crucial role in the varying meanings vague sentences exhibit. As Manor, in her [1997], also proposes a formal framework that does justice to this particular contextual constraint, I will set it out, meanwhile allowing the underlying idea gradually to become clear. One of the basic ideas is that vague predicates may denote so-called foggy objects, whose parts cannot unequivocally be determined. In order to formalize them, objects d are represented by sets of (alternative) sets of atoms belonging to the domain A: d . (A). As a set of subsets of A, any object d is characterized as a set of possible delineations. It will then be called a distinct object dd, if it has only one such a delineation (dd = {s}, with s A), or a foggy object df, if it has at least two (df = {s1,s2,}, with s1,s2, A). The opposite #d of an object d relative to A is defined as #d ={A\s with s d} (A\s reads: the complement of s relative to A). An object d1 is then a boundary case of an object d2, iff every member of the former is a subset of some member(s) of the latter and a subset of some member(s) of its complement #d2. So the boundary case d1 for an object d2 is partly enclosed in that object d2 (more exactly: some of the delineations of d1 are enclosed in delineations of d2), and partly enclosed in its complement #d2. I have to skip some other notions here, such as parthood, n-ary objects, universal vs. resticted domains, as well as the formal model Manor builds on top of all this. Nevertheless, we are already in a position to get an idea of how a similar model can be used in dissolving sorites-paradoxes. Consider the domain of people all over the world. In it, for every n, there is at least one person with n hairs on his scull. Lets therefore call Hn the non-empty extension of has n hairs. Now consider an object p = {H0, H0H1, , H0H1H2... Hn, }, the elements of which are the sets of people with less than n hairs. Obviously, one can define its complement #p. In this
38

See, however, the [1994] article refered to and cited here, as well as Raffman [1996].

15

particular domain, p and #p then denote the bald and non-bald people respectively. One can verify that theyre both vague (or foggy) in nature, for counting different subsets of the domain as members (as the delineation of the n-haired clearly cant be unequivocal). But more importantly, every set of people with a particular number of hairs, e.g. {Hn}, neither belongs to p nor to #p. This means that the corresponding object {{Hn}} can never belong to the extension of is bald or is non-bald, or, put differently, that it is always a boundary case of these two. The contextual account proves to be fruitful, for it depends on how the domain gets divided, i.e. by determining n (from occasion to occasion), whether the group of n-haired will count as bald (together with the rest of the bald) or non-bald (together with the rest of the non-bald), while on their own they cant be but borderline. 6. On the Origin of Vagueness

Gareth Evans, in a mere one page article, provided with a formal proof that met an enormous amount of fierce criticism in the twenty years that have passed since its appearance. However, I am disappointed (yet amazed), not about the quantity, but about the arguments that have given those who reacted and contra-reacted to this [1978]. As I see it, the article couldnt have done other than arouse suspicion, for one is claiming to decide on an objective characteristic of the world in a strictly logical way. Since logic clearly belongs to humans view upon this world (more particularly, the part of it expressed or expressible in language), Evans should have explained why his proof nevertheless is able to go beyond logic. He never anticipated this, nor did anyone ever after. But more and most surprisingly: nobody ever bothered asking or wondering. I will set this case aside for the time being, to present the famous proof and some of its major criticisms, but only after having quoted some inspiring words to be found in Lejewski [1976]: The ontologist is anxious to justify, with the aid of informal arguments, his existential answers to the fundamental question: what kinds of entities are there? The logician is not concerned with these arguments. The theories he develops do not imply, in the logical sense of imply, the existence of anything or should not imply the existence of anything, if he cares for logical purity (p.28, my emphasis). Ill come back to it at the end of this section. Now for Evans proof, with first some preparations: a and b are names for objects, a=b is an identity statement with undecided truth value, and the operator expresses this uncertainty. The actual proof then, is as follows. (1) (a=b) (2) x [(x=a)]b (3) ~(a=a) (4) ~x [(x=a)]a (5) ~(a=b) --------------(5) ~(a=b) Explanation. Line (1) expresses the uncertainty of the statement a=b (the object with the name a is the same as that with the name b). 39 What happens in line (2) is called intensional or -abstraction: object b has the property of being thus that it is uncertain whether it is identical to object a. This is a major step, and one who doesnt accept it, can halt the proof here, before it has actually started, as does Lowe [1994], saying there can be no such thing as the property an
39

Henceforth, I will abbreviate the object with the name a/b to object a/b or even a/b simpliciter.

16

object has just in case there is no objective fact of the matter as to whether or not it is identical with the object a (p.112). Line (3) seems to be as evidently clear as true: it is not uncertain whether object a is identical to object a. But he who defends objective vagueness would simply be giving in accepting it, for this particular statement resides at the core of the ongoing discussion: are there vague, throughout this proof interpreted as non self-identical, objects? Somewhere near the end of this section, I will refer to Quantum Mechanics (QM) as a positive line on this ontic vagueness or non-identity. I have qualms about reducing vagueness and identity to one another, but explaining this in detail would lead us too far here.40 Anyway, one can wonder if Evans is not begging the question by inserting (3). Lets move on to line (4) then, which reformulates (3) in terms of -abstraction. In the quoted article, Lowe goes on scrutinizing this logical technique, claiming that the symmetrical properties x [(x=a)] and x [(x=b)] are in fact one and the same, as they only differ by permutation of a and b. As a consequence, and even if (2) were true, (4) could never be so: the property expresses uncertain self-identity, and object a is equally submitted to it as b is. Line (5) concludes the proof, by applying CPOS41 of Leibniz Law (LL)42: as (2) attributes a property to b which (4) denies to a, both objects are non-identical (so not of uncertain mutual identity, ergo non-vague compared to one another). In (5), Evans strenghtens the former conclusion: a and b definitely are distinct objects, provided lines (1) to (4) can be prefixed by a definitely-operator . This whole operation presupposes and its freshly introduced dual to be strong enough to generate a modal logic as strong as S5, Evans confines us. But as Johnson [1989] points out, this is far too strong a claim: Evanss sentential operator , representing objective indeterminacy is in fact not the dual of that symbol which Evans defines as its dual, , or else the modal logic Evans proposes to generate obviously is not nearly as strong as S5, contrary to his indications (p.104). We turn to Michael Dummett, who explains: If it is possible to give a coherent account of this matter, then the result will be in effect a modal logic weaker than S4, in which each reiteration of the modal operator Definitely yields a strenghtened statement ([1975], p.311). Indeed, the characteristic S4-axiom pp would be disastrous for any vagueness-logic. Besides, what about the following S5-axioms: pp and pp, the first of which trivializes vagueness (as every statement is definitely indefinite), while the latter neglects any possibility of HOV (see 3.4).43 Thus far, I have only considered the technical (non)merits of the proof. Lets now focus on its philosophical interpretation, as have done numerous commentators. The most important line of criticism is the one questioning reference in Evans proof. Ill try to summarize the main points of it. We have seen that statements of the form p can be made true or false by sharpening or precisifying them (in a SV; see 3.2). When doing so for object a in (3) and (4), notice that both expressions are equivalent only in case of exact reference. But then we have to suppose what were trying to prove, which means that inference (3)-(4) is invalid, both Thomason [1982] and Noonan [1982] conclude. The latter specifies that an a with multiple sharpenings leaves us with an expression (4) uncertain in truth-value, and an a without any
40 41 42

See note 46 for some references. From a b classically to infer b a.

(a=b) (Ea Eb) for all predicates or properties E. Remarkably, and despite its crucial role in this very proof, Evans himself never bothered formalizing it (Johnsen [1989], p.104). On Evans careless use of operators, see also Over [1989].

43

17

possible sharpening even with a false expression (4).44 In his [1990], Noonan puts these remarks in some philosophical context, reminding us of the quest for vague objects this discussion is all about, and Evans failure to demonstrate any necessary incoherence in the idea of their very existence. However, he adds that if whenever there is denotation of vague objects there is also semantic indeterminacy i.e. in denotation then [] the idea that the objects indeterminately denoted are themselves ontologically indeterminate ceases to do any work in explaining the indeterminacy in truth-value (p.159-160).45 The whole question concerning relative identity is said to be closely linked to all this, but that would really take me to far in view of the purposes of this paper.46 Another, less essential, line of criticism focuses on the use of LL, or more exactly, of CPOS of LL, by Evans. This precision is important, as both Broome [1984] and Garrett [1991] note. They stand in defence of a three-valued semantics for vagueness in which the inference rule MT (which amounts to CPOS, followed by MP) is not valid, blocking the derivation of (5) from (2) and (4).47 To close off this final section, Ill come back to its opening lines, with some critical questions about the nature of logical investigation, as applied here to the case of ontological vagueness. It is a well known principle that no logic, exotic as it may be, can be allowed, in the course of a formal proof, to add information not contained by its premisses. This means that if there were vagueness in the world, the evidence about the latter would already have to be enclosed in the premisses feeding any formal proof of it, as an argumentation can never be valid if its conclusion speaks about an object with property G, while the premisses dont (Van Bendegem [1997], p.107, my translation).48 Our logical tools are able to nothing more than disclose this concealed or entangled knowledge, using some sophisticated technique. As far as Im concerned, a suitable motto for all articles refered to in this section would be the one of Richard Heck Jr.: I shall not, however, attempt to decide the question whether there are vague objects. The conclusion of the present paper is just that logic alone does not preclude the existence of such objects ([1998], p.274). Nor can it provide evidence of the opposite in its own right. Theres another form of reduction, making use of which one can easily trivialize this very discussion (as well as hundreds of others), and Ill suffice with raising it. The argument is that every piece of knowledge about the world, objective as it may be, is mediated through our representations. As a consequence, all potential vagueness should be imputed to the latter, so that nothing but an epistemical question can ever arise, as the so-called ontological problem, whether there actually is one or not, is de jure inaccesible to us. French and Krause raise this point at the beginning of their commendable [1996], which offers a survey of their joint work in the field of a
44

Mutatis mutandis, the same line of argument applies to object b in steps (1)-(2) of the proof. For an interesting, i.c. restricted interpretation of both inferences (and the proof as a whole), see Garrett [1991] (p.345).
45

Compare Everett [1996]: The claim that we cannot rigidly designate vague or indeterminate objects seems dreadfully ad hoc. [] It seems that if we can dub or causally interact with a determinate object in such a way as to provide it a with rigidly designating name, surely we should be able to dub or causally interact with a vague object in a similar way. [] The friend of vague objects really owes us some independent philosophical motivation [] and an account of how rigid designation is possible for determinate objects but why it fails for indeterminate ones (p.217).
46 47

See however Burgess [1989]-[1990a], Noonan [1990], Tye [1990], and Zemach [1991], to mention but a few.

This step can be written out in some more detail as follows. From (2) Eb, (4) Ea, to infer (4) (EaEb), as can be easily shown, and from (4) and (4) (a=b)(EaEb), i.e. LL, by MT to infer (5) (a=b). See Broomes article for some truth-tables (p.9), and Garretts for an instructive philosophical comment.
48

It should be added here that an element of relevance has to be in play. Clearly, one is warranted, e.g., to move from (x)(Px) to (x)(PxGx). As clearly as that, nothing can ever be gained by this move. Except for being allowed to speak about property G in connection with object x, that is.

18

quantum logical account, defending (a sensible approach to) ontic vagueness. Apart from them, also Rolf [1980] and Sainsbury [1989] have recognized the danger of cheap reduction at the heart of this discussion, and have tried to anticipate it, proposing more classically looking formal frameworks. In them, the distinction epistemic-ontic vagueness is (or better: can be) maintained, at least in the first stages of analysis.

7. Conclusion

In conclusion of my paper, I shall not recapitulate the sections.49 Instead, here are some tips for those who now might plan on devoting some of their available library space (and money) to vagueness. In listing the following books, viz. two extensive surveys, two specials of philosophical journals and a reader, I do not in the least pretend completeness, but nevertheless firmly believe that they offer an economical basis for further exploration. One cant help noticing that I made frequent use of Williamson [1994], arguably the best reference to date, irrespective of its (quite natural?) pronounced theoretical position (3.1). In its turn, Burns [1991] sets out for an ambitious survey, with explicit sympathy for pragmatical aspects (section 5), but the result, however commendable, is both less complete and systematic than Williamsons. Taking a big leap backwards in time, allow me to draw your attention to 1975s special issue of Synthese, a milestone in the study of vagueness, for it collects some of the key articles in various schools of thought.50 Recently, The Monist also devoted a special issue to vagueness (the 2nd number of 1998s 81st volume, to be precise), with contributions ranging from both Hecks and Hydes on ontology (section 6), and Chambers on the intuitionistic view (refered to in section 1), to exercises in renewing strategies (leaning towards what was sketched in section 5), e.g. indeterminism (Burgess) and transvaluationism (Horgan), the latter of which is explicitly based on what Sainsbury [1990] told. The few gaps of primary literature still remaining after consultation of these journals, can be filled by Keefe/Smith [1996]. True, this reader will add some doubles to your archive (viz. Fines, Dummetts and Wrights [1975]), but next to them youll now be in possession of some further essential papers: Black [1949], Machina [1976] (both 3.3), Russell [1988], an abridged version of Wright [1987] (both section 4), Sainsbury [1990] (section 5), Evans [1978] (section 6), and some more. Finally, a copy of The Southern Journal of Philosophys 33th supplementary volume (1994), reporting on the Spindell Conference 1993 on vagueness, could prove useful, especially as an introduction to some pragmatically inspired accounts (section 5).

49 50

Please refer to the introductory one for a brief overview of the article. Namely the [1975]s of Fine (3.2), Zadeh (3.3), Wheeler, Dummett, and Wright (all section 4).

19

REFERENCES
ALSTON William P., Philosophy of Language, Prentice Hall, Englewood Cliffs N.J., 1964. BLACK Max, Vagueness: An Exercise in Logical Analysis, in his Language and Philosophy. Studies in Method, Cornell University Press, Ithaca NY, 1949 (pp.25-58). BROOME John, Indefiniteness in Identity, Analysis Vol.44 1984 (pp.6-12). BURGESS J.A., Vague Identity: Evans Misrepresented, Analysis Vol.49 1989 (pp.112-119). BURGESS J.A., Vague Objects and Indefinite Identity, Philosophical Studies Vol.59 1990a (pp.263-287). BURGESS J.A., The Sorites Paradox and Higher-Order Vagueness, Synthese Vol.85 1990b (pp.417-474). BURGESS J.A., In Defence of an Indeterminist Theory of Vagueness, The Monist Vol.81 1998 (pp.233-252). BURNS Linda, Vagueness. An Investigation into Natural Languages and the Sorites Paradox, Kluwer Academic Publishers, Dordrecht/Boston/New York, 1991. BURNYEAT M.F., Gods and Heaps, in: SCHOFIELD Malcolm et al. (eds.), Language and Logos. Studies in Ancient Greek Philosophy, Cambridge University Press, London/New York, 1982. CHAMBERS Timothy, On Vagueness, Sorites, and Putnams Intuitionistic Strategy, The Monist Vol.81 1998 (pp.343-348). COPELAND B. Jack, Vague Identity and Fuzzy Logic, The Journal of Philosophy Vol.94 1997 (pp.514-534). DEAS Robert, Sorensens Sorites, Analysis Vol.49 1989 (pp.26-31). DUMMETT Michael, Wangs Paradox, Synthese Vol.30 1975 (pp.301-324). EVANS Gareth, Can There Be Vague Objects?, Analysis Vol.38 1978 (p.208). EVERETT Anthony, Qualia and Vagueness, Synthese Vol.106 1996 (pp.205-226). FINE Kit, Vagueness, Truth and Logic, Synthese Vol.30 1975 (pp.265-300). FREGE Gottlob, Begriffsschrift. A Formula Language, Modeled upon That of Arithmetic, for Pure Thought, 1879, in: VAN HEIJENOORT Jean (ed.), From Frege to Gdel. A Source Book in Mathematical Logic, 1879-1931, Harvard University Press, Cambridge Mass., 1967, 19712 (pp.1-82; translated from German by Stefan Bauer-Mengelberg). FRENCH Steven, KRAUSE Dcio, Quantum Objects are Vague Objects, Sorites (Electronic Quarterly of Analytical Philosophy, edited by Lorenzo PEA, Madrid) Vol.6 1996 (pp.21-33). GARRETT Brian J., Vague Identity and Vague Objects, Nos Vol.25 1991 (pp.341-353). GOGUEN J.A., The Logic of Inexact Concepts, Synthese Vol.19 1968-69 (pp.325-373). HAACK Susan, Deviant Logic. Some Philosophical Issues, Cambridge University Press, London/New York, 1974. HECK Richard G. (Jr.), That There Might Be Vague Objects (So Far as Concerns Logic), The Monist Vol.81 1998 (pp.274-296). HELLENDOORN Hans, Reasoning with Fuzzy Logic, Ph.D. Technical University Delft, published on his own, 1990. HORGAN Terry, Transvaluationism: A Dionysian Approach to Vagueness, The Southern Journal of Philosophy Vol.33 Supplement 1994 (pp.97-126). HORGAN Terry, The Transvaluationist Conception of Vagueness, The Monist Vol.81 1998 (pp.313-330). HORWICH Paul, The Nature of Vagueness, Philosophy and Phenomenological Research Vol.57 1997 (pp.929-935). HYDE Dominic, Rehabilitating Russell, Logique et Analyse Vol.35 1992 (pp.139-173).

20

HYDE Dominic, Why Higher-Order Vagueness Is a Pseudo Problem, Mind Vol.103 1994 (pp.35-41). HYDE Dominic, Vagueness (Review), Mind Vol.104 1995 (pp.919-925). HYDE Dominic, From Heaps and Gaps to Heaps of Gluts, Mind Vol.106 1997 (pp.641-660). HYDE Dominic, Vagueness, Ontology and Supervenience, The Monist Vol.81 1998 (pp.297312). JOHNSON Bruce, Is Vague Identity Incoherent?, Analysis Vol.49 1989 (pp.103-112). KAMP Hans, Two Theories about Adjectives, in: KEENAN Edward L. (ed.), Formal Semantics of Natural Language, Cambridge University Press, Cambridge Mass., 1975 (pp.123-155). KAMP Hans, The Paradox of the Heap, in: MNNICH Uwe (ed.), Aspects of Philosophical Logic. Some Logical Forays into Central Notions of Linguistics and Philosophy, Reidel, Dordrecht/Boston/ London, 1981 (pp.225-277). KEEFE Rosanna, SMITH Peter (eds.), Vagueness. A Reader, MIT Press, Cambridge Mass., 1996. LEJEWSKI Czeslaw, Ontology and Logic, in: KRNER Stephan (ed.), Philosophy of Logic, Basil Blackwell, Oxford, 1976 (pp.1-28). LOWE E.J., Vague Identity and Quantum Indeterminacy, Analysis Vol.54 1994 (pp.110-114). MACHINA Kenton F., Vague Predicates, American Philosophical Quarterly Vol.9 1972 (pp.225-233). MACHINA Kenton F., Truth, Belief, and Vagueness, Journal of Philosophical Logic Vol.5 1976 (pp.47-78). MANOR Ruth, Solving the Heap Paradox without Sacrificing Vagueness, draft presented at the Centre for Logic and Philosophy of Science, University of Ghent, Belgium, 1997. NOONAN Harold W., Vague Objects, Analysis Vol.42 1982 (pp.3-6). NOONAN Harold W., Vague Identity yet again, Analysis Vol.50 1990 (pp.157-162). OVER D.E., Vague Objects and Identity, Analysis Vol.49 1989 (pp.97-99). PRIEST Graham, ROUTLEY Richard, NORMAN Jean (eds.), Paraconsistent Logic. Essays on the Inconsistent, Philosophia Verlag, Mnchen, 1989. PRIEST Graham, Sorites and Identity, Logique et Analyse Vol.135-136 1991 (pp.293-296). PRIEST Graham, Fuzzy Identity and Local Validity, The Monist Vol.81 1998 (pp.331-342). PUTNAM Hilary, Vagueness and Alternative Logic, Erkenntnis Vol.19 1983 (pp.297-314). QUINE W.V.O., Word and Object, MIT Press, Cambridge Mass., 1960. RAFFMAN Diana, Vagueness without Paradox, The Philosophical Review Vol.103 1994 (pp.41-74). RAFFMAN Diana, Vagueness and Context-Relativity, Philosophical Studies Vol.81 1996 (pp.175-192). ROLF Bertil, A Theory of Vagueness, Journal of Philosophical Logic Vol.9 1980 (pp.315325). RUSSELL Bertrand, Vagueness, The Australasian Journal of Psychology and Philosophy Vol.1 1923 (pp.84-92), reprinted in: The Collected Papers Volume 9: Essays on Language, Mind and Matter 1919-1923, Edited by SLATER John G., with the assistance of FROHMANN Bernd, Unwin Hyman, London, 1988 (pp.145-154). SAINSBURY Mark, What is a Vague Object?, Analysis Vol.49 1989 (pp.99-103). SAINSBURY Mark, Concepts without Boundaries, Kings College, London, 1990. SANFORD David H., Borderline Logic, American Philosophical Quarterly Vol.12 1975 (pp.29-39). SANFORD David H., Competing Semantics of Vagueness: Many Values versus Super-Truth, Synthese Vol.33 1976 (pp.195-210).

21

SCHIFFER Stephen, Two Issues of Vagueness, The Monist Vol.81 1998 (pp.193-214). SORENSEN Roy, An Argument for the Vagueness of Vague, Analysis Vol.45 1985 (pp.134-137). SORENSEN Roy, Blindspots, Clarendon Press, Oxford, 1988a. SORENSEN Roy A., Precisification by Means of Vague Predicates, Notre Dame Journal of Formal Logic Vol.29 1988b (pp.267-275). SORENSEN Roy, The Ambiguity of Vagueness and Precision, Pacific Philosophical Quarterly Vol.70 1989 (pp.174-183). SORENSEN Roy, The Metaphysics of Words, Philosophical Studies Vol.81 1996 (pp.193214). SORENSEN Roy, Ambiguity, Discretion, and the Sorites, The Monist Vol.81 1998 (pp.215232). THOMASON Richmond, Identity and Vagueness, Philosophical Studies Vol.42 1982 (pp.329332). TRILLAS Enric, Mengers Trace in Fuzzy Logic, Theoria Vol.11 1996 (pp.89-96). TYE Michael, Vague Objects, Mind Vol.99 1990 (pp.535-557). TYE Michael, Why the Vague Need Not be Higher-Order Vague, Mind Vol.103 1994a (pp.43-45). TYE Michael, Vagueness: Welcome to the Quicksand, The Southern Journal of Philosophy Vol.33 Supplement 1994b (pp.1-23). VAN BENDEGEM Jean Paul, Tot in der Eindigheid. Over Wetenschap, New Age en Religie, Hadewijch, Antwerpen/Baarn, 1997. VAN KERKHOVE Bart, Vagueness Unlimited: A Look at Some Pragmatic Accounts of the Sorites Paradox, to appear, 200+. VARZI Achille C., Supervaluationism and Paraconsistency, to appear in the proceedings of the First World Congress on Paraconsistency (Summer 1997, University of Ghent, Belgium), 199+. WILLIAMSON Timothy, Vagueness, Routledge, London, 1994. WRIGHT Crispin, On the Coherence of Vague Predicates, Synthese Vol. 30 1975 (pp.325365). WRIGHT Crispin, Further Reflections on the Sorites Paradox, Philosophical Topics Vol.15 1987 (pp.227-290). ZADEH Lofti, Fuzzy Logic and Approximate Reasoning, Synthese Vol.30 1975 (pp.407-428). ZADEH Lofti, KACPRZYK Janusz (eds.), Fuzzy Logic for the Management of Uncertainty, John Wiley & Sons Inc., New York, 1992. ZEMACH Eddy M., Vague Objects, Nos Vol.25 1991 (pp.323-340).

22

You might also like