Physical Identity Eli Hirsch The Philosophical Review, Vol. 85, No. 3. (Jul., 1976), pp. 357-389. Stable URL: http://links.jstor.org/sici?sici=0031-8108%28197607%2985%3A3%3C357%3API%3E2.0.CO%3B2-2 The Philosophical Review is currently published by Cornell University.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/sageschool.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact
[email protected].
http://www.jstor.org Mon Dec 3 22:38:11 2007
The Philosophical Review, LXXXV, 3 (July 1976).
PHYSICAL IDENTITY
O N E way to present a philosophical problem about the identity through time of physical objects is initially to conceive of an object (or, strictly, an object's history) as comprised of a temporal succession of momentary stages. Evidently not just any succession of object-stages corresponds to a single persisting object; some do and some do not. T h e early stages of one object do not ordinarily go together with the later stages of a second to add up to the history of some single third object. So in order for object-stages to add up to a single persisting object they must be related in some special way. What we want is an analysis or definition of what that relationship is. In a sense, of course, any succession of object-stages, however arbitrary, does add up to something: perhaps to an event, or to a state-of-affairs or, if nothing else, at least to a "merely arbitrary succession of object-stages." What is important, however, is that not every succession adds u p to apersistingobject or body (I use these here interchangeably), where this fundamental category is to be understood as loosely comprising items which can straightforwardly be said to occupy space and to persist through time. Clearly only certain privileged successions are accorded the special status of uniting into a single persisting object in this sense, which gives rise to the question as to what the unity-making relationship is in virtue of which some successions enjoy this special status. One can usefully distinguish three kinds of responses to this question. The first is in a way ananti-response, and is associated with one traditional notion of substance. According to this view the unity-making relationship which constitutes the persistence of material substance cannot be analyzed or explained in terms of any ordinary observable properties or relations. This mode of persistence is consequently regarded as in some sense ultimate. I shall present an interpretation of this view in section V of this paper when I discuss the identity of matter. A second response is to maintain that we can formulate a general
ELI HIRSCH
analysis of the unity-making relationship in terms of such notions as qualitative and spatiotemporal continuity. This approach is fairly prevalent in empiricist literature on the subject and a version of it which is found in Russell and Broad will be discussed in section 111. Finally there is the sortal theory which Wiggins and others have lately espoused and which I shall initially examine. T h e basic idea here is that the only way to give a proper analysis of our concept of identity through time is first to divide objects into different sorts (where different sorts are roughly associated with different nouns). We cannot, as in the Russell-Broad approach, provide one general analysis which would be applicable to the unity-making relationship of every different sort of object (we cannot state "identity criteria" for objects in general), but what we can d o is provide, with respect to each differentiated sort of object, an analysis of the distinctive unity-making relationship which applies to that sort. On the whole I intend to argue for a position which is something of a compromise between the three just mentioned. One of my contentions will be that our general concept of the persistence of a body can be partially analyzed without recourse to sortal differentiations. Though I agree with the sortal theory to the extent of holding that various crucial nuances of our concept of physical persistence depend upon sortal differentiations, I think it is a mistake to ignore the potential scope of the sortal-neutral kind of analysis essayed by Russell and Broad. Furthermore I will maintain, in the general spirit of the substance tradition, that our overall conception of physical persistence contains an essentially unanalyzable dimension. Let me emphasize at the start one general limitation on the intended focus of this discussion. An object's unity through time, which is what I intend to discuss, is by no means the only philosophically challenging mode of object-unity. In particular one can raise questions about an object's unity through space which parallel in many ways questions about its unity through time. The spatial question would have to d o with our basis for treating some, but not all, aggregates of matter as unitary objects. This question will not be directly addressed in this paper, though some of the issues raised have a bearing on it.' I n order to focus properly on the question which will be addressed, a question essentially about identity Quinton presents a helpful but overly brief account of an object's unity through space under the heading of "spatial identity," in Anthony Quinton, The Nature of Things (Routledge & Kegan Paul, London and Boston, 1973), pp. 75-77.
PHYSICAL IDENTITY
through time, the perspective to adopt is one in which an aggregate of matter has (on whatever basis) already been delineated as a unitary object and our primary concern is to understand what it means to trace that object's career through time.
The sortal theorist's main idea, as I understand it, is that we cannot state criteria of identity through time for objects in general since there are different identity criteria for different sorts of objects. Wiggins expresses this idea when he says that there could not be "any usable account of what it is, in general, to make a mistake or avoid a mistake in tracing [an object] a . . . T o trace a I must know what a is."2 He then explains that to know "what an object is" in the relevant sense is to be able to apply a special kind of terrn to the object, namely, a sortal. It is only by reference to an applicable sortal that we can understand what it means to trace the ~ b j e c t . ~ Where F is a general term, to be able to give an account of how to trace the career of an F-thing is the same as to be able to state identity criteria for F-things; and this in turn is the same as being able to analyze the conditions under which a succession of object-stages can correctly be treated as stages of a single persisting F-thing. Wiggins believes that we can give such an account for some general terms but not for others. The terms for which we can give such an account are the sortals. (This might be taken as a definition of "sortal.") Thus we can state identity criteria for tables, and for apples, and for cars. But Wiggins thinks that we cannot state uniform identity criteria for objects in general, since these criteria will vary from sortal to sortal. Nor, according to him, can we associate identity criteria with such David Wiggins, Identzty and Spatzo-Temporal Continuzty (Basil Blackwell, O x f o r d , 1967), p. 35. C f . P. T. Geach, Reference and Generalzb (Cornell University Press, Ithaca, N.Y. 1962), p. 39 ff. I shall not be concerned here with the differences between Wiggins and Geach (see Wiggins, o p . cit., Part O n e ) , or with possible variations o n these differences (see Sydney Shoemaker, "Wiggins O n Identity," Philosophical Revieu, L X X I X [1970],pp. 530-535). W h a t seems m o r e important is the fundamental point o f agreement between sortal theorists that o u r identity judgments are dependent u p o n sortals i n that t h e sortals supply o u r identity criteria, different sortals supplying different criteria.
ELI HIRSCH
nonsortals as "red object" or "red round object" since these criteria will again depend upon whether we are talking about, for example, a red round apple, or a red round ball, or a red round comet, and so on. Wiggins is apparently claiming that our identity criteria are dependent upon sortals in a very radical way. Certainly when he says that we can give no usable account of our identity criteria in sortalneutral terms he must mean to imply something much stronger than simply that we could give no completely exact account in such terms. He must mean that we could not even formulate a usefully close approximation to our identity criteria without appealing to sortals, that we could not even formulate an account which works for the most part.4 This position strikes me as intuitively quite implausible. I am prepared to believe (and will in fact argue presently) that our sortal classifications affect our identity criteria in various significant respects. But should it not also be possible to formulate some underlying and general rule of identity which cuts through these sortal classifications, a rule which it would be correct to follow, if not always, at least almost always? When we consider what it means to trace the careers of such sortally diverse objects as, for example, trees, tables, cars, and the wheels of cars, certainly our intuitive impression is that, though there might be certain nuances of difference between the tracing rules which we follow in these cases, there must also be a significant common denominator running through all these cases, one which could supply a usable, even if not absolutely accurate, handle on all of them. T h e implausibility of Wiggins' contention that our identity criteria are in a radical way relativized to sortals can be brought out by reflecting upon the following obvious fact: A person will frequently be able correctly to trace the career of a new sort of object without requiring any information as to what the identity criteria are for that new sort. Certainly the simplest (though, perhaps, not the only possible) explanation of this fact is that the person is applying the Though I feel confident that there is ample textual evidence for attributing this extreme position to Wiggins I am not too concerned with the exegetical truth, and if this position is not really Wiggins'then I want to consider it anyway. I should perhaps mention that this extreme position (which I am attributing to Wiggins and which I now intend to attack) is one which I myself espoused in a earlier piece. See "Essence and Identity" in M. K . Munitz, ed.,IdentityanriIndiuiduation(N.Y.U.Press, 1971), esp. p. 39.
PHYSICAL IDENTITY
same criteria to the new case that he has already learned to use in the old cases. But this implies that, contrary to Wiggins, thereare usable general criteria which cut through sortal divisions. We can, for example, easily imagine a child raised on a farm who knows a substantial amount of English but who has never seen nor heard of a car, or, to make the case even purer, has never seen nor heard of any transporting vehicle. He is now shown for the first time in his life a blue car moving across an open field. There is no doubt that he is immediately in the position to say such things as, "That big blue thing (with the four round black things on the bottom) is moving across the field." In these circumstances it would seem natural to assume that the child is using the expression "that big blue thing (with the four round black things on the bottom)" to refer to the same object that we might refer to as "that car." And, what is critical in the present connection, having referred (in his way) to the car he seems perfectly competent to trace the car as it moves across the field. Moreover there seem to be no very obvious limitations on his ability to reidentify the car over longer periods or in more complicated circumstances. He seems, in short, capable of getting on quite well without the allegedly special identity criteria relativized to the sortal "car." This is evidence that in fact no such radical sortalrelativization is necessary, that in fact the general concept of the identity of an object provides, if not all, at least a significant part of the identity criteria that we ordinarily need. I want to distinguish carefully what I am saying here from something which I am not saying. I am saying that on the basis of our general concept of object-identity a speaker of our language, who knows how to operate that concept, will be able to trace objects in an essentially correct manner quite independently of any sortals that he can apply to the objects. But I am not saying that it would be logically impossible to speak a dfferent language, involving a different concept of object-identity, in which objects are traced in terms of identity criteria radically different from the ordinary ones. In fact I believe that there is a sense in which it is correct to say that any given scene could in principle be described in accordance with an unlimited variety of identity criteria. We can, for example, construct a language in which the word "car" is replaced by the two words "incar" and "outcar." The criteria of identity for incars and outcars can be roughly (only roughly) indicated as follows. The term "incar" will apply to any car that is entirely inside a garage, and where a car is partly inside and partly
ELI HZRSCH
outside a garage, "incar" will apply to the segment of the car that is inside; correlatively, "outcar" will apply to any car that is entirely outside a garage, and where a car is partly inside and partly outside "outcar" will apply to the segment outside. When (as we would say) a car moves from inside a garage to outside, the description in that other language would be: "An incar moved towards the exit whereupon it commenced to shrink in size until it eventually vanished; simultaneously with the shrinking of the incar an outcar appeared at the outside of the exit, and gradually grew until it attained the size and form of the original incar." In this description the original object inside the garage (the object which coincides with what we would call a car) is traced in such a way as to render it numerically identical with what is later a smaller object inside the garage, and numerically distinct from any object that is ever outside the garage. This description strikes us as utterly bizarre, but once we grasp the identity criteria that lie behind it (and this might take a little trying) we can see that it is a perfectly intelligible description and that it describes in different language precisely the same scene that we would describe without mentioning anything that shrinks or vanishes. T h e somewhat astonishing lesson to derive from this kind of a consideration is that even so seemingly logically simple a question as whether or not something has moved from inside to outside, or whether or not something has decreased in size, can be answered only on the basis of our commitment to certain identity criteria rather than to others5 But the above assertion that our concept of persistence and change is dependent upon identity criteria does not imply the sortal theorist's assertion that these identity criteria are themselves dependent upon sortals. The first assertion leaves open the possibility that our ordinary concept of the identity of a physical object already The incar-outcar language suggests a number of questions (related to "conventionalism" and "relativism") which I shall not attempt to answer here. For example, could there possibly be persons whose natural language (whose primary way of thinking) contains descriptions as strange seeming as the incar-outcar one, and not just in some isolated case but in general? And if there were such a language would it make sense to ask whether it or English gave in some sense a better or truer description of the world? My discussion is not intended to presuppose any special views about how to answer these questions. The possibility of artificially constructing the incar-outcar description serves at least to point up the (somehow very surprising) logical complexity of our concept of object-identity and the dependence of this concept on certain criteria rather than others.
PHYSICAL IDENTITY
commits us to a general range of identity criteria that are neutral with respect to sortal divisions. If this were so our identity criteria would not be selectively attached to our sortal concepts (at least not primarily), but would be attached instead in a general way to our understanding of what it means for something to persist through time, to move through space, and to have and change typical physical properties of shape, size, and so on. And again, the strong prima facie evidence that this is so is our confident expectation that anyone (who is operating with our ordinary concept of a persisting object) will trace new sorts of objects in at least an essentially correct manner. We are confident, for example, that someone who has never before heard of a car would undoubtedly describe the scene of a car moving out of a garage in terms of the essentially correct idea of an object maintaining its size while moving from inside to outside, and would not, in those circumstances, trace a shrinking object in accordance with the incaroutcar criteria. T h e latter criteria, it seems, clash in some general way with our ordinary concept of what the identity of an object consists in, and this is presumably why we are confident that no speaker of our language would apply those criteria. Our confidence seems then to be based on the assumption that a speaker of our language operates with a general concept of what the identity of an object consists in, and that quite independently of any sortals which he can apply to an object, this general concept of object-identity will allow him to make correct identity judgments in at least most ordinary circumstances. Certainly this seems to imply that our concept of the identity of an object does incorporate a generally usable range of sortal-neutral identity criteria.
If we try to formulate an account of what these sortal-neutral criteria are it would seem that an absolutely minimal condition of adequacy for the account is that it render inadmissable tracing an object in accordance with identity criteria of the incar-outcar variety. It should be evident that these criteria are merely one instance of a ~vhollygeneral kind of abberant criteria (abberant, that is, relative to our ordinary identity concept). These criteria are abberant insofar as they would permit us to trace an object in such a way as to combine what (in our language) ought properly to be regarded as stages of
ELI HIRSCH
the object and stages of some of its parts. A minimal condition of adequacy for any account of our identity concept is that it at least render inadmissable such drastic whole-part tracing confusions. Now it is a fairly remarkable facc chat most (possibly a]) of the sortal-neutral accounts of bodily identity that one finds in the literature d o not even satisfy this minimal condition. At any rate this is certainly true for the very common account which attempts to analyze bodily identity exclusively on the basis of considerations of continuity. For example, in the course of discussing "the durations of physical objects" C. D. Broad states: "A thing . . . is simply a long event, throughout the course of which there is either qualitative similarity or continuous qualitative change, together with a characteristic spatiotemporal ~ n i t y . Evidently Broad thinks that our criteria of "~ identity for objects consist essentially of two considerations: continuity of qualitative change and spatiotemporal continuity. T h e trouble is that these two considerations by themselves d o not give us the slightest clue as to what is wrong with the incar-outcar criteria, since the shrinking incar will exhibit a maximum degree of continuity of both sorts. Indeed if we tried to base our identity concept exclusively on continuity considerations, as Broad suggests, we should not ever have the faintest idea of how to trace an object's career since we can always move by continuous gradations from a whole object to its parts, or from part of an object to the whole, or from part of an object to adjacent parts. Broad's mistaken account is rife in the literature. Just to cite one more example, Russell says in one place: "Given any event A it happens very frequently that, at any neighboring time, there is at some neighboring place an event very similar to A . A 'thing' is a series of such events . . . It is to be observed that in a series of events which common sense would regard as belonging to one 'thing', the similarity need only be between events not widely separated in space-time."7 I take this to mean that the unity-making relationship which binds a succession of "events" into a single persisting object is nothing more than spatiotemporal and qualitative continuity. As we -
C . D. Broad,Scient$c Thought (Routledge and Kegan Paul, London, 1949), p. 393. See also p. 346ff. ' Bertrand Russell, Human Knowledge, Its Scope and Limits (Simon and Schuster, @
N.Y., 1948). p. 488.
PHYSICAL IDENTITY
have seen, this is hopelessly ~ v r o n g . ~ It is in the recoil from the Broad-Russell account that one first begins properly to appreciate the force and motivation of the sortal theorist's insistence that, as Wiggins puts it, an object must al~vaysbe traced under a sortal concept. We cannot possibly rely on considerations of continuity alone, but if we relativize those considerations to a sortal concept then we get a tracing rule (for example "Trace a continuous succession ofcar-stages") which it makes sense to follo~v. Nevertheless, for the reasons I gave in the last section I think it should be possible to formulate a sortal-neutral account of some usable range of identity criteria. I can think of more than one way to attempt this formulation, but the idea which seems to me most promising consists in emending the Broad-Russell account in a fairly simple way. T o the two continuity conditions mentioned by Broad and Russell I propose adding the follo~vingprinciple of minimizing change: Trace an object in such a way as to minimize as far as possible changes other than mere changes of location. This principle would explain why the incar-outcar sort of criteria are inadmissable. In tracing the shrinking incar we do follow a path that is perfectly continuous, but we violate the principle of minimizing change because we judge an object to be drastically altering in size (and shape, and weight, and so on) when we can (in the ordinary way)judge it to maintain a more constant size (shape, weight, and so on). Admittedly the shrinking incar will preserve constancy of location (the thing always remains wholly in a garage) but this kind of constancy is not to count. I think that the principle of minimizing change captures one intuition that we have about the incar-outcar sort of criteria. We want to say that when a car is leaving a garage there is no reason to judge that the object has altered in size (shape, and so on). I read this to mean that our attempt to trace the object continuously through space and time presents us with no need to countenance such alterations, and where there is no need the principle tells us we must not. My suggestion, then, is that our general concept of the persistence Russell sometimes adds the condition that the events which comprise a single object must be causally connected. (See, e.g., Our Knoulledge of the External World [W. W. Norton &Company, Inc., N.Y., 19291 p. 112.) But whatever thiscondition of causal connectedness amounts to it certainly cannot help us in the present context since the stages of the shrinking incar will be as causally connected as are the stages of any object (e.g., trees, rivers, melting ice cubes) whose material composition is continuously altered.
ELI HZRSCH
of an object allows for an approximately correct analysis in terms of a basic sortal-neutral rule of identity. The basic rule is: Trace an object's career by following a spatiotemporally and qualitatively continuous path which minimizes nonlocational changes as far as possible. This rule is, I believe, usable in a great number and variety of circumstances, though it will also lead to many problem cases which I will discuss shortly. For the moment I want to note that the basic rule will allow for a distinction between cases in which an object persists through change and cases in which the object goes out of existence. An object goes out of existence according to the basic rule (in the basic way, I would say)just in case it is no longer possible to trace its career in a sufficiently continuous and change-minimizing manner. I say "sufficiently continuous" because it must not be supposed that our ordinary identity criteria exact the most extreme form of conIn fact, whenever an object persists through the tinuity im~iginable.~ addition or substraction of a part we can only think of this as occurring, in a sense, instantaneously, and as involving a certain degree of discontinuity. When, for example, a branch is chopped off a tree we must think of the tree as suffering a somewhat discontinuous alteration in volume, shape, and overall spatial coordinates. This can be seen pretty easily by considering that if the tree's initial volume was, say, 15 cubic feet and its volume without the branch is 14 cubic feet (the branch having taken up one cubic foot), then certainly the tree could not be said ever to have taken up 14%cubic feet. Thus its volume "jumped" from 15 to 14; and a correlative degree of discontinuity must also have been suffered with respect to its shape and spatial coordinates. It is impossible to state precisely what constitutes a "sufficient" degree of continuity (and this might eventually depend, to some extent, on the sort of object being traced), but we may perhaps assume that in the most typical cases the outer limit of this conception is already reached at the point that we would have to think of the object as discontinuously doubling or halving its volume. We might then interpret the notion of continuity, at least at the level of the basic rule, as implying that an object cannot be traced continuously when it is precipitously broken up into fragments that are all as small as half of the original. In such a case the basic rule would enjoin us to -
-
For a more detailed discussion of continuity see Hirsch, op. cit., pp. 39-43
366
PHYSICAL IDENTITY judge that the object has gone out of existence.1° Another way that an object can go out of existence according to the basic rule is if it vanishes by merging indiscriminately into its environment. In such a case we cannot continuously trace the object along a path which has any claim to being change-minimizing. If, for example, a number of cars are melted down into a single indiscriminate mass then, though we could preserve continuity by arbitrarily identifying each car with some specified portion of the mass, there would be no plausible way to regard these identifications as change-minimizing since any number of other arbitrary identifications would be on an equal footing. Thus, in general, if there is no path which can be plausibly regarded as the distinctively changeminimizing one we can only say that the object has vanished and no longer exists. At the level of the basic rule, then, objects go out of existence either when they are precipitously rendered into fragments by breakage, burning, and so on, or when they vanish by merging into their environments. These ways of ceasing to exist d o of course correspond to some of the most standard cases. But there are also other ways to cease to exist which the basic rule cannot explain, since there are cases in which objects are said to go out of existence even though their careers could be extended along paths which are perfectly continuous and change-minimizing. There is in fact a certain general disparity between thejudgments of identity which we would make if we relied entirely on the basic rule and the judgments we actually make relying on sortal-relativized criteria. T h e scope and meaning of this disparity is what I next want to discuss.
OFTHESORTAL-NEUTRAL IV. LIMITATIONS ACCOUNT Perhaps the most immediately obvious reason why the basic rule can only serve to provide a good approximation to our identity scheme, but not a wholly accurate account of it, is the point just alluded to, that we sometimesjudge objects to go out of existence in l o I am assuming for this discussion that our concept of spatiotemporal continuity can be satisfied by a n object only if it is spatiully continuous at every moment of its career. Hence we cannot identify an object with the discontinuous summation of fragments that emerges from it when it is divided up, even though the path from the object to the summation is, in a certain sense, continuous.
ELI HIRSCH
ways that the basic rule cannot explain. But before returning to this point I want to discuss another limitation of the basic rule which I regard as more fundamental. Though the rule would enable us successfully to trace objects, parts as well as wholes, in a wide variety of circumstances, it suffers from a general, and ultimately destructive, kind of vagueness. For there will be too many circumstances in which the rule will not clearly guide us in choosing which of a number of paths is to be treated as minimizing change. The kind of case which gives rise to this problem can be characterized generally as one in which, starting from a given object, we can trace continuous paths in different directions each of which minimizes change in a different respect. I shall not here attempt a very detailed or systematic treatment of these problem cases. I think a few rather simple examples will suffice for my purpose of pointing u p the fundamental limitation of the basic rule, and the eventual necessity of invoking sortal criteria. What we will see in these examples is that our attempt to force the basic rule to work leads to indefinitely mounting complications, and there is finally no indication that we can formulate for even the most common cases a sortal-neutral procedure, accurately matching our actual identifying practices, for unambiguously choosing between conflicting paths which minimize change in different respects. T o begin with a very simple example, imagine that you have a red table and you decide to paint half of it black in order to wind up with a half-red half-black philosophical example. As you apply the black paint the red expanse which initially coincides with the full extent of the table gradually diminishes in size as it is encroached upon by a widening black expanse. Now if you tried to trace the table according to the basic rule you would be faced with the following choice. You might decide that the way to minimize change is to preserve as far as possible constancy of color. You would then judge that the original red table gradually diminished in size until it is now only half of a table. O r you might decide that the correct way to minimize change is to preserve as far as possible constancy of size. You would then make the ordinary judgment that the table's color has altered. Perhaps it will be suggested that this is not really a hard conflict since in tracing the table in the ordinary way we preserve not only constancy of size but also constancy of shape, so that we thereby minimize change in the greatest number of respects. Well, but suppose that the black paint significantly alters the texture and temperature of the surface to which it is applied. How should we
PHYSICAL IDENTITY
then weigh up all the constancies preserved by one possible path against all those preserved by its competitor? This might still be dismissed as too easy a case since I have omitted a possibly crucialconstancy which we preserve in the ordinary way of tracing the table, namely, its constancy as a separately moveable thing (that is, roughly, a detached thing that tends to move together with its parts). But any special appeal to this specific constancy fails to provide a general solution to our difficulty, since we would still have no way of dealing with a large variety of objects that are not, at least in any perfectly straightforward way, separately moveable: for example, a fence, or a radiator, or a tree, or the wheel of a car, or, for that matter, a table that is securely fastened to something else (for example, to the floor). We certainly cannot rely on any general presumption that objects either are, or must remain, detached from other objects. It might still be suggested that the kind of change-minimizing conflict which could arise in the case of painting an object can easily be resolved by simply adding a proviso to the basic rule that minimizing change in size and shape weighs more than minimizing change in other respects. This suggestion, however, will again fall far short of providing a generally applicable interpretation of the basic rule, since the mentioned proviso could certainly not qualify as a general principle, even if it works well enough for the specific kind of example just considered. There are many other examples in which our ordinary tracing procedure shows no bias towards preserving size o r shape. Consider, for example, what happens when you add bumpers to a car which previously had none. T o make this vivid suppose that the bumpers are imposingly large, prominently curvacious, and colored conspicuously different from the rest of the car. Here our ordinary identity criteria determine thejudgment that the car which was orginally bumperless is now bebumperred, with the necessary implication that the car has altered somewhat in size, shape, and color distribution. But at the level of the basic rule we could just as well have traced the original car along a path which perfectly preserved size and shape, as well as color distribution. We could have done this simply by judging that the original car is now sandwiched between the bumpers but does not contain them as parts. Our operative identity criteria, which are relativized to the sortal "car," determine us to trace a path which might be said, in crudely unanalyzed terms, to preserve the object's constancy as a whole car. From the sortal-neutral point of view of the basic rule,
ELI HIRSCH
however, there could be no decisive reason to favor this constancy over all the others that could be preserved in the alternative way of tracing. It is this last sort of case, in which objects are attached to each other, that presents what is perhaps the most serious dificulty for the basic rule." If the small objecty is attached to the larger object x to yield the composite object x-with-y the basic rule will not clearly instruct us whether or not we are entitled to trace a continuous path which would identify x-with-y with the original x. (Presumably the identification of x-with-y with the smaller y can be ruled out on grounds of insufficient continuity). Whether this identification of the composite object with the larger original component is permissible can only be determined by reference to the sortal under which we are tracing the object: it will depend upon whether we can treat bothx andx-with-y as coming under the same sortal. Though there is unquestionably a great degree of potential latitude in this decision, and we can think up any number of borderline cases, at least with respect to many typical and obvious cases a sortal concept will provide a basis for deciding about the composition of an object that is brought under the sortal. Thus we identify the car-with-bumpers with the original car-without-bumpers, and accordingly judge the car to have altered in size and shape, because we treat both the object-with-bumpers and the object-without-bumpers as coming under the sortal "car"; but we would almost certainly not say that the car altered in size and shape if, say, a small trailer or sled was attached to it, since we would not normally think of a trailer or sled as entering into the composition of a car. Nor, pretty obviously, would we say that a tree has altered in size o r shape if some car bumpers are nailed onto it, since the car bumpers would not naturally be treated as part of the tree's composition. Here, at least, are judgments of identity that are in a fairly compelling sense sortal-relative.12 I suggested earlier that someone, perhaps a child, who did not
I ' Actually an exactly paralled difficulty arises when objects are detached from each other. Cf. the discussion later about the trunk-tree. But 1would say that, independently of sortals, the basic rule can be made to deal easily enough at least with the case in whichy is merely brought into contact withx but not attached to it (e.g., an ashtray is placed o n a table), since here we can safely lay down the general proviso (which would no doubt admit of a few exceptions) that tracing a change-minimizing path should not involve identifying the intactx with the fragmented x-with-y.
PHYSICAL IDENTITY
know the sortal "car" could still pick out a car as, for example, "the big blue thing", and successfully trace its career in many ordinary circumstances. We are now in the position to understand why this is so. In many ordinary circumstances all that the child will need to trace the car is the basic rule. But we can now also take note of the kind of limitation which his ignorance of the sortal will impose upon the child's tracing ability. Imagine indeed that for the winter a small yellow sled is attached to the back of the car. (It will not, I think, necessarily matter whether or not the child knows the sortal "sled," but to simplify let us assume that he does not). How will the child see this? Will he see it, in the way we do, as the original blue thing maintaining its size and color but being attached to the yellow thing? O r will he see it as the original blue thing becoming bigger and partly yellow, that is, as now containing the yellow thing as a part? The basic rule provides no definite guidance. Tracing the first way, as we do, preserves constancy of size and color, but tracing the second way preserves (what may seem important) constancy of separate movability. Which of these two ways of reidentifying the object might strike the child as most natural is unclear. There is here a genuine illusion of clarity because the grip of the sortal on our thought prevents us from experiencing the conflict which could be generated in such a case at the level of the basic rule. But that a real potential for conflict does exist can scarcely be questioned once one stops to compare this case, of attaching a sled to the car, with the case of attaching bumpers. Our sortal-relative criteria determine a relatively clear (though not, certainly, an absolutely exact) basis for making discrepant identity judgments in these cases, but it is surely evident that from the child's sortal-neutral vantage point there can be no relatively clear difference between the cases (though there may be any number of obscure differences which might point him in one direction or the other). Both cases essentially leave the child with an option that is properly resolved only by appealing to sortal-relative criteria. It is worth noting that some cases of change-minimizing conflict correspond to, and in a way explain, a peculiar feature of our ordinary identity criteria, namely, that sometimes in tracing an object we will branch off in two different directions under two different sortals. These are the kinds of cases which Wiggins describes as two objects of different sorts occupying the same place at once. A rather clear example of this occurs if you chop off all of a tree's branches so that all that is left of it is its trunk. Then in a sense
ELI HIRSCH
the tree and the trunk are now one, but in a sense they are not since they have different histories, the tree having once been larger and the trunk never having been larger than it is. If we trace the trunktree backwards in time in order to determine its past we reach a conflict point where we can trace along different paths in order to preserve different constancies. These different paths in fact correspond to tracing under the sortal "trunk" and tracing under the sortal "tree." Tracing under the first sortal (that is, determining the history of the trunk) allows us to preserve constancy of size, whereas tracing under the second sortal (that is, determining the history of the tree) allows us to preserve, roughly, constancy of separateness (that is, the tree, unlike the trunk, is always relatively separate from anything like it). I n the last case our sortal-relative criteria resolve the conflict in a compromise fashion, by allowing us to trace both paths under different sortals. In other (perhaps in most) cases (for example, the case of painting the table and the cases of attaching the bumpers or the sled to the car) the sortal criteria resolve the conflict by choosing one path and discarding the other.13 But the basic rule would leave us essentially stranded in all of these cases. And, that, I suggest, is the primary reason why we ultimately need the sortal criteria properly to fill out our identity conception. The rule which tells us how to use sortals in judging of the identity of objects can be stated in rough general terms as follows: Trace an object's career by following a spatiotemporally and qualitatively continuous path under a sortal concept. This rule (which could be l3 That some of these paths are discarded is evident at least if we confine ourselves to the most typical and unquestionably correct sortal-covered tracings (e.g., tracing under "table" or "car"). But here it must be emphasized again that our sortal-relative identity criteria, though they are much clearer than what would be provided by the basic rule, are quite flexible and admit of indefinite borderline possibilities. For example, in the case of attaching the bumpers it may seem borderline correct to trace the career of thepart ofthe car other than the bumpers, and this "object" could perhaps be said to have first constituted the whole car prior to being sandwiched in between the bumpers (thus giving us a marginal possibility for tracing what 1 have called the "discarded" path). A related (but different) complication has to do with our judgments about the identity of matter (a topic which I will address at length in the next section), since we can also trace the career of the aggregate of matter which made up the car just before the bumpers were installed, and this aggregate of matter can be said to have then become sandwiched in between the bumpers. These various complications do not adversely affect my central position that the basic rule provides a reasonable approximation to, but not a wholly accurate account of, our ordinary identity criteria.
PHYSICAL IDENTITY
expressed much more elaborately)14 is constititutive of what we mean by "sortal concept." In terms of the rule we can roughly define "sortal" as follows: A general term F is a sortal if it is a conceptual truth (a rule) that whehever one traces a spatiotemporally and qualitatively continuous path under the term F one is thereby tracing the career of (what counts as) a single persisting F-thing.15 Terms like "table," "car", and "tree" seem to qualify as sortals on this definition. But a color term like "red" does not, as came out in the case of painting the table, where trying to trace a path under "red" gave us the (conceptually) wrong answer. Nor could a term like "takes up n cubic feet" qualify as a sortal since tracing a path which preserves constancy of volume will frequently yield the wrong answer in cases like that of adding bumpers to a car. What we have seen in the few cases examined, and could see in any number of other cases, is that the sortal rule enables us to trace objects through those junctures at which the basic rule is helplessly vague. And my view is indeed that the sortal rule is in essence nothing more than a clarzjcation of the basic rule. T h e excessively vague idea of tracing a continuous path that minimizes change now gives way to the relatively clearer idea of tracing a continuous path under a sortal concept. But these two ideas are not logically independent. When we trace under a sortal we are ordinarily tracing a path which incontrovertibly minimizes change; and even when we allow the sortal to guide us through points of change-minimizing conflict we nevertheless continue to trace a path which minimizes changein a certain respect (that is, in respect of satisfying the sortal concept). The sortal, we might say, orients us towards an object in terms of a specific viewpoint that clarifies for that object which constancies count, and which do not, in minimizing change. If I am right the basic rule continues to operate latently in providing our most fundamental standard of what the persistence of an object consists in. This implies, among other things, a principled restriction on the range of sortals which can coherently be introduced into our conceptual scheme. In order for a concept potentially to figure as a sortal it must be such that when we typically trace '"ee Wiggins, op. cit., pp. 34-40, and Hirsch, op. cit., pp. 32-28. Also relevant is Quinton, op. cit., pp. 66-71. This definition roughly squares with the one that I gave earlier when I defined a sortal as a term with which one could associate identity criteria, because the primary way to state identity criteria for F-things is, first, to state that tracing u n d e r F yields a persisting F-thing, and then, to explain how to trace under F.
ELI HZRSCH
a continuous path under it we are tracing a path which, at least from some intelligible viewpoint, minimizes nonlocational changes. This is why the introduction into our language of a sortal like "incar" would be, if not outright impossible, at least conceptually jarring in the extreme. T h e reason is that a shrinking path traced under "incar" fails to minimize nonlocational change in any respect which is remotely conceivable. l 6 Of course this general constraint on sortal introductions, imposed by the basic rule, does not explain why we have just the sortals that we have and not an indefinite number of other possible ones. T o explain this more adequately would presumably require an appeal to a wide variety of factors. (We would want to understand, for example, the distinctive role some sortals play in our attempts to formulate laws of nature, and the distinctive role some sortals play in our social and legal fabric. We would also need eventually to come to grips with the general issue whether our identity scheme is in any important sense "just a matter of convention.") But it would take me beyond my present limits to pursue this point further. The primary disparity between the sortal rule and the basic rule is that the former guides us with relative clarity in cases where the latter would guide us only very vaguely. This disparity, however, carries with it a derivative and somewhat more obvious one. The sortal rule implies that an object goes out of existence at the moment that we can no longer apply to it the sortal under which its whole career is traced. We can, admittedly, shift from one restricted sortal to another (for example, from "red car" to "green car") as an object passes through different phases, but the object's whole career, from beginning to end, must be traced under one unrestricted sortal (for example, "car"), and the moment this sortal can no longer be applied the object's career must be extended no further. But this can easily happen in such a way that, from the point of view of the basic rule, there is no reason at all to judge that the object has gone out of existence. If a car, for example, is thoroughly charred and twisted by fire, but suffers no discontinuous breakage, then the basic rule would have us say that it still exists in a different form, whereas the l6 Why is it that mereconstancy of location cannot count (much)?Atone level this is connected, 1 think, with the fundamental idea that a n object must be self-complete, which implies that its identity must not depend (too patently) o n how it is related to other things. But, more obviously, if constancy of location did count, then (location being relative to any number of frameworks) the idea of minimizing change could never have an even tolerably unambiguous application.
PHYSICAL IDENTITY
sortal rule forces us to say that, if it is no longer a car, it no longer exists. The sortal rule is the operative one, but we can still quite definitely sense in such a case the latent pull of the basic conception. For one thing, people (philosophers not excluded) are simply not that quick to admit (if they ever do) that the car has to go out of existence just because it ceases to be a car. Furthermore we feel distinctly inclined in cases like that of the charred car to stretch the sortal as far as possible to keep our identity judgment in line with the basic conception. ("Well, it's still a car in some sense.") Eventually though, despite the understandably opposite inclination, I think we must yield to the pressure of the operative sortal criteria and say that we are no longer presented with a car and therefore the car no longer exists. A tendency to avoid in discussing these matters is that of implying that the central role of our sortal criteria is precisely this, to have us judge that an object ceases to exist when it ceases to be the same sort. This makes it sound as if the sortal rule strikes like an arbitrary bolt from above to drive objects to an early doom. That an object ceases to exist at all is only a depressing corollary of the logical conditions of its persistence through change. And the central, and possibly indispensable, function of our sortal criteria is to clarify those conditions, which at the basic level are too amorphous. A somewhat earlier demise is only the necessary price that an object sometimes pays for having enjoyed a logically clearer mode of persistence. Let me now summarize the disagreement that exists (or appears to me to exist) between Wiggins and myself. T h e overall content and tenor of his discussion leaves the overwhelming impression that, according to Wiggins, if someone did not know any sortal which applies to an object (if he had no idea "what the object is") then he would be virtually at a total loss as to how to trace the object's career. This, I am quite certain, is false. Even if I have no idea what sortal applies to an object I can, for the most part, confidently trace it in accordance with the basic rule. Of course there will always be the outside chance that what I am identifying as a persisting thing has, according to the operative sortal criteria, gone out of existence only to be continuously replaced by something else. But this is by necesszty only an outside chance since it is certainly a general feature of our conceptual scheme to think of objects as persisting through far and away the large majority of changes that they can suffer. And so long as I am tracing a path which unambiguously minimizes change the reliability of this general presumption of persistence through
ELI HIRSCH
change is very significantly reinforced by the logical pressure, mentioned earlier, for us to keep our sortal leveljudgments about "going out of existence" more o r less in line with the basic level ones. Thus I can proceed to trace the object with a high degree of confidence, even if not with total assurance. T h e really fundamental difficulty only occurs when I reach a juncture of change-minimizing conflict. Here without any sortal to orient me I may indeed be at a loss. The situation may naturally strike me in a certain way, or I may just guess, but unless I can somehow bring a sortal into play I can have no confident expectation that I have made the right move. Such cases of change-minimizing conflict are in one sense extremely rare, since they occur in only a tiny proportion of the times that make up an object's duration. (Consider the small proportion of times in a car's total duration that the car is painted, or a sled is attached to it, o r a bumper is changed, o r anything else happens that could conceivably give rise to change-minimizing conflict). I n another sense, however, such cases are common, since they occur repeatedly and regularly. If we tried to rely entirely on the basic rule these cases would regularly afflict our thought and communication about the identity of objects with a far greater degree of obscurity and indefiniteness than our ordinary identity criteria tolerate. So the sortal criteria are ultimately needed, but they are not needed with nearly the insistence that Wiggins appears to think.
I want now to raise a certain puzzle about the way that we conceive of the material composition of things. The persistence of an object is defined, according to the foregoing analysis, in terms of the idea of a continuous path which minimizes change, if not change in all respects at least in some intelligible sortal-relative respect. Now in order for it to be possible to trace an object along a changeminimizing o r sortal-covered path the object must, in some sense, stand out from its environment; it must be, as I will say, articulated. For if the object were merely a wholly indistinct portion of some larger uniform mass we could trace continuous paths from the object in any and all directions, none of which could claim the unique status of minimizing change in any relevant respect." l7
The intuitive idea of "standing out" or "being articulated" might indeed be
defined by reference to the idea of traceability. We might say that an object is articu-
376
PHYSICAL IDENTITY Suppose, for example, that I specify some portion of wood which partially makes up the uniform wooden table in front of me. I might do this by momentarily laying my hand on the surface of the table and thereafter referring to the hand-shaped mass of wood which had momentarily extended directly downwards beneath my hand. Since this mass of wood is nonarticulated, since, that is, I am assuming that it does not stand out in any way from its environment, I cannot possibly trace it along a change-minimizing or sortal-covered path, for there will an indefinite number of paths extending away from that mass of wood, all of which contain qualitatively indistinguishable hand-shaped masses of wood. And this presents a puzzle. For we do think of that wood, of the wood under my hand, as having an identity of its own, as being spatially related to other masses of wood that surround it, and as in fact havingalmost certainly persisted for a longer time than the table has. But what can this concept of persistence mean if it cannot be understood in terms of the idea of a change-minimizing or sortalcovered path? Perhaps this puzzle can be made a bit more stark by slightly altering the image. Suppose that I break the table to pieces so that I am left with a large number of wooden fragments. Holding one of these fragments in my hand I certainly want to say, "This fragment of wood came out of the table." Now this is, so far, no problem since we can in principle trace the fragment backwards in time to some portion of the table, tracing, that is, along a change-minimizing path or, if necessary, a path covered by the concept "fragment ofwood."ls There is no problem here because a fragment ofwood is articulated, and therefore "fragment of wood" can function in a straightforward way as a sortal which covers a tracing path. The problem arises insofar as I also want to say something else, namely, "This wood used to make up part of the table." (Not, strictly, "This fragment of wood
lated at a given moment if and only if it can be traced at that moment along a path which minimizes nonlocational change in at least some intelligible respect. T h e detailed implications of this definition are, however, far from clear, and the discussion which follows depends only on a minimal grasp of the intuitive idea. '"t is really more complicated than this. Depending on how the table broke use might have to say that this fragment is traceable back to some portion of a larger fragment, which is in turn traceable back to some portion of a still larger fragment, which is in t u r n . . ., which is in turn traceable back to some portion of the table. Still, so long as "coming out of" is transitive, there should be no major problem in explaining at least what it means to say that the fragment came out of the table.
ELI HIRSCH
used to make up part of the table," since my table was not made up of articulated fragments). But what can 1 possibly mean by this? What can I mean by saying that this wood was in my table yesterday though it was not yet articulated? If I want to talk about that self-same wood persisting from yesterday to today I should, it seems, have criteria of identity which explain what this means, which explain what the unity-making relationship is that binds together the stages of that persisting wood. But it seems that I have no such criteria of identity, at least none provided by the idea of a change-minimizing o r sortalcovered path.lS Now the very natural first impulse upon hearing this puzzle is to protest that the whole thing is really very obvious. (It is a philosophically important feature of this puzzle that we find it extremely difficult to take it seriously). "Look, that wood that you're holding in your hand came out of the table, right? If it came out ofthe table it had to be in the table. Now it certainly didn'tjump into the table when we weren't looking. So it's been in there all the time." This seems like an outright evasion, for 1 was not asking for a jzlstzjication (for evidence) but for an analysis (for criteria). I was not asking, "How do we know that the wood persisted before it was articulated?" which would imply that I already understand what that state of affairs amounts to, but rather, "What can we mean by speaking of the persistence of something which is not articulated?" T h e challenge is to show me the criteria of identity. (But we will see shortly that this natural response, which apparently ignores the request for criteria, may be in a certain sense quite apt). Our next impulse might be to try to state the identity criteria for the nonarticulated wood in terms of its location relative to the outlines of the table. T h e suggestion would be that our concept of the identity of a nonarticulated mass of wood is criterially determined by the rule that you are referring to the same wood so long as you are referring to the wood which occupies the same place on (that is, relative to the outlines of) whatever it makes up, in this case the table. This is wrong not merely because we already decided that mere constancy of location cannot properly define an object's identity. It is l 9 Russell consistently noted this puzzle about the identity o f matter, though he formulated it in the context o f his previously discussed erroneous account o f the identity o f ordinary objects. See, e.g., Bertrand Russell, "The Relation o f Sense-Data to Physics," in Mysticzsm and Logic (London: George Allen & Unwin Ltd., 1917), pp. 169-173.
PHYSICAL IDENTITY
wrong because it is certainly not the case that the same wood cannot alter its location on the table. Suppose that yesterday (before I broke the table) I had filed the table down to make it several inches shorter. Certainly this would warrant my now saying, "This wood in my hand was closer to the outlines of the table yesterday than it was a day before yesterday." And quite apart from filing or anything like that, anyone would have to admit that it is conceptually coherent to suppose that my table suffered from some chemical quirk which made some of its wood regularly contract while the rest expanded, in which case possibly this wood in my hand regularly altered its size and its place on the table. No, the location of the wood on the table does not criterially define its identity. So we have here a general problem about the concept of the identity through time of matter insofar as that concept carries with it the problematical idea of a quantity of matter maintaining its identity through periods of n'oriarticulation. Once this problem is grasped it is a short step to realizing that it pertains even to those cases where some matter is articulated. This is because the identity of the matter will never depend on its articulation, so we will always have to distinguish between the identity of the matter and the identity of the articulated object which it makes up. This distinction is required even where the only sortal under which we can trace the articulated object that the matter makes up is some term with roughly the force "articulated bit of matter" (for example, "puddle of water," "fragment of wood"). Thus we have to distinguish between the identity of a puddle of water (fragment of wood) and the identity of the water (wood) which makes it up. This distinction is very obviously required whenever a smaller articulated bit of matter is separated from, or joined to, the larger articulated bit which we are tracing. For example, we would not necessarily expect that in tracing a puddle (or pool or expanse) of water we are thereby tracing the very same water, since someone may have removed a glass of water from the puddle, or added one. Or if a splinter of wood is separated from a larger fragment (or lump or chunk) of wood then, of course, we no longer have the very same wood. But even where no smaller articulated bit separates away from, or joins, the larger articulated bit which we are tracing this still does not give us a critemal guarantee that we are tracing the very same matter. This is perfectly obvious in the case of the puddle of water since we know that water may evaporate out of, or condense into, the puddle. And a moment's thought should convince us that even in tracing an apparently intact
ELI HZRSCH
fragment of wood it must always remain a conceptually coherent possibility that some of the wood is vanishing (changing) into thin air while some other wood is materializing into the fragment. (Consider that a tree is in a way just a big chunk of wood, whose matter is continuously changing). So it seems that the idea of a changeminimizing or sortal-covered path gives us no criteria for the concept of the persistence of matter. How then does this concept operate? It might be suggested that the appropriate move at this point would be to search diligently for some perhaps complicated and ingenious formulation of our identity criteria for matter (where these criteria might vary significantly from one sort of stuff to another). Before directly addressing this suggestion I want to explain in some detail an alternative approach, which is the one that I hold to be correct. My position, to put it somewhat incautiously at first, is that we do not have any identity criteria for matter. But this idea needs to be explained. T o begin with it should be clear that the sort of identity criteria that I have been discussing in this paper are observational criteria. We have observational criteria for an identity judgment if we are able to explain these criteria by reference exclusively to conditions of a straightforwardly observable sort. This seems tantamount to saying that observational criteria must not go beyond the ordinary manifest properties of things (for example, something's being a table o r something's being red) and the ordinary manifest relations between things (for example, spatial and temporal). Now for the case of the persistence of an ordinary articulated object we can draw a fairly clear distinction between criteria of identity and evidence of identity. The criteria are those observational considerations in terms of which we can, at least roughly, analyze or define what our concept of the object's persistence consists i n z 0 (These criteria, as I argued, basically amount to the requirement that the object trace a continuous change-minimizing o r sortal-covered path.) And evidence for identity are facts from which we can inductively conclude that the criteria are satisifed. But for the case of the 2 0 Cf. Wiggins, op. cit., p. 43. Identity criteria, as I understand this, may be exceedingly vague and allow for any number of borderline possibilities. But we have criteria for "same F" if and only if in the most typical nonborderline cases our judgments about "same F" are analytically entailed by straightforwardly observable conditions.
PHYSICAL IDENTITY
persistence of matter I think we can draw no such distinction: here, in a sense, we have only evidence and no criteria. By this I mean that the only way to characterize our general procedure for judging of the identity of matter is to say that we reidentify matter in such a way as to arrive at the most coherent and theoretically satisfying account of what we observe. In this way we arrive at various principles which, both at the common sense leveI and at scientific levels, specify how bits of matter of various sorts may be presumed to behave under different observable circumstances. (For example, one such principle might be that, other things being equal, a nonarticulated bit of wood is presumed to maintain a constant location within a table.) But these principles are both partial and provisional, and may, within broad limits, be supplemented or revised in the light of scientific progress. These principles cannot therefore provide an analysis or definition of our concept of persisting matter. This account implies that our concept of persisting matter, even at the most common sense level, incorporates something in the way of a theoretical-explanatory posit of an underlying mode of physical persistence which ultimately accounts for the observed behavior of ordinary articulated objects. Even the most limited level of common sense contains an abundance of secure and well founded views about that underlying domain, though it is ultimately for science to fill in the details. But no belief about the observable manifestations of matter, however well founded, is analytic of our concept of persisting matter. It must always remain a conceptually coherent possibility that our explanations have been faulty and that matter behaves differently from the way that we think. Though I am maintaining that we have no ordinary observational criteria of identity for matter it might still be possible to provide some level of analysis or explanation of our concept of persisting matter. We can, in fact, broadly distinguish between two philosophical (and scientific) approaches to such an explanation. (a) It might be held that the persistence of matter is ultimately to be understood in terms of the persistence of certain articulated particles such as atoms or electrons. These particles are articulated by various unobservable properties in terms of which a sortal such as "atom" can be defined. T h e persistence of an atom is then analyzed in the standard way as depending upon the continuity of the atom's path under the sortal "atom." (Alternatively we can say that the atom's path minimizes change with respect to its unobservable prop-
ELI HZRSCH
erties.) T h e persistence of a bit of matter is thus ultimately analyzed in terms of the sortal-covered persistence of the atoms which make it up. (We can then say, if we like, that we have "theoretical identity criteria" for matter since we can give an analysis of the persistence of matter in theoretical terms.) (b) If we want to avoid a commitment to atomism we might say simply that the persistence of matter depends upon some unobservable relationship which binds the successive stages of a single bit of matter. This relationship has sometimes been called "genidentity."21 So we can, in a sense, explain the identity of matter by reference to that relationship. (Though this "explanation" seems very thin indeed, there is still nothing to prevent us from saying that in terms of the unobservable relationship of genidentity we can state "theoretical identity criteria" for matter.) The (a) approach has the advantage of allowing a closer analogy between the persistence of matter and the ordinary sortal-covered persistence of familiar articulated objects. The (b) approach, however, has the advantage of not forcing us to wed the concept of persisting matter a priori to atomism. (The (b) approach does not exclude atomism since there is nothing to prevent the relationship of genidentity from turning out to depend upon the sortal-covered persistence of atoms.) I shall not here attempt to penetrate the cloudy issues raised by these alternative^.^^ What is essential for my purposes is only to note that on either approach we do not have ordinary observational criteria of identity for matter. There is a philosophical tradition, loosely associated with the word "substance," to the general effect that our concept of the persistence of matter points to something beyond the reach of what can be straightforwardly observed. (This is the main gist of Descartes' discussion of the wax in Meditation 2.) This seems closely akin to the position which I am here advancing. My position would also entitle us, I think, to say something else which is close to the heart of the substance tradition, and this is that the unity through time of matter, in contrast to that of familiar articulated objects, is in a senseultirnat~. This sense of ultimacy derives from the two related points that the 2 1 See Rudolph Carnap, Introduction to Symbolic Lopc (Dover Publications, N.Y., 1958), p. 198. See also the remarks about genidentity in Hans Reichenbach, The Philosophy of Space and Time (Dover Publications, N.Y., 1958), p. 270ff. 2 2 See the conflict between atomic theories and "continuum" theories as discussed in Stephen Toulmin and June Goodfield, The Architecture of Matter (Harper & Row, N. Y., 1962), p. 64ff. and p. 158ff.
PHYSICAL IDENTITY
unity through time of matter goes deeper than (because it is not analyzable in terms of) the ordinary manifest properties and relations of things, and this unity is posited as playing a central role in the ideally most complete explanation of physical p h e n ~ m e n a . ~ ~ This, then, is my view on the identity of matter. Let me return now to the previously postponed suggestion that perhaps, contrary to me, there really are observational identity criteria for matter and we ought to look harder for them. I would not expect this suggestion to induce a great deal of enthusiasm at the present stage of the discussion. For we now have before us two possible accounts, two possible models, for understanding the nature of our judgments about the identity of matter. According to the criterial model thesejudgments are to be seen as analytically entailed by some complex conjunction of straightforwardly observable conditions. T h e model that I propose denies that any such analytic entailments apply to these judgments. The latter model must seem far more promising than the former in light of the cases considered earlier (the hand-shaped bit of wood in the table, the fragment of wood, the puddle, the tree.) For these cases certainly did not make it appear remotely hopeful that we could specify even vaguely some observable conditions which analytically entail the relevantjudgments about the identity of matter. T o recapitulate briefly, we recall that the identity of the nonarticulated bit of wood in the table could evidently not be analyzed in terms of any straightforward considerations of sortal-covered o r change-minimizing continuity. Nor, we saw, did the wood's location in the table afford a criterial basis for judging of its identity. And notice now the more general point that our ability conceptually to divide an object like a table top into such parts or portions as the part i n the middle, the curvedpart, that corner, that edge, and so on, will never give us the sought-after criterial basis for the identity of the bits of matter which compose the object, since it is by no means an analytic truth that the same parts, in this sense, must be composed of the 2 3 Cf. Sydney Shoemaker's discussion of the connection between the substance tradition and the analyzability of identity judgments, in SelfKnowfedge and SelfIdentity (Cornell University, Ithaca, N.Y., 1963), pp. 57-63 and pp. 254-260. My notion of a n ultimate and (observationally) criterionless mode of identity through time also relates to (though it differs significantly from) Chisholm's ideas about "strict" identity. See Roderick M. Chisholm, "Identity Through Time" in H.E. Kiefer and M.K. Munitz, ed., Language, Belief, and ,Wetaphysics (State University of New York, Albany, N.Y., 1970), and "Problems of Identity" inldentity andJndividuation, op. cit.
ELI HIRSCH
same matter. (Or, to consider a slightly different sort of case, when a piece of clay is deformed we may be able straightforwardly to observe the differing movements of such parts as, for example, the top part, the middle part, the thin part, the bumpy part. But none of this gives us a criterial basis for judging of the identity of the matter which makes up these parts.)24Are there any other even remotely plausible candidates for such a criterial basis? I certainly cannot see any. The reasonable conclusion seems to be that we simply d o not rely on observational criteria when we judge of the identity of some matter which partially composes an object. In a case like that of the wood in the table what we do rely on (at a common sense level) is the simplifying assumption that the wood's location in the table probably remains pretty much constant. That is, we rely on some such general simplifying principle as this: Other things being equal the location of a nonarticulated bit of matter within an articulated object may be presumed to remain pretty much constant. But this principle provides nothing like a criterial guarantee, and there is no saying a priori in how many different ways the principle might have to be augmented and reshaped both by common sense and science. (There is no saying a priori in how many different ways the "other things being equal" clause would have to be filled in to yield the simplest and most coherent account of the careers of different bits of matter.) In those various cases where we judge that some nonarticulated matter has altered its location within an object (for example, in the case of a piece of clay that is being deformed, o r in the case of a tree, o r in the case of a river) we rely on the most diverse evidential considerations and ultimately on the best theory of matter we have available. Even when we turn to the much simpler kind of situation, in which the quantity of matter under consideration is fully articulated, we find nothing like a criterial guarantee of our identity judgments about the matter. Suppose, for example, that I am holding a perfectly solid block of wood in my hand, stationary, not squeezingit too 2 4 One could certainly raise problems about the applicability of the basic rule to some of the parts or portions mentioned here, whose identities seem to depend upon their locations. But notice, then, how very disinclined we are to thinkof these items as genuine objects with identities of their own. (And notice too how natural it is to think of some of these items as places rather than objects; for example, the corner of the table is a place on the table.) In any case, my main concern here is only to show that these items d o not provide us with identity criteria for matter.
PHYSICAL IDENTITY
hard, in ordinary atmospheric conditions, with nothing observably weird going on (for example, there are no observable changes in either the size o r shape of the block). In such circumstances I would venture the following identity judgment: "The wood which now makes up this block is identical with the wood which made u p this block a moment ago." We are searching for some observable conditions which might criterially (analytically) entail the truth of this judgment. What needs to be ruled out, among other things, is the possibility that some of the wood vanished (changed)into air (or into the flesh of my hand) and/or that some additional wood materialized out of the air (or out of my flesh) into the block. What observable conditions could conceivably be such as to entail analytically that this has not happened? Surely not the conditions mentioned earlier (that the block was not squeezed, that its size and shape remained constant, and so on). It might be startling, but certainly not incoherent, for scientists to tell us that under just those conditions some bit of wood turns into air and vice versa. (The constancy of the block's size and shape might be accounted for by positing suitable expansions and contractions of the wood inside the block.) Is it not clearly hopeless to seek some other observable conditions which somehow would analytically entail the identity judgment? If someone is not convinced that this is hopeless then let him reflect upon the following point (which seems to me fairly decisive). In order for it to be true to say "The wood which now makes up this block is identical with the wood which made up this block a moment ago" it must be the case that all the original wood of the block is still in the block. This means that even if some minute and possibly invisible speck of wood in the block turned into air the identity judgment in question is, strictly speaking, false. But it seems obvious beyond the need for further argument that no straightforwardly observable conditions (no facts about the ordinary properties and relations of things), however complex, could analytically guarantee that no such minute speck turned into air. And since the identity judgment requires just this guarantee it follows immediately that no observable conditions can criterially guarantee the identityjudgment. * j (Is 2 5 It would certainly seem a mistake to try to suggest that though we have n o observational criteria for the judgment "x (which exists now) is the same matter as y (which existed before)" we d o nevertheless have such criteria for thejudgment "Some large portion ofx is the same matter as some large portion ofy," or, colloquially, ''x is
ELI HIRSCH
it really fair to intrude considerations about minute and possibly invisible specks of matter into a discussion of our most common sense concept of persisting matter? This decidedly is fair. For if someone, say a child, could not appreciate the relevance, with respect to his judgments about "same wood," of scientific theories about minute and invisible specks of matter, this would show us that he was not really employing the concept of the identity of matter but was instead still at the more elementary level of thinking only about, for example, "same block," "same stick.") The inescapable conclusion seems to be that in judging of the identity of the wood in the block we do not rely on identity criteria, but we rely instead on some such simplifying principle as the following: Other things being equal an articulated object (like a block of wood) may be presumed to alter its material composition, if at all, only partially and very slowly. Common sense and science augment this principle (fill in the "other things being equal" clause) not on the basis of some a priori criteria1 strictures but on the basis rather of the widest and in principle most unlimited variety of facts about how things alter in size and weight, decompose, mix together, and so on. The proper model here is not that of criteria application but rather that of working towards the most coherent theory of an underlying level of persisting matter. That our common sense concept of persisting matter involves the positing of an underlying reality may seem unbelievable because the concept strikes us as absolutely obvious and inevitable. Indeed the premise that the concept is not criterially definable in conjunction with the recognition of its utter obviousness could naturally lead to the surmise that the concept must be in some sense a priori. But I think that a more straightforward explanation of why the concept is so obvious is that our experience with ordinary articulated objects provides us incessantly with a literally overwhelming barrage of evidence that there exists underlying matter. That is, our most immediate and surface level observations of ordinary objects, whose persistence conditions are grasped in terms of observational criteria, more or less the same matter as y." Surely our understanding of these latter judgments presupposes our understanding of the former, presupposes, that is, our understanding of what it means to say of a particular bit of matter that it has persisted over some period of time. And since, by hypothesis, we have no observational criteria in terms of which toanalyzeor define what is meant by the formerjudgment it follows that we cannot have observational criteria in terms of which to analyze or define what is meant by the latter judgments.
PHYSICAL IDENTITY
present us with a range of facts which point unavoidably to the conclusion that these ordinary objects are composed of, and are ultimately to be understood in terms of, persisting items of a quite different sort. The range of facts to focus on contains as an instance just the sort of case discussed earlier, breaking a table to pieces. What we cannot avoid noticing in such a case is that the fragments which emerge from the table go together to add up to an object of at least roughly the size, and even form, of the table. And this is virtually a universal phenomenon: An object which contains no articulated parts is split up into a number of objects which add up to the original. How can one avoid trying to explain this by invoking the idea that the smaller objects which came out of the bigger one were in some sense already in there prior to their articulation? This conclusion seems so inevitable that one could scarcely take seriously my earlier question about our basis for judging that (the wood in) the wooden fragment had always been lurking nonarticulatedly in the table. And as far as the question where in the table it was, well, we adopt initially the simplest and most obvious hypothesis that it was always located right at the place in the table from whence it came. Such assumptions as this constitute a rudimentary common sense theory of underlying matter, and all that is now required is for someone like Thales to enter the scene and the rest happens by itself.26 But we can, I feel convinced, imagine what it would be like to live in a world whose articulated objects did not display those patterns of behavior which provide the primitive basis for our concept of persisting matter. In such a world the concept would have no use, at least not at a common sense level. I will only sketch this peremptorily. But imagine that whenever you break a table to pieces the resulting fragments add up to five times the size of the table. Imagine that if ever you start out with a basin full of water and remove a glassful 2 6 Perhaps the most rudimentary common sense theory of matter could be characterized in terms of the three presumptive principles which have emerged in the course of this discussion, viz.: (1) the presumption that the material composition of a n articulated object remains pretty much constant, (2) the presumption that the location of a nonarticulated bit of matter within an articulatedobjectremains pretty much constant, and (3) the presumption that when a smaller articulated object comes out of (merges into) a larger one then the matter which makes up the smaller one is subtracted from (added to) the matter which makes u p the larger one. I think that these three principles, working together, may be suff~cientto generate a fairly extensive body of beliefs about matter.
ELI HZRSCH
from it then when you pour the water back the basin overflows enormously. Imagine that whenever you dig a hole in the ground you wind u p with a pile of dirt which looks thirty times higher than the hole next to it. These imaginings would have to be generalized indefinitely before we reached the image of a world in which the concept of persisting matter had no immediate application. But I can see no reason to doubt that we can coherently broach this image. That it is extremely dfficult for us to do so shows how deep in our experience the concept of matter penetrates; but that it is possible for us to d o so shows that the concept is not, in the most ultimate sense, unavoidable. If we lived in that imagined world we would still have immediate use for the idea that "you can't get something out of nothing," since to get, for example, a new fragment of wood you would have (rather than pray) to make one by separating it out of something else. T h e idea for which we would have no immediate use is that when you create a new articulated object by separating it out of another object there was all along something, the persisting matter, waiting there to be articulated. This idea (which is, I think, in some sense really a strange idea) would not impress itself upon us. It follows from this account that I would regard our concept of the persistence of ordinary articulated objects as decisively more primary than o u r concept of persisting matter. This implies a repudiation of the common procedure (for example, in Locke's discussions of identity) of stating identity criteria for ordinary objects in terms of a prior notion of persisting matter. Often this procedure will take the form of substituting for the requirement of spatiotemporal and qualitative continuity the requirement that an object's material composition can alter only gradually. Now these requirements do in practice amount to virtually the same thing, but it is nevertheless a major error of principle to inject the concept of persisting matter into the very center of our ordinary identity criteria. T h e clear and observational concept of the persistence of an ordinary object deserves to be kept relatively disentangled from the more difficult and theoretical concept of persisting matter.27 We have, and have the 2 7 I say it deserves to be kept "relatively" disentangled since there seems to be no way to keep an absolute lid on pressures from the "underlying reality." For example, if I managed to glue those wooden fragments back together into a table I should probably want to say that it was the same table again, rehabilitated, basing myself, I suppose, on the ("theoretical") judgment that "it's still the same matter." In this general connection see Quinton's discussion of "compositional criteria" in Quinton, op. cit., p. 69.
PHYSICAL IDENTITY
right to have, a concept of persistence which operates essentially at the most straightforward observable level without much concern for what may be happening in the theoretical depths. I have discussed altogether three ideas of physical persistence: the basic conception, the sortal conception, and now the theoreticalexplanatory conception of persisting matter. The connection between the first two of these was that the sortal conception serves primarily to clarify the basic one. Can we see any other connections here? I think we easily can. T h e core of the basic conception was the rule that an object's career should be traced so as to minimize changes. But there seems to be a very natural movement of thought from saying "Minimize changes" to saying "Simplify changes." We can think of this movement as passing through the intermediary step "Do not countenance a change in an object unless it is necessary." This can be interpreted to mean, "unless it is necessary for tracing the object in a continuous path," which gives us the basic rule, and its eventual sortal clarification. O r it can mean, "unless it is necessary for providing the simplest explanation of the phenomena," which gives us the general procedure for judging of the persistence of matter. As I see it, then, at the root of our concept of physical persistence is the vague rule to minimize changes. From this rule there emanates two rather contrasting conceptions, on the one hand the relatively concrete and definite conception of the persistence of different sorts of articulated objects, and on the other the relatively abstract and indefinite conception of persisting matter.28
Dean Kolitch worked with me on an early draft of this paper a n d I a m indebted to him for any number of key corrections and clarifications. I also benefited greatly from the comments and suggestions made by Pamela Belkin, Georges Dicker, and the editors of The Philosophical Review.
