Chapter 1 Rough Draft

[Jeff]

I will post the rough draft of this chapter. It divides into two large parts.

I. Defends the idea of First Philosophy with a critique of its greatest modern detractor, W.V.O. Quine

II. Tries to make some sense out of the legitimate subject area of First Philosophy using a notion of metaphysical possibility.

These sections are not, as yet, integrated. Also, I make some outright errors. But this is a place to start.

Note that all footnotes and bibliographical references have been omitted.

[Jeff]

Chapter 1: Metaphysical Possibility

Section I: Inter-theoretical Truth

“The quest of a simplest, clearest, overall pattern of canonical notation is not to be distinguished from a quest of ultimate categories, a limning of the most general traits of reality.” W.V.O. Quine, Word and Object p161

Ideally, when we sat down to talk about the world, we would have already agreed upon both a logically perspicuous language and an inventory of the world's contents. The language, we would know, is simply the clearest shorthand for talking about what actually exists; the world's contents, we would also know, remain unchanged however we choose to talk about them. At least this is what most of us realists hoped for when we first became interested in metaphysics. Alas, it is not to be. It appears that we cannot escape talking about the world when we talk about talk and talking about talk when we talk about the world. But it remains controversial to claim that there is nothing to be salvaged from our naïve intuition about the relative independence of logic and ontology. Despite Quine’s pervasive influence on analytic philosophy, many contemporary philosophers still consider metaphysics to be a study of categories of a very general sort, categories that are expressed by very general predicates like being, relation, individual, class, entity, &c… On such an understanding metaphysics seeks to identify, describe, define, and express the interrelationships of these and other general categories. On this view, logic may provide an abstract framework for all our thought, but it is the business of ontology—however broadly construed—to provide its content. In this short paper I will defend the claim that there is something important worth saving in this distinction. I will do so by examining the best argument against it: Quine’s proposal for determining our ontological commitments.

I. The Ontology of Sand
Quine’s early work was on the subject of ontological commitment, and his essay “On What There Is” (1948) represents something of a watershed in metaphysics, which had hitherto sanctioned putative entities relatively casually. Further, Quine’s views about the extent and the determination of these commitments have perhaps reached the status of orthodoxy, so they are worth examining closely. Why do we need a theory of ontological commitment? The short answer is that in our language use, we are generally careless about committing ourselves to all kinds of spurious entities. For example, the sentence, “John is wearing a broad grin,” if taken in all seriousness, would seem to commit us to the existence of grins as a legitimate ontological element. A theory of ontological commitment, then, is a theory that tells us when we are, in fact, committed to the existence of certain entities. The goal of the theory is to enable us to paraphrase talk about entities that we do not regard as ultimate, grins say, in terms of other fundamental entities.

Quine’s theory involves both a criterion for ontological commitment and several rules for its correct application. The criterion is this: A certain sentence is committed to the existence of an entity if either the entity is named in the sentence or if the sentence implies an existential generalization in which that entity is needed as a value. The second disjunct is the important one; for many names can be paraphrased in terms of acceptable ontological elements. The rules for the application of the criterion provide the guide for determining which entities are, in fact, acceptable. Thus, the essence of Quine’s criterion is encapsulated in the rather obscure bit of jargon: “To be is to be the value of a bound variable.” But Quine means nothing very exotic by this. Most singular terms, Quine holds, can be paraphrased as definite descriptions. And these can be eliminated in such a way that no true claim contains a commitment to obscure entities. For example, the name ‘Pegasus’ is eliminated by the predicate ‘Pegasizes.’ We then say that nothing Pegasizes. Note: ‘Pegasizes’ appears here as a predicate, so the fact that subjects and predicates appear in sentences doesn’t necessarily commitment us to the existence of either particulars or properties generally. It takes more than an unparaphrased sentence to commit us to the existence of entities. If we take there to be universals, say, we do so because we have other grounds for thinking that a sentence like, "Humility is a virtue," refers outright to fundamental ontological elements.

However, it is important to recognize that Quine’s theory, without some additional guidance for specifying what should be ontologically sanctioned, is trivial. So what are the rules that Quine uses as a guide for applying this criterion? The first is that we restrict our ontology to the minimum that would be theoretically necessary for the expression of our beliefs. This restriction is familiar enough and has been, in one form or another, a commonplace in philosophy since Ockham. We might call it the rule of ontological economy. The second is that the ultimate vocabulary of our paraphrases must be both clear and scientifically acceptable. This restriction requires more discussion. First, what is it for a vocabulary to be clear? For Quine this amounts to an object having specifiable identity conditions. "We have an acceptable notion of class, or physical object, or attribute, or any other sort of object, only insofar as we have an acceptable principle of individuation for that sort of object. There is no entity without identity" (Quine (1981), 102). Classes, Quine points out, are identical when their members are identical (Quine (1997), 13). Properties aren’t so easily individuated. "The positing of attributes is accompanied by no clue as to the circumstances under which attributes may be said to be the same of different" (Quine (1969), 19). Indeed, there is no room for any intensional objects among the values of the variables of Quine's extensional language for these cannot be clearly individuated. It is not simply because such objects are abstract that they fail this condition, for Quine has no problem recognizing classes as acceptable ontological elements.

Quine also recommends the vocabulary of our paraphrases should be scientifically acceptable. In one sense this is something that no philosopher would seriously question. No ontologist would want to countenance entities which science rules out. But Quine’s claim is more radical in that he thinks that ontologies shouldn’t attempt to get beyond the objects sanctioned by science. In fact, he thinks that they cannot do so. To think otherwise is to embrace the ideal of a first philosophy, a body of intertheoretical truth concerning the ultimate categories of thought. Quine’s argument against this notion is his famous argument for the indeterminacy of translation (IT): “Manuals for translating one language into another can be set up in divergent ways, all compatible with the totality of speech dispositions, yet incompatible with one another” (Quine (1960), 27). If divergent manuals of translation are considered as different conceptual schemes, then the thesis becomes this: Different conceptual schemes can be compatible with the totality of sensory stimulations, which give rise to our speech dispositions, yet incompatible with one another. If true, such incompatibility implies that there is no objective criterion for evaluating equally adequate conceptual schemes.

Quine’s well-known argument for IT involves natives, jungles, Junglese, at least one mysterious utterance—“Gavagai,” and a linguist with a putatively impossible project—radical translation. Omitting all but the pertinent details, here is a brief description: Imagine a newly discovered tribe whose language is without known affinities to our own. The linguist has to learn the language directly by observing what the natives say. She first notices some promising one-word sentences with a direct link to a native’s current sensory stimulations. Translation then proceeds with the linguist compiling simple native sentences for objects that she recognizes. It can proceed in no other way, for the linguist’s task presupposes that she frame some general hypotheses concerning the equation of native utterances to English expressions. Quine argues that there will be many sets of such hypotheses that fit the native’s behavior and are yet incompatible with one another. For example, the linguist initially translates the native expression “Gavagai” as “Rabbit.” This translation is, however, not the only one possible. Based on the native’s behavior alone, the linguist could never be sure whether the native means by “gavagai” rabbits, all the various temporal segments of rabbits, all the un-detached parts of rabbits, etc… Quine contends that the linguist can, in principle, never get clear on the matter. For to question the native so as to refine the tentative translation presupposes the particular set of hypotheses selected by the translator. The linguistic apparatus needed to form the questions can only be supplied by a more extensive application of the hypotheses in question.

If the linguist is to produce a translation manual, surely a much more complicated task than translating mundane announcements of observable current events, then she has no choice but to read our ontological point of view, our conceptual scheme, onto the native language. Thus, it is possible to compile incompatible manuals of translation which both adequately account for all observable native behavior. If this is true, then there is no basis on which we could ever determine which was the “correct” manual. Incompatible manuals which equally account for native behavior are on a par, given that they are equally parsimonious, conservative, and modest in their posits, equally general in their range of applicability, and equally refutable. These theoretical virtues, namely: simplicity, conservatism, modesty, generality, and refutability, are our sole guides in choosing between translation manuals. But they are also our sole guides in choosing between equally adequate theories, generally. For Quine’s point is not confined to problems facing linguists engaged in radical translation. His point is the more general one that different conceptual schemes can be compatible with the totality of sensory stimulations yet incompatible with one another.

If Quine is right, then there is no fact of the matter concerning what the ontological commitments of a given conceptual scheme are. This is true even of our own naïve scheme. As we have seen, the names wantonly assigned in natural languages cannot be used to determine our own ontological commitments. Depending upon our choice of translation manuals, we will find a given scheme committed to any number of arrays of objects. One might suppose, then, that Quine would hold one set of ontological commitment to be as good as any other. But this is not the case. Quine believes that it is the business of science to tell us what there is. "Given physical objects in general, the natural scientist is the man to decide about wombats and unicorns" (Quine (1960), 275). His faith in science seems to be a product of two sources. First, science has had unparalleled success in allowing us to predict natural regularities. Second, the scientific method is the only method that embodies the theoretical virtues that were listed a moment ago. This method "produces theory whose connection with all possible surface irritation consists solely in the scientific method itself, unsupported by ulterior controls" (Quine (1960), 23). Thus, Quine commits himself to the most current ontology of science, whatever that may be. No philosopher can escape the adoption of some theory or other in her endeavor to describe the nature of things, and every theory carries its own ontological commitments. Science is just the most useful theory for most human purposes.

Quine takes his theory of ontological commitment to commit him to a sort of dualism of physical objects and classes—numbers being reckoned as classes of classes. "In a contest for sheer systematic utility to science, the notion of physical object still leads the field" (Quine (1960), 238). He also thinks that the admission of classes as values of variables of quantification "brings power that is not lightly to be surrendered" (Quine (1960), 266). In a 1997 response to process philosopher Leemon McHenry, Quine wrote, “My ontology is as dualistic as Whitehead’s and indeed the same as his except that classes have a clean-cut principle of individuation, namely coextensiveness, whereas properties have none.” Though some have taken Quine’s hyper-Pythagorean speculation in “Whither Physical Objects?” (1976) as a definitive endorsement of class-monism, Quine himself has resisted this move. At least it can be said that Quine endorses no intensional objects, e.g., propositions, meanings, and attributes, and finds most abstract objects—save classes—likewise dispensable, eliminating facts, measures, unactualized possibles, ideal objects, and geometrical objects. Thus, his position approximates nominalism, though Quine denies that he is, in fact, a nominalist (Quine (1960), 243). Typically, his procedure for the divestiture of some class of entities proceeds on Quine's principle that "explication is elimination". A term that is considered an object in a fairly limited array of contexts is found to be eliminable, for its useful function in the theoretical vocabulary can be performed using less troublesome forms of expression.

Quine’s criterion of ontological commitment does not itself imply that there is no difference between the world and the language that we use to describe it. It is only when it is combined with his rules for its application and his IT thesis that this is implied. Only then do we understand that 1) science is theory-building and as such its sanctioned objects are theoretical posits, which are not real in any intertheoretical way and 2) that

“[e]ach reduction that we make in the variety of constituent constructions needed in the building of the sentences of science is a simplification in the structure of the inclusive conceptual scheme of science… The quest of a simplest, clearest, overall pattern of canonical notation is not to be distinguished from a quest of ultimate categories, a limning of the most general traits of reality” (Quine (1960), 161).

II. Illicit Oases
I have identified three aspects of Quine’s theory that lead him to dissolve the distinction between the world and the language used to describe it, namely: the criterion of ontological commitment and the rules to be followed in applying it, the second of which implies the IT/IR theses. Before proceeding to criticize each of these elements, it should be noted that Quine thinks that there is some middle ground on the issue of the language-world distinction.

“Everything to which we concede existence is a posit from the standpoint of a description of the theory-building process, and simultaneously real from the standpoint of a description of the theory that is being built” (1960, 22)[italics mine].

But it seems that one cannot have it both ways. Either things are real apart from our theories or they are not. If objects are posits then they aren't real, and if objects are real then they aren't posits. This is simply a logical dilemma. Further, if I have given Quine’s argument correctly, the horn of the dilemma which confronts him is obvious: Quine’s position should be that everything to which we concede existence is a posit, and we simply have no way of determining if our posits are the right ones. Quine could—and does—argue that asking for such a determination itself makes no sense. But it is worth asking whether Quine’s repudiation of first philosophy, i.e., his naturalism, is a product of the argument that I have just given or one of its motivations. In some places he suggests that naturalism is a starting point and not the outcome of a chain of reasoning.

“Naturalism has two sources, both negative. One of them is despair of being able to define theoretical terms generally in terms of phenomena… The other negative source is unregenerate realism, the robust state of mind of the natural scientist who has never felt any qualms beyond the negotiable uncertainties internal to science" (Quine (1981), p72).

But it truly must be admitted that neither ‘despair’ nor a ‘robust state of mind’ constitute a reason for naturalism. At most these are merely motivations, and in an argument that concerns the legitimacy of anti-naturalism, construed as some kind of first philosophy, they must in all fairness be discounted. So if I have him right, Quine’s best argument for naturalism is the same as his argument for embracing the ontology of science. And this is the argument that I will now critically examine.

A. Problems with the criterion
Many of the arguments that I am about to give are fairly well-known. I repeat them because I have not seen them assembled in support of the point that I am now making. The first set focuses on the status of the paraphrases called for by the criterion. There seem to be two problems here. First, one might wonder if it is really possible to paraphrase all non-sanctioned entities out of our sentences. To make this difficulty felt, one might suggest the need for some standard to govern adequate paraphrases. This—or these—of course, would themselves be debatable. At least it seems fair to raise the question about how much can really be paraphrased away in a manner that does not change the original content of a sentence.

A more pressing problem is that paraphrase itself seems to suffer from something like the paradox of analysis. Paraphrase is a symmetrical relation. If two sentences are logically equivalent, then apart from some a priori metaphysical principles of one’s own, it is unclear how paraphrase could negate a real ontological commitment in an unparaphrased sentence. There is a dilemma here: Either a paraphrase is equivalent to an original sentence, or it is a replacement of it. If it is equivalent, then it is unclear how the paraphrase can be privileged with regard to its ontological commitments. If the paraphrase is a replacement of the original, then the advantages of a theory of ontological commitment are lost. After all, the idea was to find those claims that are agreed to be true and discern their ontological commitments (Quine (1953), p1). P.F. Strawson, who raises this point in against Quine, writes, “You do not abolish your commitments by refusing to be explicit about them, any more than you can get rid of unpleasant realities by employing euphemisms” (49).

B. Problems with the identity restriction
The second set of problems concerns Quine’s restriction on entification: “No entity without identity.” There are at least three problems associated with it. First, even if there is no principled criterion of identity, say for attributes, it is not obvious why this fact should be supposed to undermine the reference of sentences containing them. Perhaps a desire for theoretical simplicity is at work here, but if the choice is between simplicity and our prima facie duty to objects that we never deny in practice, then at least a strong presumption should favor the objects. Further, as was noted in the last section, it is not obvious how appeal to the lack of such a criterion can enable Quine to avoid commitment to the existence of such objects anyway. Secondly, the claim that all sanctioned terms must be backed by identity criteria appears to be a metaphysical claim itself, since it concerns the questions of what an object is and what objects there are. Such claims are ruled out by Quine’s IT/IR theses. But identity criteria are certainly metaphysical principles, though they may also serve the semantic purpose of allowing us to grasp the meanings of terms. Finally, it is arguable that not all kinds of objects can be provided with criteria of identity. The identity of some kinds of objects has been taken not to consist in anything else. For such basic objects identity is primitive and irreducible. For example, this may be true of persons.

C. Problems with IT/IR
As has been mentioned IT implies IR. IR, again, is the thesis that for any given class of sentences, there will be more than one assignment of referents that will produce a certain truth-value assignment, leaving a linguist with nothing by which to choose between them. If this is correct, however, then not only is there no fact of the matter concerning which conceptual scheme correctly mirrors the world, but also there is no fact of the matter about what the ontological commitments of any theory are. The various assignments of referents provide an array of ontological commitments, none of which can be said to be the true ones. So, IR militates against any doctrine of ontological commitment. Quine attempts to address this worry in his "Ontological relativity," but even there he is forced to concede that there is no fact of the matter about what the ontological commitments of a theory are, except relative to a background theory (1969, 64-7).

Another problem with the IT/IR theses is that they deny fundamental constraints upon theory-building by insisting that all such constraints are internal or relativized to a background language whose constraints are internal. A.N. Whitehead proposed four fundamental constraints upon theory building, namely: consistency, coherence, applicability, and adequacy. The first two are internal and rationalistic, consistency being logical consistency and coherence being a measure of how our concepts hang together. The last two are external and empiricistic, both referring to different ways experience is interpretable in terms of a theory. Applicability stresses the requirement that our theories be about the topic at hand. If theories are not interpretable in terms of their subject matter, then odd—crazy even—consequences follow. For example, a theory whose terms all deal with the refraction of light might be thought to apply to the frequency of divorce among children of the clergy (Ferré, 3). Here indeterminacy runs in reverse, and we begin to loose the ability to even interpret our theories. Adequacy amounts a standard of evidential completeness. It requires that all relevant evidence be considered before any conclusions are reached.

Quine would not be tempted to admit either of these last two constraints, because he would take them to be sufficiently captured by his own theoretical virtues. But I doubt this. For there is a tension in theory-building between coherence and adequacy, which is only one of the tensions between rationalism and empiricism that theories must strive to balance. The tension that I speak of is that between the coherence requirement that all evidence be relevant in some sense and the adequacy requirement that facile theories be broadened to include ever deeper levels of experience. In short, for theories to be meaningful, two conflicting tasks are constantly pursued: the task of rational reconstruction in light of new insight and the task of empirical reinterpretation in light of new experience.

Quine runs the risk of being over rationalistic in this regard for he cannot admit, in a meaningful way, standard empiricist concepts of theory-building. Don Garrett has distinguished five kinds of empiricism. First, methodological empiricism is concerned with the proper relation between theory and observation. It emphasizes observation over rationalistic theory-construction. Second, conceptual empiricism is concerned with the origins of the content of our concepts. It is echoed in the familiar refrain that experience precedes thought. Third, nomological empiricism is concerned with the possibility and status of scientific laws. A statement by such an empiricist would be that there are no a priori physical laws, rather the laws of nature are known through experience. Fourth, explanatory empiricism is a position regarding explanations that makes the metaphysical assumption that there are brute facts in nature, that we finally come up against the way things just are. Finally, reductive empiricism is a style or pattern of explanation in which some state of affairs, which is usually regarded as evidence for something else, turns out to be all there is to that state of affairs. Quine certainly qualifies as a conceptual, nomological, and reductive empiricist. But I doubt that he could be said to be a methodological or explanatory empiricist. And this charge should have some bite: For it is in just these ways that science itself, Quine’s refuge against his own ontological relativity, is best characterized. On the one hand, science involves a commitment to a careful balance of observation and theory-building. But this is usually thought to favor the observation side to some degree, so a commitment to methodological empiricism seems evident. Further, Einstein’s idea that the speed of light is a universal speed limit seems to involve a commitment to explanatory empiricism, as does the common scientific postulation of brute facts of nature which we may never be able to fully understand.

[Jeff]

III: Jungle Drums
Unless Quine has a better argument for denying what seems to be a natural human impulse to understand the world at its most comprehensive level, he should abandon his attack on first philosophy. It is, of course, true that the discipline attracts more than its fair share of crackpots and dogmatists, but that, in itself, is not an argument against pursuing it. In fact, as a good contextualist should know, every context carries implications for belief. Borrowing an example from process philosopher Frederick Ferré, if the chief of police turns out to have been the head of a crime syndicate, then all the detective’s plans and actions, laid in a false context, undergo a massive change in significance, both theoretical and practical. If only our sleuth had thought with more care about the possibilities! Larger contexts count, and thinking carefully about the small context while failing to do our best thinking about the largest context is simply irresponsible.
Quine has always known this, and most of his thinking does appear, upon reflection, as about as metaphysical as any other philosopher’s. In short, I find in Quine no convincing reason to give up the responsibility entrusted to philosophers—of all people—to preserve and advance at least a minimal commitment to first philosophy. As arrogantly as this has been done in the past, the strongest Quinean lesson worth heeding is that we should take more care in the future. We should especially consider carefully our ontological commitments. But what has been done poorly in the past may be done better in the future, perhaps even well. However I suspect that future advances in philosophy will not come from the desert-dwellers, but from the jungle natives themselves who through their closer contact with the world understand that we must be careful to distinguish what is real from the theoretical usefulness of a particular datum. Those who inveigh against first philosophy, those who contend that there is no such distinction, or that nothing philosophically important would be lost without it, have perhaps seen too much of the desert to appreciate life in the jungle.

[Jeff]

Section II: Metaphysical Possibility

Most philosophers would agree that there are several sorts of possibility. For example, if a set of physical laws are applicable in a certain possible world regardless of time and space, i.e., if a possible world is defined with certain invariable physical laws, then there seems to be an unbridgeable gulf between real and logical possibilities , at least in that world. For surely other laws than those that obtain there are logically possible, and if real possibility means, as a minimum, compatibility with the actually obtaining natural laws, and if these are not logically necessary, yet are eternally valid, then many things logically possible must always have been and will always remain really impossible (Hartshorne, p595). Unless we are prepared to countenance possible world chauvinism, we should admit that there are events in our own world that while logically possible are, at least here and now, really impossible. We might put all this another way: Most of what is logically possible does not actually exist in our world. Logical possibility is thus insufficient to establish what is really possible, since anything less than a sheer contradiction is logically possible. Similarly, there is a meaningful distinction between saying that a state of the universe is really necessary and saying that it is logically necessary. Again, the non-occurrence of the merely really necessary need not imply any contradiction.

This much is relatively uncontroversial. The questions begin to mount, however, when we start probing what we mean by real possibility. Does this "space" admit of divisions similar to that between real and logical possibility? If so, how are we to characterize this space—or these spaces? What are the categories needed for its—their—adequate description? These issues are not merely idle. They have profound implications for the very possibility of traditional metaphysics. For if a distinct realm of metaphysical possibility exists, then the revisionary metaphysics that have been the target of so much ire in the twentieth century still seems a fruitful discipline. Many still believe that metaphysics cannot be pursued in this grand style. It would certainly be a worthwhile project to show that revisionary metaphysics is, in fact, possible or, perhaps even, inescapable for all who attempt to adequately describe the world. The latter thesis is sweeping and beyond the scope of this paper, but I hope here to at least provide some reason to think that there is a meaningful notion of metaphysical possibility to be had. As I see it, the project has two components. First, I will attempt to show why something like metaphysical possibility must exist in any adequate description of the world. Second, I will try to give the concept of metaphysical possibility some characterization, though I admit that it will remain vague.

I. Analytic and Physical possibility
We might start by distinguishing a subcategory of analytic possibility within the realm of the really possible. Historically, this has been a relatively uncontroversial thing to do. Leibniz, for instance, distinguished truths of reason from truths of fact. Kant, of course, made the distinction central to his philosophy. But in the 20th century, analytic philosophers came to doubt the existence of this class of truths, W.V.O. Quine presenting the best-known attack on the idea. The structure of Quine’s argument against analyticity consists in a series of unsuccessful attempts to provide a necessary and sufficient condition, or criterion, for analyticity. He has a point, but many have raised objections. Here is one: Some philosophers have found Quine’s assessment of our lack of such a criterion inappropriate. It may be true that the concepts of analyticity, necessity, and synonymy are all inextricably bound up together; nevertheless, it does not obviously follow that we should make do without them in describing the world. In other words some have failed to find anything vicious in this circle of concepts. Though no noncircular account of these terms is in the offing, we needn’t reject them unless we are already drawn to a rejection of “mentalism” or find ourselves under the spell of Quine’s extreme externalism.

There is an established tradition that distinguishes logical possibility from analytic possibility, marking the former as strictly logical possibility and the latter as narrowly logical possibility (Konyndyk, p12-4). The point of the terminology is, I presume, to show the close connection between these types of possibility. Assuming that we can agree upon the laws of logic, strict logical necessity is truth in virtue of these laws alone. Narrow logical necessity is truth in virtue of these laws together with the definitions of any included nonlogical terms. Several examples have reached near-canonical status. I’ll go with the stallions. “It is not the case both that Ed is an uncastrated male horse and that Ed is not an uncastrated male horse,” is strictly logically necessary, being an instance of the relatively noncontroversial law of noncontradiction. On the other hand, “It is not the case both that Ed is a stallion and that Ed is not an uncastrated male horse,” is not strictly logically necessay since it depends for its truth upon the definition of the nonlogical term “stallion”. Rather, it is narrowly logically necessary. Since the use of the terms “logical possibility” and “analytic possibility” is a far less confusing way to mark this distinction, and since I can find nothing but a recent tradition that favors employing the more cumbersome nomenclature, I will break with tradition.


Besides analytic possibility, most philosophers see fit to distinguish a category of physical possibility. It appears to be analytically possible for a cow to jump over the moon, i.e., there appears to be no contradiction even if we are given the definitions of all the included terms; however, most of us would say that this is not physically possible, i.e., it violates the physical laws that obtain in our world. Similarly it seems analytically possible for the lights in the night sky to be tears in the celestial sphere through which an otherworldly light pours into our world. I say that this is analytically possible, for there doesn't seem to be any reason for us to rule out these possibilities simply through an examination of either the logical particles or the meanings of the any of the words in these sentences. If one thinks otherwise, if one finds our current scientific vocabulary coming from the mouths of babes, then these nonanalytic real impossibilities can be rewritten in an acceptable form, one that makes no tacit appeal to any law of nature. For example, perhaps we could agree that it is analytically possible but physically impossible that a domestic four-stomached ruminate could propel itself by a single voluntary movement over an object occupying more space than the ruminate in the same proportion as the moon is larger than a cow (Hartshorne, p594). To a first approximation then, it is sufficient for a state of affairs to be classed as physically possible if its obtaining would not violate the physical laws of that world, whereas a state of affairs can be said to be physically impossible if it would violate those laws. Physical possibilities could vary from world to world and in one world in different times and places. To say that something is physically necessary is to say that given the current world, time, and place, a certain state of affairs must obtain.

II. Metaphysical possibility
My characterizations of both analytic and physical possibility have been brief. But what I’ve in mind by them will be further clarified in what follows. For I will argue that there is a level of possibility between them, so to speak. I will attempt to show (a) that certain necessary truths (or possible truths) are neither logical, analytic, nor physical (b) that these modalities are situated in our broader picture of modalities in the required way, and (c) that these truths are classifiable under a certain criterion (or a set of related criteria), attempting to state these insofar as possible.

It is probably worth mentioning that the use of a distinct realm of possibility to provide a domain for the categories of metaphysics differs from many of the traditional methods of classifying this branch of philosophy. For example, the a priori / a posteriori distinction could be thought to provide a methodological classification of the discipline. Some philosophers have argued that what is distinctive about metaphysics is the fact that it uses an a priori method, i.e., a procedure whose rules require that all metaphysical conclusions are grounded in premises whose truth is not observational. This methodology is distinguished from an a posteriori one in which the truths of metaphysics are generated from observational premises. Whitehead held that the methodology of metaphysics was a posteriori in this way, hence it could not be distinguished from other disciplines on this basis (p6-8). Indeed several other methodological differences might be thought to distinguish metaphysics, e.g., that it is inductive rather than deductive, that it is prescriptive rather than descriptive (or vice versa), &c… And methodology is just one criterion itself. Other conceptions of metaphysics have it distinguished according to the objects that it classifies, by its aim, and by several other types of criteria.
While I cannot remain wholly neutral to this debate in my espousal of a realm of metaphysical possibility, neither do I have the space to adequately address it here. Suffice it to say that I consider metaphysics to be a study of categories of a very general sort, categories that are expressed by very general predicates like being, relation, individual, class, entity, &c… Metaphysics seeks to identify, describe, define, and express the interrelationships of these and other general categories. I take the realm of metaphysical possibility to be the home of many of these categories. In fact, I would use this realm of possibility to give some specific content to what it means to be a general category as opposed to a derivative one. However, my position is not that all the categories of metaphysics lie in the realm of the metaphysically possibility, for there is much difficulty associated with determining ultimate and irreducible categories. The boundaries of this domain are bound to be fuzzy. Further, categories may shift and change from time to time as other disciplines develop which had hitherto been the province of metaphysics. This is all too vague, but it must suffice for now. Perhaps it helps to know that I see my view on these matters as a conceptual descendent of the categoreal scheme of A.N. Whitehead.


I now turn to the first part of my task, i.e., showing that certain necessary truths (or possible truths) are neither logical, analytic, nor physical. I shall proceed by adducing three examples. The first two are not original, but I have not seen any of them assembled with in an attempt to make the point that I am urging.

Case 1: For any x, x is water if and only if x is H2O.
Again, we are looking for truth that is neither analytic nor merely physically true, nevertheless truth that we would be inclined to call necessary in some sense. One candidate is that water is H2O. Is this an analytic truth, i.e., is it true solely in virtue of the rules of logic and the definitions of the terms? No, water is a mass noun that applies to that natural kind of colorless, odorless, but drinkable stuff that descends from the clouds as rain, forms streams, lakes, and seas, and is a requirement of all living beings that we know. H2O on the other hand is a chemical description denoting a substance whose simplest part consists of two parts hydrogen and one part oxygen. Of course, water is H2O, but what makes this statement true is the nature of water not its definition. It was a substantive discovery of science that water is H2O. Is this then a physical necessity, i.e., given the current world, time, and place, must water be H2O? Certainly this is true, but one would not want to say that this is all we mean by the assertion. It is physically necessary that water is H2O, but one would say this is partially due to its metaphysical necessity. For we do not mean simply that water is H2O only in our world here and now. Rather we mean that in any world at any time and place in which there is something that is water, that stuff will be H2O.

A definition, as such, is universally employable in any scheme of relationships whatever its characteristics. Definitions are simply vacuous if nothing exists that fulfills them. So definitions are, by themselves, insufficient to adequately describe possible worlds. Hence another scheme of relationships is needed to describe them. However, schemes of physical laws are likewise insufficient to do this, for physical possibility is too lean a vocabulary to account for the scope of our statements about the world. Thus, there is another scheme of relationships more general than physical law and presupposed by it. This scheme some philosophers call metaphysical possibility.

Case 2: Things exist that are not identifiable with mere collections of qualities.
The next two examples will each focus on a different side of the category of metaphysical possibility, the first on the distinction between analytic and metaphysical possibility. It concerns what some philosophers have conceived, perhaps unfortunately, as substance. A common objection to the admission of substances into ontology is that though substances might be conceptually necessary for humans to conceive of the world, they are not ontologically required to describe it. In one sense this is almost certainly true. Modern science has shown not only that a category of static being is not necessary to describe the microphysical nature of the world, but also that it is probably inadequate to do so. This is not surprising, since the category of substance has had since Aristotle strong ties to a linguistic accident, namely the subject-predicate structure of sentences in natural languages. Nevertheless, some such category—perhaps a nonstatic one—does seem ontologically necessary if we are to account for things, which are not reducible to or identifiable with entities of other ontological categories, e.g., collections of qualities, ideas, or impressions. Things do seem to be necessary across possible worlds in a metaphysical way, i.e., that things exist seems to be a natural requirement of any state of affairs whatever. The point here is not epistemic—how can we know that our concept of thing really applies to the world? The point is ontological—if we are to describe the world then we must have recourse to this category. On the assumption that we can at least make some progress in describing the world, ontological categories are transcendental in the sense that they are to be invoked in construing what experience reveals of the world. This is only to say that on a denial of an extreme externalism that holds that “the quest of a simplest, clearest overall pattern of canonical notation is not to be distinguished from a quest of ultimate categories” (Quine Word & Object, 161), the categories that we do invoke to describe the world are ontological. It is both unnecessary and impossible to deny that these are concepts.

Case 3: It is possible that causation is an instantiation of an effective-receptive property pair.
Failure to distinguish physical possibility from metaphysical possibility has caused problems in understanding the way both science and philosophy work. David Ray Griffin has argued that we need to distinguish hard-core commonsense from soft-core commonsense, our metaphysical ideas roughly corresponding to the former type (Griffin, p18-21). Hard-core commonsense is made up of “those notions that all human beings inevitably presuppose in practice, even if and when they deny them verbally” (p18). Although our soft-core commonsense notions might be overridden by science, our hardcore notions cannot be. One has to ask what it would mean for these notions to be giveupable. Science itself is a rational edifice. To the extent that it militates against certain metaphysical notions, I leave these unspecified here, it attacks its own foundations. Our hard-core beliefs, then, can serve as a compass telling us when we have gotten seriously off course intellectually.

One might attempt to demonstrate this distinction by arguing that physicalism is necessarily conceptually incomplete. If it could be shown that physicalism makes certain ontological assumptions that it does not adequately acknowledge, then those assumptions might reasonably be thought to correspond to our hard-core commonsense beliefs. However, our current project requires significantly less than this. For all that needs to be demonstrated is merely that it is possible that physicalism makes these assumptions. For if it is possible, then we must acknowledge a class of possibility more general than and presupposed by physical possibility, consisting, minimally, of the alternatives to the view that I present but which can, for present purposes, remain unspecified. The following argument aims to do just this.

The guiding idea here is that causation requires the instantiation of a receptive property for every effective property within a causal interaction. This might be more perspicuously stated: The claim x caused y to become F implies both that x has the power to effectively bring about F in y and (crucially) that y has the power to be causally affected in a manner that results in its becoming F. Only if y has such a power is it possible to distinguish between y becoming F and y being caused by x to become F. This idea, though an old one, is currently a minority position advocated mainly by process philosophers, like David Griffin. But the basic distinction between effective and receptive properties has been recognized by canonized Western philosophers, like Plato and Locke.

For the processists (and perhaps for Plato and Locke), the existence of receptive properties is a necessary condition for the instantiation of causal interactions. In other words, any instance of causation will involve instantiations of both types of properties. The metaphysics of causation will have recourse to both because effective properties cannot be effective unless some individual is receptive to them.

Does modern physics include receptive properties in its ontology? We may begin to answer this question by noting that physics has never accepted the existence of a purely receptive being. The reason is obvious: Such a being could never causally affect us or any of our scientific instruments. We would be causally isolated from such an entity. Does physics accept the existence of entities that are a mix of effective and receptive properties? Here we must be careful. It is one thing for physics to be compatible with the existence of receptive properties. It is quite another thing for receptive properties to appear among its ontologically sanctioned properties. Since physicalistic parsimony admits only these sanctioned properties, they are the ones that concern us here.

My argument for the conclusion that receptive properties do not appear among the ontologically sanctioned properties of physics is a modified version of a similar argument given by Rosenberg:

(1) Physics at least accounts for the effective properties of the world's fundamental entities.

(2) When proposing laws, physicists at least postulate regularities in the instantiation of effective properties.

(3) However, physical theories represent the minimal set of properties needed to explain experimental results.

(4) Receptive properties are not needed to explain experimental results; however important they are for ontological completeness, they need not figure in physical laws.
_______________________________________________
(5) Therefore, receptive properties do not appear in the descriptions of physical laws.


I take this argument to be valid. Is it sound? I don't think that the first two premises are controversial. Prima facie, their falsity would render physics a hopelessly inadequate description of the world. The third premise is only slightly more controversial. It merely states the commonly held belief in scientific parsimony. Roughly, it is the claim that other things being equal, the simplest theory, i.e., the one that posits the fewest forces, entities, properties, and laws, is the one we should accept. The fourth premise is the controversial one, and I will argue for it explicitly by examining the way properties are actually posited in this science. Of course, in following such a method, we may fail to examine some atypical explanations. So this second method of defending (4) can only make it likely to be true. I'll take a single case, which I think is easily generalized. It might be useful to remind ourselves of the intuitive distinction here: Receptive properties are simplistically described as powers to be affected by something while effective properties are powers to affect something.

Consider the textbook definition that physicists offer for "mass", i.e., the measure of an object's resistance to changes in its motion. Shouldn't such resistance be counted as a receptive property? It needn't be. Indeed, it shouldn't be so considered by a parsimonious physical theory. So long as resistance can be "cashed out" by effective properties alone, then the physicist's use of the term "resistance" as a designation for some receptive property is illicit.

However, the very measurement of mass confirms that physics need only be concerned with the correlation of effective properties. Consider the experimental situation in which mass is usually measured. Frequently this task is accomplished with an instrument (a balance) with the following property: When force Y is exerted on the pan, then a specific value can be read from the balance by a human observer. But force is an effective property--perhaps the very definition of "effective property". So, what the balance actually accomplishes is a correlation of certain effective properties. As anyone who has worked in a lab knows, force need not be exerted on the balance by any mass at all to get a reading for it. Balances are easily "tricked"—and the more sensitive balances are, the easier it is to trick them. This can be done by moving them suddenly or by blowing on the measuring pan. In such cases there is a force exerted on the instrument but no mass is being weighed. It would add nothing to the physical explanation of mass, then, to consider it as having some property that an instrument designed to measure mass does not in principle detect. Since measurements of mass suffice to account for all the regularities associated with mass, receptive properties are not associated with mass by physical theory, which is simply concerned with the correlation between forces of different types.

[Jeff]

III: All possible worlds
I turn now to the second part of my project, namely providing a criterion under which metaphysical necessities are classifiable. This project is considerably more difficult for a number of reasons, not the least of which is that so many different things have been classed as metaphysically necessary. Thus far, my procedure has been to remain pretty much agnostic about the specific propositions that populate the class of metaphysically necessary truths. Without specifying the contents of this class, a project far beyond the scope of this paper, it is difficult to define the criterion of their selection. Nevertheless, I think something helpful can be said.

Let’s start by distinguishing two types of questions about possibility. One might ask, “Is it logically (analytically, metaphysically, &c…) possible that x?” By this one is generally understood to be asking if a certain proposition could be instantiated in some possible world. And it is generally thought that unless it implies a logical contradiction, every proposition is instantiated in at least one possible world. On the other hand, one might ask, “Is x a logically (analytically, metaphysically, &c…) possible world?” By this one is understood to be asking if a total state of affairs meets some standard of possibility. Especially if we agree that there are different levels of possibility, it is important to realize that the second question, which ranges over worlds, is logically prior to the first question, which ranges over propositions. For the existence of levels of possibility implies that at least some propositions are not instantiable in certain worlds despite the fact that they are not strictly logically contradictory. Thus, answers to the first sort of questions can very easily go wrong when logical possibility is assumed to be the only standard a proposition must meet to be instantiable in all possible worlds. Specifically, we must understand that propositions can be logically and analytically possible without being metaphysically possible. Similarly, a certain proposition could be metaphysically necessary yet fail to be either analytically or logically necessary, though it would be both logically and analytically possible.

So our search for a criterion for metaphysical possibility has two constraints. First, it must distinguish some propositions as both physically necessary and merely metaphysically possible. Second, it must distinguish some propositions as both metaphysically necessary and merely analytically possible. Now, it is the possible worlds framework that is appealed to in most contemporary attempts to understand our modal concepts. And it is unlikely that a better way of understanding metaphysical possibility is to be found. So, this is the appropriate place to begin our search for the criterion in question. What is needed is a way of segmenting possible worlds such that necessary propositions of a certain type turn out to be true in all possible worlds of a certain type. Specifically, metaphysically necessary propositions should range over a certain domain of possible worlds, the metaphysically possible ones.

Now, there are various systems of modal logic that are defined and interrelated in well-known ways according to an accessibility relation (Loux, pp20-8). In T-models this relation is merely reflexive; in Brouwer-models the relation is reflexive and symmetrical; in S-4 models the relation is reflexive and transitive, and in S-5 models the relation is reflexive, symmetrical, and transitive. I pass over these well-known distinctions without further comment. If the desideratum is that metaphysical necessity applies to a class of possible worlds without exception, then the S-5 system of modal logic is the minimum system necessary to capture it. “For it is only an S-5 model structure, where accessibility is an equivalence relation, that allows us to construe every world as accessible to every other” (Loux, p28). So, we go some way toward providing a criterion by saying that metaphysical necessity ranges over all possible worlds, where those worlds are understood to be accessible to each other according to the constraints sanctioned by the S-5 model of modal logic.

This criterion helps us distinguish what is physically necessary from what is metaphysically necessary, for the former does not hold across all possible worlds. But the criterion fails to distinguish metaphysical necessities from analytical and logical ones. For these, too, apply in all possible worlds. Thus, there seem to be more analytically possible worlds than there are metaphysically possible ones. So, a conjunctive criterion appears to be in order. One condition will state the requirement that metaphysical necessity is, minimally, necessity in all metaphysically possible worlds (as that has been understood), and the other condition will mark the distinction between a proposition’s being metaphysically necessary but merely analytically possible.

I want to suggest that the notion of world already does this. So that metaphysical necessity can be adequately described as truth in all possible worlds. The all in this familiar bit of jargon distinguishes physical from metaphysical necessity, and the notion of a world distinguishes metaphysical from analytical necessity. A world is a total way that things can be, a maximally comprehensive state of affairs. If one is skeptical about the existence of states of affairs, this might be put another way: A world is an entity such that every possible proposition is made either true or false by it. In many ways a world is like a picture, for the interrelationships between the things that exist in a world—or the propositions that are made true by it—are given at once by the world itself. Thus, a world, conceived as a totality of relationships, is itself a constraint on possibility. For although anything is possible, as they say, not everything is compossible. I want to suggest that compossibility is a specifically metaphysical constraint beyond either the laws of logic or the definitions of nonlogical terms. To say that a proposition is metaphysically necessary is to say either that it is true in every genuine world or that it is a logical consequence of some other metaphysical necessity.

In a recent lecture at the University of Oklahoma, Stanford’s John Etchmendy discussed his work on diagrammatic reasoning. He showed a hundred-step proof in standard quantificational logic that was accomplishable in five steps when the auto-consistency of diagrams was appealed to. This seems to be a clear case in which the compossibility relation provides a further constraint on both representations and proofs than do ordinary logics. Of course, anything that one could prove using diagrammatic reasoning could also be proved by traditional means. The point is that diagrams provide a new constraint that is a function of the fact that only certain states of affairs, or propositions if you like, are compossible.

If one accepts that worlds are appealed to only at the level of metaphysical possibility, then one must also accept that it is unnecessary to appeal to them at the level of either logical or analytical possibility. However, it needn’t be illicit to make such an appeal so long as it is not made in an attempt to prove that certain propositions are not metaphysically necessary. Indeed it is customary in composing a semantics to appeal to both a meaning assignment and a world assignment in stating analytic truths. Of course, what is analytically true in all possible worlds will be metaphysically true in them as well. It is only when analytic possibilities are held to be necessarily metaphysical possibilities that an appeal to a world as a maximally possible state of affairs is required to distinguish genuine metaphysical necessities. In this sense then, there can be analytically possible worlds only in a Pickwickian sense. These are not worlds at all, for the constraints of language are not the constraints of metaphysical possibility. The alternative doctrine, that language itself fully discloses the metaphysical relations of all possible propositions is tantamount to the claim that there is no difference between talking about the world and generating the simplest, clearest overall pattern of canonical notation. So long as there is a difference here, however slight, a class of metaphysical necessity must exist, which we might be able to characterize as truth in all possible worlds.

[Jeff]

I'll be revising this chapter next week. I'll post the second draft here by next Friday 1/20/06.

The following week I will revise Chapter 2. I am hoping to have a complete draft of the entire dissertation by March 17th. This may be overly optimistic. We'll see. I'll post all the pieces as I proceed. And I'll probably be posting much less in other threads for a while...though I am working on a review of Brokeback Mountain right now.

Jeff

[Tyler]

You misspelled "antidisestablishmentarianism" in paragraph 82 of the 16th section.
: )
Holy shit, Jeff, that's a lot of writing.

[jtmx1]

I write a lot. You guys know this. If you read half of the bullshit that I post on the board here, then you are better friends than I ever deserved.

I guess that, like a few of you, I am driven by a near constant need to creatively express myself. If I watch a film I expect myself to say something about it. If I read a book or hear a song… On the inside I feel like Ennis Del Mar in that I know what I think, speak, or write is hopelessly inadequate. Unlike him, my inadequacy does not drive me to silence. I cannot keep the loves of my life to myself.

I am just really lucky that you guys choose to hang out with me. Otherwise, I think I would be quite lost.

Jeff