, Volume 39, Issue 2, pp 375–391

Rea’s Revenge and the Persistent Problem of Persistence for Realism


    • Philosophy DepartmentUniversity of Connecticut

DOI: 10.1007/s11406-010-9284-3

Cite this article as:
Strawser, B.J. Philosophia (2011) 39: 375. doi:10.1007/s11406-010-9284-3


Realism about material objects faces a variety of epistemological objections. Recently, however, some realists have offered new accounts in response to these long-standing objections; many of which seem plausible. In this paper, I raise a new objection against realism vis-à-vis how we could empirically come to know mind-independent essential properties for objects. Traditionally, realists hold kind-membership and persistence as bound together for purposes of tracing out an object’s essential existence conditions. But I propose kind-membership and persistence for objects can conceptually come apart and function epistemologically distinctly from one another—in which case the usual reliance by realists on an assumption of persistence to determine kind-membership conditions is unjustified. Thus, present realist attempts to explain how empirical detection of mind-independent essential properties for objects could possibly occur inevitably results in circularity. The charge against the realist is to explain why we don’t have to first discover persistence conditions for an object before we can ascertain kind-membership conditions for an object. If no answer is forthcoming, then it seems the weight of the epistemological objection to realism is back in full force.


Metaphysical realismPersistenceMaterial objectsElderRea


Realism about medium-sized material objects faces a number of challenges. Many can be lumped together as epistemological challenges questioning how we could access facts about objects’ supposed mind-independent essential properties of kind-membership and persistence through time. If it is a fact about the world that Ks go out of existence when they lose property X, how could we come to know that fact? Conventionalists have an easy answer: we can know such modal facts about objects because we impose them on our interpretation of the world. That is, we carve up the world into objects; kind-membership conditions or persistence conditions for any given object are not fully mind-independent, for if they were how could we access them? So realists, who wish to defend the claim that objects’ essential properties are mind-independent, have the task of explaining how we could empirically discover such properties. Some realists think they have an answer to the challenge via a story about detecting property clustering in objects and co-variant sameness across divergent kinds. Realists of this stripe hold that any object mind-independently belongs to some kind, and that (at least) some of the properties by virtue of which it belongs to that kind mind-independently amount to persistence conditions for that object.

But even if these realists are right in all this, I argue that there remains a further epistemological challenge resting on the association between kind-membership conditions and persistence conditions. Namely, there seems to be no empirical means by which we could ascertain persistence conditions for objects apart from first ascertaining kind-membership conditions. And therein lies a fair question: why assume kind-membership conditions should, or even can, come first? To make this challenge, I’ll contend that persistence conditions and kind-membership conditions for material objects can and do conceptually come apart. The association that persistence conditions and kind-membership conditions run together is deeply seated at the core of most realists’ presumptions about object classification.1 But once we view these as distinct, realist stories explaining how we empirically discover them become inadequate to account for how kind-membership could be ascertained before (or apart from) persistence conditions. That is, granting the realists their claim that we can access some (mind-independent) facts about objects’ essential properties, I still ask: “Why don’t we have to first empirically discover which changes a given object can persist across before we can discover which properties characterize the kind to which the object belongs?” Most realist accounts trying to respond to the epistemological challenge presume persistence for objects in order to arrive at empirically detectable kind-membership; and then use kind-membership to establish persistence conditions.2 But, I protest, such a move begs the question for empirical detection of objects’ persistence qua persistence itself and thus does not answer this further “discovery problem” against realism. For this is a rather simple circle if I am right in thinking that we can conceptually separate persistence conditions from kind-membership conditions. That is, these realists think that we rely on persistence judgments to make kind-assignments, and rely on kind-assignments to deduce persistence conditions. Of course, they don’t think they’re involved in such simplistic circularity, but I’ll argue that really, unwittingly, this is precisely what is occurring—or, at least, could be occurring, for all we know, until the realists provides an answer to this challenge as to why we should or shouldn’t think that one comes before the other.


My use of the phrase “discovery problem” follows Michael Rea’s aptly named epistemological challenge in World Without Design (2002) which centers on any naturalist version of realism about material objects (hereafter, RMO). RMO holds that objects exist mind-independently in the world. These objects have certain properties essentially—that is, without these properties the object would not exist. And, according to RMO, at least some objects have at least some of these properties essentially independent of human beings thinking about them or other relations external to the objects themselves.3 So, the true claims “necessarily, water is H2O,” and, “a gold atom essentially has 79 protons,” have straight-forward readings for the realist. Rea’s discovery problem (hereafter, DP) argues that there are no means by which we could ascertain that these claims are mind-independently true that is consistent with naturalism. Hence, a realist committed to a naturalist epistemology does not have proper justification for knowledge claims about intrinsic essential properties for objects (and, thus, lacks proper justification for RMO). I noted above that realists who hope to defend that objects’ essential properties are mind-independent have the task of explaining how we could empirically discover such properties. And Rea’s DP is designed to show that no such empirical means of discovery exists. Hence, more accurately, I should say that this task of explaining how we could empirically discover such properties is only true of realists committed to naturalism. If we have some non-natural means of detecting such properties, then there is no discovery problem for RMO.4 But, setting aside such possibilities, in this paper I will focus solely on the challenge facing realists committed to naturalism.5

Conventionalists, such as Alan Sidelle, offer very similar arguments against RMO following in the same epistemic vein; although their aim is against all forms of RMO, Rea’s non-naturalist version included. Sidelle argues that we are confined by the limits of our own concepts because our epistemic methods are incapable of providing us knowledge of a wholly independent reality. Thus, any knowledge we have of the modal realities of the world (such as, “necessarily, if that rabbit is run over by a steam-roller, it will cease to exist”) must derive their modal import from our conventions, not from the world.6 Conventionalists derive all of this from the perceived impossibility of explaining how we could gain knowledge of necessity merely through empirical observation. That is, even if objects have modal properties “out in the world,” Sidelle thinks there would be no way for us to tap into them.7

It seems the best response for the realist could be to simply dodge the issue of how the mechanics of epistemic access to essential properties would actually function and instead argue that we have strong pragmatic reasons for assuming that we somehow do have such access—or at least good justification for believing we do. And, as it turns out, there are some plausible arguments for pragmatically warranted belief in RMO to be found which Rea explores at some length. 8 These pragmatic arguments make the case that the success of science and our successful interactions with the world (enabling us to survive) give us warrant for believing in RMO, or, better, for believing that we somehow identify and access mind-independent essential properties for objects. And there is a variety of other evidence that could count for naturalism as good justification for believing in RMO, such as the relational agreement humans have when trying to pick out objects, cross-cultural agreement and language convergence on how we identify objects, and that we can revise our understandings of objects’ essential properties through empirical observation.9 But Rea and Sidelle think these arguments all fail to adequately justify RMO for a variety of problems they find with this kind of defense and pragmatic arguments in general.10 I won’t here rehash the debate over pragmatic arguments for realism, nor come down on whether or not they are fully successful in warranting belief in RMO.

Instead, I wish to say that here a mistake is made by Rea and others who raise epistemic challenges to RMO (such as the DP) in overly focusing such critiques on broad justification issues.11 For, I argue, these kinds of pragmatic defenses for the realist against the epistemic challenge to RMO simply miss the mark. The central dispute of the epistemic challenge to RMO should not be that pragmatic justifications fail to adequately justify belief in objects’ having properties which are mind-independently essential. Rather, the real challenge of the epistemic objection to RMO is conceptual as to the mechanics of how we could ever possibly come to know such things; and this is where the DP and all epistemic challenges to RMO should focus. This is what I call Rea’s Revenge: even if pragmatic arguments are successful in providing proper warrant for belief in RMO, the DP’s challenge runs deeper in questioning whether such knowledge is even possible (as I’ll elaborate below.)

And this is where Rea fails to focus the DP. Rea does, however, start the objection here and discusses this most crucial point of the DP early on. Rea writes, “it is widely taken for granted that we are somehow able to learn the modal properties of material objects; the difficult question is how.” (Rea 2002, 77.) But he is not satisfied to show merely that a realist has no story explaining how they can access facts about essential properties for objects. Rather, he wants to argue that naturalists cannot therefore be epistemically justified in RMO—but I argue that this actually neglects the DP’s real force. That Rea thinks modal plenitude could resolve the epistemic problem for the realist belies that he focuses the DP (incorrectly) on the issue of warranted beliefs in RMO. He writes, “...modal plenitude can stand alone as a way of accounting for the reliability of our beliefs [about objects’ essential properties]… one might accept a suitably extreme version of modal plenitude… [in such a case] the reliability of our beliefs [would be] trivially guaranteed.” (Rea 2002, 95.) On a realist account, of course, this would cause “rampant colocation” (Rea 2002, 95.) But what good does this do in resolving the real crux of the DP challenge? That is not the heart of the problem. A modal plentitude account would still face the question of answering how we can access material object’s essential properties, even if we had full justification for believing that we get them (trivially) right. This, then, helps show what the proper challenge of the DP should be: explaining the naturalistic mechanics of how our epistemic access would even be possible.

Thus, pragmatic arguments showing that we may be warranted in believing RMO do not answer the real epistemic challenge that the DP presses. There are actually two separate questions nestled within all epistemic objections to RMO: The first is whether or not we have warrant for believing that RMO must be true. The second is how we could possibly come to have knowledge about material objects’ mind-independent essential properties. The pragmatic arguments for RMO, if successful, only answer the first question. We can know things without knowing how we know them; that does not mean we do not have warrant for believing them. But how the mechanics of our knowledge about objects’ essential properties could possibly work is a problem which runs much deeper because of the nature of that knowledge; that is, our knowledge of objects’ mind-independently essential properties (assuming we have such knowledge) is quite different from other kinds of knowledge.

To see why this is, consider the case of the “chicken sexer.” Chicken sexing is a skill possessed by a very small number of people worldwide. These sexers detect the sex of a chicken at a very early stage of development and are able to do so very quickly. It is not presently understood how these chicken sexers are able to do this. They themselves do not fully understand it and have trouble explaining how they consistently and accurately detect the sex of the chick. At several weeks of development obviously detectable features of being male or female emerge, but when these elite sexers distinguish them at a younger age there is no obviously detectable way of telling them apart.12 Time and again these sexers get it right at high rates of accuracy. (They are therefore very valuable to the poultry industry because they can set chicks on a breeding/feeding track based on sex much earlier and, therefore, more profitably.)13

So a realist might argue that just as (at present) it is not fully understood how chicken sexers successfully determine the sex of newborn chicks, we may not be able to understand how we are able to determine an object’s essential properties, yet we have good reason for believing that we do (somehow) successfully ascertain the essential properties of material objects. Just as we know the chicken sexer does somehow successfully determine the chick’s sex (and we can test this by examining the chicks at a later stage of development), we have justification (of the kinds outlined above) to believe that we somehow determine material objects’ essential properties.14

But this kind of defense fails to appreciate the full extent of the problem for the realist, assuming, that is, a naturalist based RMO.15 Perhaps, indeed, there may be good warrant for believing that we do somehow access facts about essential properties for objects—that’s what the pragmatic defenses deliver, if they are indeed successful. But notice, however, that this kind of knowledge (somehow accessing facts about objects’ essential properties) is quite different from the chicken sexers’ skill—the two are in principle different for there is some clear empirically identifiable fact that eventually verifies the chicken sexers’ knowledge claim.16 Thus, we have good reason to believe not just that the chicken sexer somehow accesses this knowledge but that they do so empirically and that we are just (as of yet) unaware of how they empirically perceive this fact. That is, there is a determinate and substantive physical fact that we can (eventually) empirically point to explaining the chicken sexers’ knowledge. But in the case of a material object’s mind-independent essential properties, there is no such determinate physical fact. Indeed, it seems there cannot be—and this is the deep problem that should properly be at the core of all epistemic challenges to RMO. And notice that this problem cannot be resolved by appeal to pragmatic arguments for warranted belief in RMO.

Perhaps the realist could respond that the “fact” that makes it true that a rabbit no longer exists after it is flattened by a steam roller is just that: the rabbit is flattened by a steam-roller. That is, when posed with the question of whether or not a rabbit will go on existing after it is flattened by a steam-roller, the realist could suggest we try it and find out.17 But, again, this misses the objection. The question is not if the rabbit continues to exist post-flattening, but why we think it no longer exists. The answer, of course, will be because we think there are certain properties that a rabbit must have if it is to exist at all (such as, say, a body of such and such minimal shape and extensionality), and once those properties are lost, the rabbit no longer exists. So be it. But why are those properties (and not some other properties) essential in this way? And how do we know it or how could we possibly discover it? And there is the rub for the realist.

Consider: there is nothing in the world which is the “truth-maker” for whether or not the object Bob continues to exist after his particles are dispersed across the galaxy. There are many physical facts about those particles and about the aggregate of particles. But there is no physical fact about whether or not the dispersion of particles constitutes Bob’s going out of existence. Assume the following claim X is true: “Bob no longer exists after his particles are dispersed across the galaxy.” But what makes claim X true? Whatever it is, it is not a physical fact, nor is it empirically verifiable.18 And, thus, the chicken sexer analogy does not work for the realist.19 There might be something in virtue of which claim X is true. But how could that something be anything physical? Or, better, how could that something be anything empirically detectable?20 This, then, is Rea’s Revenge and the real force of the DP. It challenges (naturalist committed) realism to not merely give a justificatory account of why we may be warranted to believe RMO (as pragmatic defense attempt to do), but rather to explain how it could even be possible that we could ever empirically come to know such facts about objects.


Crawford Elder, however, thinks the realist can empirically detect some mind-independent essential properties that some objects hold. That is, he offers an account that attempts to explain not only why we should presume realism about material objects (and the correlative essential properties), but actually how we could come to know facts about an object’s persistence and kind-membership conditions. Thus, he attempts to directly answer the true force of the DP and all such epistemic challenges head-on.21

So how could a realist story go for the empirical detection of objects’ essential properties? Elder answers the challenge by first explaining how we empirically ascertain kind-membership properties and then he uses kind-membership discovery to explain how we ascertain persistence conditions.22 To start, he offers an evolutionary account for how we can identify certain kinds as a survival tool, explaining that our knowledge concerning the sorts of properties that individuate natural kinds might have been “hard-wired” into our brains via natural selection. That we were better able to survive and reproduce because our brains were preprogrammed to detect (again and again) nutritious substances and to distinguish them from each other seems plausible enough. Elder argues that we were able to track these sorts that form the basis of kind-membership by finding combinations of detectable properties via smell, taste, and texture that aided us in survival. Moreover, Elder argues, it’s also possible we simply learned which collections and sorts of properties distinguished such substances over a given range. We learn how these properties individuate kinds by observing strong patterns of property clustering in objects across several repeated encounters and, in fact, the pattern clustering is so strong that we then conclude the clustering must not be accidental to these objects. With every repeated encounter (and “success”) we garnered further evidence that the clustering of properties we observed must be causally governed. We deduced this by seeing concurrent “patterns of co-variation” with each clustering of properties. That is, some observed properties within a given cluster contrasted with some other properties in a comparable cluster, and, not only that, but they do this in such a way that there is a pattern to the contrasting uniformities across kinds which can be reliably predicted. The variations of detectable properties across clusters of properties create their own patterns of variance—this flanking uniformity of variations teaches us that some of the properties themselves individuate certain sorts upon which natural kinds are formed. These variant patterns of clustering across objects tell us that the different clusters discriminate kinds of a common family or other. Hence, Elder concludes, we have an empirical process through which we differentiate kind membership for various objects we encounter.

Perhaps this much could be true; for the sake of argument, let us grant Elder’s story about kind detection via co-variation across detectable property clustering. But then Elder moves on to explain how we empirically detect persistence conditions for an object, and here is where he gets into trouble (and any realist account of empirical access to essential properties for objects will have the same trouble). He argues that we possess a “hard-wired” ability to track material objects such that we can perceive short term property configurations of “before and after” in the various states objects within a certain kind can take on. From here we can then apply the same idea for cognitive detection of long term patterns used above for kind-membership and fill in the details of persistence conditions via our knowledge of behavior patterns for the kind to which the object in question belongs.23 That is, Elder builds our empirical detection of an object’s persistence through time off of our initial empirical detection of kinds and correlative kind-membership conditions.


I propose, however, that an object’s persistence through time and an object’s belonging to a kind are themselves separate questions that conceptually “come apart.” In such a case, we could imagine the assertion that detecting the persistence boundaries of an object is epistemologically prior to discovering an object’s membership within a kind. The realist has the task of explaining why they are not forced to endorse such a questionable position—with which, presumably, they’d be rather uncomfortable. We can ask of the realist: why don’t we have to ascertain persistence conditions for Ks before we ascertain which properties reliably characterize Ks—which properties are membership conditions for the kind K? It seems, at least, that detecting the persistence boundaries of an object could be epistemologically prior to discovering an object’s membership within a kind and we have no reason to think or know otherwise.24 Yet Elder’s solution to how we access facts about an object’s persistence conditions depends on our ability to first detect an object’s kind. And this seems impossible; for if we have not yet detected an object’s persistence conditions, then we have not yet detected an object, simpliciter.25 And if we have not yet detected an object, then we can hardly compare that object to other objects to determine its kind membership conditions. But if we cannot compare an object to others for kind membership detection, then (according to Elder’s means of accessing facts about object persistence) we cannot deduce facts about that object’s persistence.26 And thus we’ve gone in a circle. Laid out formally, it is a reductio against Elder’s account:
  1. 1.

    We are able to empirically detect an individual object’s sameness in kind to various other objects via kind-membership indicating properties X, Y, Z. (Elder)

  2. 2.

    We are able to empirically detect an individual object’s persistence through time by identifying it as a member of a particular kind K (because we have knowledge of how members of K behave over time.) (Elder)

  3. 3.

    Epistemologically prior to detecting whether or not an object belongs to a kind K, we must be able to detect that there is an object itself. (Premise)

  4. 4.

    But we can only detect if there is an object before us if we can detect persistence through time. (Elder/Rea)

  5. 5.

    But we can only detect persistence through time for a given object after we have identified its kind membership (from 2).

  6. 6.

    Thus, we must detect persistence epistemologically prior to determining kind-membership and we must detect kind-membership epistemologically prior to detecting persistence. (3 & 5)

1 and 2 fall out of Elder’s account. Both Rea and Elder should (and, it seems do) affirm 4.27 5 and 6 follow from the previous premises. Hence, the most contentious claim would likely be 3.

To shore up some intuitive appeal for 3, then, consider what a process identifying kind-membership would actually entail. We see a rabbit. We might say, “here is a rabbit.” In saying, “here is an F,” where F is a name for any natural kind, it may at first blush appear that our identification of kind-membership just is the same as identifying the persistence rules for the object (such as that rabbits cease to exist when flattened by steamrollers). But it is not. Note that it could be asked by a fellow observer, “what is a rabbit?”28 And the answer would have to be, “this thing [ostensive pointing] is a rabbit.” Elder is correct: placing the “this thing” into kind “rabbit” does then mark out certain empirically obtained essential facts about the object—including certain persistence conditions that have been empirically derived for all objects belonging to kind “rabbit.” But that there is “this thing” with which to first attach a kind to entails that basic (or simple) object persistence (or numerical sameness across time) is prior to the kind-membership conditions for the object.29 It may very well be that once we take an object to be in kind K that fact then entails certain (elaborated or even changed) conditions for persistence. But it seems there had to be some persistence conditions of a something to first be able to place the object under consideration into a kind.30

Or, to show more clearly how this trips up Elder’s account (and I think any suitably empirically based realist account); consider again his explanation of how we discover persistence for an object. The story would begin, “We use highly developed cognitive perceptions to uncover patterns of co-variation across short and long-term detectable property clustering. Once we’ve observed enough about an object to identify it as a member of kind K we can then determine its persistence conditions according to –”31 but here I must interrupt. For what is this “it” that the realist references in order to compare “it” to other objects to place “it” in a particular kind through which we can then ascertain persistence conditions of the given object? Certainly, once he has placed “it” in a kind K, membership to that kind brings with it a variety of particular persistence conditions. But, critically, many (if not most) of the properties that characterize kind-membership are properties the instantiation of which only occurs over time. So until the realist can know that “it” is the same object he is observing, persisting through some period of time, then he cannot tell he’s detecting a kind-identifying property. Simply put, the realist must first pick out an object through time in order to identify it as belonging to a kind. So one must know that they are observing an instance of persistence before they can tell what particular properties characterize the kind to which the persisting object belongs. And how, of course, could the realist (empirically) know that? How could one know that they are observing an instance of persistence “all by itself,” as it were? Without first observing the correlating properties of kind-membership (which occur over time) already built into an object, one cannot know that they are observing persistence for a given object because there is no “given object” at which to point.

Thus, if we don’t first know something about persistence conditions for an object, then it seems reasonable to ask how we would detect membership patterns, and so forth. So Elder (or any realist who thinks persistings are empirically observed rather than stipulated or constructed) cannot use kind-membership to get persistence; for it seems some kind of minimal persistence conditions are a prerequisite for identifying an object at all.

But perhaps a realist will still resist the claim that knowledge about an object’s persistence is epistemically prior to knowledge about its kind membership conditions. That is, one may still wish to deny the move in 3. Certainly Elder would argue against this and Rea himself resists the metaphysical priority of persistence in typical fashion:

Whether a rabbit can survive flattening is one thing; whether it belongs in the kind rabbit is another—and the answer to the first question obviously doesn’t entail any answer to the second. As to which way the dependence goes—persistence conditions depending on kind-membership, or vice versa—I’m not entirely sure. I’m inclined to speak of kind properties ‘giving’ objects their persistence conditions; but I’m also inclined to say (e.g.) that objects are grouped into kinds on the basis of differences in persistence conditions… objects come with essential properties, and those properties determine both kind-membership and persistence conditions.32

Note that in making this claim, however, Rea does admit that the two issues, even if they “come together” via an object’s essential properties, are still conceptually distinct. If this is the case, then my above argument against Elder’s realist story still holds. Rea is arguing that persistence conditions and kind-sameness come at the same “time.” That is, Rea (and presumably Elder and most realists) thinks we can’t have one without the other.33 But if that is so, then it seems my argument, mutatis mutandis, still works against the realist discovery account. That is, we cannot determine if an object belongs to the kind rabbit apart from also determining its persistence conditions; and Elder sets aside the persistence questions to determine the kind questions (which then in turn determine persistence conditions). And this the realist cannot do, or else we need some explanation as to why we would be justified in so doing. Either way—if kind-membership conditions first require persistence conditions; or, as Rea suggests, they come together—the realist cannot determine kind-membership logically prior to or, at least, apart from persistence conditions.34 The challenge to the realist is to explain how we can do so that does not beg the question.

Another potential response the supporter of an Elderian position could appeal to here is a peculiar distinction between “particular objects” (or tokens) versus “object kinds” (or types). The claim would be that the object referenced in 3 and 4 is of the former category. As such, objects qua particulars could possibly have different persistence conditions than objects qua kind-instances do—or at least not have their persistence tied to their kind-membership conditions (as 2 and then 5 demand). That is, objects as particulars may have persistence conditions simply and only as, say, mere spatio-temporal continuants which could be distinct from persistence conditions that persisting members of a specific kind K would have. If that was the case, then the persistence conditions which would need to be presupposed for the detection of kind-membership conditions—the persistence conditions necessary to be able to pick out the object particular in order to do the empirical work Elder thinks we could then do in order to determine its kind-membership conditions—would not be the same persistence conditions specific to a kind K. If this response is feasible, then there is no circle in the empirical detection process Elder laid out.35 I’ll call this the “token object” response.

This reply does not succeed, however, for both Elder’s and Rea’s views of objects do not allow for a distinction between objects qua particulars and qua kind-instances in the way this response proposes. The reason is straight-forward: to be able to apprehend an object before you, you must think it is presently and continuously existing. But to know if something is persisting in its existence, we must know something about what we take it to be—what kind of thing it is—in order to know its essential persistence conditions. This is precisely why Rea, above, took persistence conditions and kind-membership conditions to “come together.” That is, to apprehend something as merely a “token object” that has spatio-temporal continuance through time is not enough to apprehend an object at all. This is because there would be no reason to take any particular set of sense data as consisting as a particular object over any other particular set of sense data.36 What would it mean to take something as an object but yet know nothing about it that could bear on what kind of object it is? Once we are claiming there is an on-going object presently before us that is in itself claiming we know something of what it would mean for this object to go out of existence. And that kind of knowledge, Elder and Rea will both demand, says we know at least something minimally about what kind of object it is. Note, again, that one of the primary reasons for this is that many (if not most) kind-membership conditions are dispositional in nature, and, thus, require observation over time. That is, for any given object, to say something of its persistence—even in this minimal and meager “token” way the response suggests—entails making claims that bear on its kind-membership.

So, because Elder is committed to a view of objects such that persistence and kind-membership cannot come apart, then my reductio above still holds against his view. And, again, Elder is the only philosopher yet to offer a naturalist account as to how the mechanics of empirical knowledge about objects’ essential properties could actually work. That is, his account is the only response to this “Rea’s Revenge” aspect of the DP (how we could know such things, whether or not we are warranted in believing them) currently on offer.

However, there is a view of objects that might at first seem to be of assistance to the realist pursuing the token object response that could provide an escape route from the circuitous reasoning. This is the position recently advanced by Stephen Schwartz that rejects the necessary connection between an object’s existence and its kind-membership conditions.37 Schwartz argues that Elder, Rea, and others holding the “essential membership view” believe that when a thing is for a time a member of kind K and then undergoes a change so that it is no longer a member of kind K, then it ceases to exist. But he holds to the “non-essential membership view” which claims “that there are forms of transformation that ordinary objects undergo that involve change from one natural kind to another without destruction.” (Schwartz 2009, 617.) Schwartz offers the example of a caterpillar that changes into a moth wherein it appears some object x has changed kinds yet it is numerically the same object and has not gone out of existence. If this view of object-hood could work, then persistence need not hinge on kind-membership and perhaps the “token object” escape to the circle can proceed.

As we examine the view more closely, however, the appearance of usefulness for the realist simply dissipates. For, granting for the sake of argument the plausibility of such a position, it does not seem that applying Schwartz’s non-essential membership view would aid the realist in the epistemic mechanics of essential property detection. Following the “token object” response above coupled with Schwartz’s non-essential membership view, it’s not clear just what could count as persistence conditions for an object qua particular that would not have at least some overlap with the persistence conditions specific to belonging to a kind K. Remember, Schwartz’ view argues that persistence need not hinge on kind-membership, but it seems that even on his view kind-membership could still hinge (in at least some respects) on various persistence conditions. That is, if X is observed to persist through time, whatever it is that makes it that specific object X, as opposed to some other mereological composition that we label as X, will be part of the conditions for kind-membership of X. Take Schwartz’s example of a caterpillar turning into a moth. Notice that some persistence conditions for both the object as caterpillar and the object as moth will be subsumed into those properties which Elder says we detect in order to identify kind sameness (the XYZ of 1). That is, there must be something that serves as the “placeholder” for the object through transformation, if it is to have persistence through time as the same object. But whatever it is that makes us think it is the same object, one would think, would also be part of the persistence conditions for a given kind. So, in identifying persistence through time, we’d be playing (at least part of) the game of identifying sameness in kind that Elder thinks we do.

To press this further, let us try conceiving of a token object Z as the response suggests: a mere spatio-temporal continuant. Why have we picked out Z as a continuing object instead of Z plus some other gerrymandered aggregate matter around it as the boundaries for the object Z? The only answer to this mereological mystery seems to be that we’ve drawn the boundaries around Z as we have because of certain assumptions as to what kind of object Z is and how objects of that kind go in and out of existence.

Notice, then, that even if a Schwartzian view of objects could allow for persistence wholly apart from kind-membership, it’s going to be difficult (if not impossible) to apply such a view to an Elderian account of how we empirically come to have knowledge of objects’ essential properties. Again, as noted above, this is because many (if not most) of the properties for kind-membership instantiate over time and Elder uses this feature of those properties to explain how we could detect such essential properties. But applying Schwartz’s view to Elder’s account would change it considerably. One wonders if it could ever get off the ground because “raw” persistence apart from kind-membership would simply not give Elder the tools he claims we’d need in order to detect kind sameness—there would be no observational resources left in such a paltry “token object” that one could use to derive kind-membership. Take again Schwartz’s story about a caterpillar turning into a moth. Yes, this seems like an instance of an object undergoing a transformation, one might say, but notice that the object is still an object of some kind throughout the transformation (an organism, say) and, indeed, an organism that we take to be the kind of object that undergoes such transformations. If, however, the caterpillar turned into a gaseous cloud, or a rock, or a bottle of wine, it would be hard to claim that the cloud, rock, or bottle of wine before us is the same object as the caterpillar. Indeed, we’d say the caterpillar went out of existence. And we’d make such a claim because of the kind of object we take the moth to be. But, by adopting a Schwartzian non-essential membership view of object-hood which allows us to detach persistence entirely from kind-membership, we could propose that the gaseous cloud, rock, or bottle of wine before us is the same object persisting through time as the caterpillar. So be it. But now we are not going to be able to derive any kind-membership conditions from observation of such an “object.” And, as noted above in discussing why Elder and Rea both reject such a view of object-hood, why not apprehend the gaseous cloud plus some other gerrymandered parts of spatio-temporal continuants before us as one object? The answer must be that we apprehend the objects persisting before us as the objects we do because of the kind of objects we take them to be.

So, indeed, a Schwartzian view of object-hood (if such a view is even plausible) may allow for a technical escape to the circle created by the relationship between persistence and kind-membership, but it does so at a high cost. For now the naturalist account Elder offers can no longer explain how the mechanics of empirical detection of objects’ essential properties is supposed to work. And so the challenge of Rea’s Revenge still stands, unless a rival account can be offered that weds Schwartz’s non-essential membership view of objects with some new account of how we empirically come to have knowledge of objects’ essential properties. But it is, to my lights, unclear how this could be done.


I’ve argued that the DP has a second challenge which remains regardless of whether pragmatic (or other) arguments can justify warranted belief in RMO. This challenge (Rea’s Revenge) is the obstacle of explaining how we could empirically discover objects’ properties which are mind-independently essential.38 Elder proposes a potential method: first, we empirically determine whether or not object X belongs to the kind rabbit (through recognizing variant property clustering patterns across objects). Then, through our knowledge of kind-membership conditions, we can empirically determine whether or not a particular rabbit can survive flattening, because we can ask whether not things of that kind can survive flattening. But I argue this is presently unjustified and, for all we know, it may be that he gets this backwards. Epistemologically before we can determine if an object belongs to the kind rabbit, we must first determine its persistence conditions—for if we do not, then we do not (yet) have an object in our conceptual framework to determine to what kind it belongs. That is, without prior persistence, there is for us no object of which we can detect essential kind-membership properties in the first place. Unlike Schwartz’s view, I do not suggest that persistence is metaphysically separable from kind-membership, but rather I aim to show only that this relationship between the two conditions (once they are viewed as conceptually distinct) is an apparently insurmountable epistemic problem for any realist story replying to Rea’s Revenge. This is because we seem to need both persistence conditions and kind-membership conditions to be able to ascertain either of the two for an object. So to get our apprehension of objects’ essential properties off the ground, it seems we have to assume one to get the other (as Elder does), and this we cannot do because it results in circuitous reasoning. In other words, to be able to know if a thing before us goes on persisting through time, we must know what kind of thing it is. And, to be able to know what kind of thing there is before us, we must first know that there is an object persisting through time before us at all. So we are stuck. And, hence, Elder has not given us a successful naturalist explanation of how we could access persistence conditions apart from accessing kind-membership conditions. This means there is at present no naturalist-committed realist accounting for how we can empirically detect material objects’ properties which are mind-independently essential. Until such an account is made, the heart of the epistemic objection against realism remains.


It is a deeply held association for most realists, but, to be clear, it is not universally agreed that an object can’t depart from its kind without ceasing to exist—it would depend on which kinds we’re talking about. And some will point out that, likewise, persistence conditions are not invariably necessary properties: provided object O in fact belongs to K, O cannot go on existing without continuing to belong to K, but, for all that, O could have belonged to a kind other than K. (On this point, see Penelope Mackie 2006 and 2008) So be it. I am not taking a stand on either side of this particular point; but rather that for most realists about material objects such an association of the kind I am attacking is central and standard practice.


Elder, for example, does precisely this (and it is the clearest exposition of the move). The general idea is found in several places throughout Elder’s work. The theme is discussed in chapter 2 of Elder 2004 and in “Carving up a Reality in which There Are No Joints,” in Steven D. Hales, ed., A Companion to Relativism. (Oxford: Blackwell, 2010a); but it is laid out most clearly in Familiar Objects and their Shadows (Cambridge University Press: 2010b.) The project is discussed tangentially in Elder 2007 and 2006.


Rea calls such properties “intrinsic” as opposed to “extrinsic.” So, biological kinds which many (including Elder) view as having external relational properties which define their kind membership would have extrinsic essential kind-membership properties (because a biological individual’s essential properties derive from their origin, which is derived from some other biological individual); whereas water having the property of “being essentially H2O” would be intrinsic.


And this is precisely what Rea argues for: that only non-naturalist versions of RMO can escape the DP. See Chapters 8 and 9 of Rea 2002 for some proposals on how such versions could work.


And, henceforth, my use of the term “realist” can be taken as shorthand for “realist committed to naturalism.”


As Sidelle writes, “I shall argue that our methods can be understood to give us modal knowledge only if the ‘modal structure of the world’ is closely tied to our conventions; our methods cannot give us knowledge of a completely independent reality, but only of the limits of our conceptual scheme.

Thus, we must reckon the necessary truths of which we actually have knowledge as dependent on our conventions for their necessity,” (Sidelle 1989, 87). See also Alan Sidelle 1998, 423–48. Amie Thomasson is a conventionalist who provides a strong attack against the realist that does not rely on the epistemic challenge. See Thomasson 2007.


Note: I am trying to avoid the assumption that objects not having mind-independent essential properties necessarily entails that those objects do not mind-independently exist. Some contest this entailment, and so I am focusing merely on the objects having mind-independent essential properties.


See Chp. 6 “Pragmatic Arguments” of Rea 2002. Also see Elder 2004 who offers his own pragmatic arguments for warranted belief in RMO throughout the work.


Thanks to Alexis Elder (no relation to Crawford Elder) for pointing out just how many resources are available to the realist for RMO justification.


In Rea’s case he thinks that these pragmatic arguments entail further metaphysical commitments that go beyond the methods available to naturalism.


And for Rea there is the further mistake of trying to focus his DP on pragmatic defenses for the naturalist in particular. It does not seem that the DP establishes the epistemic pressure against naturalism’s employment of pragmatic arguments (as opposed to non-naturalist employment of them) that Rea wants it to, but I will not argue this here.


There is, of course, corresponding differences at the micro-biological level of the chicks, but there are no obvious surface level detectable corresponding differences—or, at least, there appears not to be any.


The skill possessed by these chicken sexers is of great interest to cognitive scientists for this very reason. See Richard Horsey, “The art of chicken sexing.” (, accessed March 1st, 2009.) Much more here deserves to be said. We could look at stories of how highly successful poker players “just know” or “have a feeling” as to how certain cards are going to play out in a given hand, as another example. Assuming we do not believe that they nor the chicken sexers have some kind of “magic” ability, we assume, rightly, that they have some empirical access to the knowledge in question—but that we simply do not yet understand how their access works.


To clarify, presumably by “determine” we mean something like “learn” or “detect” these properties, in which case we are taking a realist stance. The conventionalist might agree with the claim that we somehow “determine” the object’s properties, but mean by it to “fix,” “stipulate,” or “structure,” them. But clearly in the chicken sexer case the determination of gender is a discovery, and so, the realist aims to use this kind of defense in a similar manner. I’ll show below precisely why the defense does not work; it is due in large part because the “determination” in the essential property case cannot be so easily assumed to be an empirical discovery as the chicken sexer case can be.


Again, Rea holds to RMO but thinks it is not compatible with naturalism and thus focuses his attack against any naturalist realist. Sidelle thinks RMO is unjustifiable period.


In this case it is the sex organs of the chickens that develop so as to be easily visible at later stages.


Many thanks to Jesse Mulder for the suggestion of this possible response from a realist on this point.


The realist might protest and here claim that we can learn that certain properties essential to Bob go away once Bob’s particles are dispersed and that this fact (that those properties have gone away) is empirically verifiable—that is, those properties existence is a fact in the world. Certainly, I agree. But the fact that those properties’ are essential to Bob is not a physical fact anywhere to be found in the mind-independent world.


Again, if one says it is simply analytic that the claim is true, they solve the epistemic challenge, but they are no longer realists about material objects, but instead have opted for some version of conventionalism. Notice in this move there is also no truth-maker for the claim—there is literally nothing (in the mind-independent world) that makes it true. Or, if one takes an even harder-line approach it is we who make it true—our mind in carving up reality—in which case clearly the objects’ essential properties are not mind-independent.


Note: I colloquially use the “truth-maker” locution not as an endorsement of truth-maker alethic theories but as a helpful moniker to point to a lack of any empirically detectable fact that makes claim X true. Below we’ll see Elder’s attempt to give us an empirical story for the realist, but I’ll argue that it fails for internal reasons regarding the relationship between persistence conditions and kind-membership conditions.


And to my knowledge he is the only contemporary philosopher to make such an attempt.


The story I outline below is cobbled together from a variety of Elder’s works, most specifically from Familiar Objects and Their Shadows (2010b).


As Elder writes: “More sophisticated cognitive activities enable us to detect patterns that are more long-term: we observe enough about an object to identify it as a member of kind K, add to it knowledge about how members of K move (or do not move), sometimes together with partial knowledge as to how members of K alter qualitatively over time, and arrive at the knowledge that this member of K is the very member we earlier observed.” (Elder, Familiar Objects and Their Shadows, 2010b.)


And, as I’ll note below, if persistence is not prior to kind-membership then it is at least simultaneous with kind-membership. In either case, kind-membership cannot be used to establish persistence, for kind-membership conditions first (or simultaneously) require persistence conditions.


Rea points out this aspect of persistence early on and is right that, as he writes, “it is a conceptual truth that material objects have persistence conditions...” (Rea 2002, 83.) That is, to even consider an object, you must assume some basic persistence conditions in order to be able to pick out an object. (Later it appears Rea might be inconsistent on this point when he speaks freely about an object in question “prior” to our picking out its persistence conditions.)


This will be elaborated below, but one of the primary reasons for this is that many (if not most) of the kind-membership conditions we detect are dispositional in nature, and, thus, require observation over time.


Rea clearly affirms this (see note 25 above) and Elder affirms something like the necessity of persistence in a variety of places cited in this paper. Below I’ll look at what the outcome for my argument would be on a view that rejects the necessary connection between even an object’s existence and its kind-membership conditions in considering Stephen Schwartz’s recent position as elucidated in, “The Essence of Essence,” (2009) Australasian Journal of Philosophy.


Nota bene: The observer is not asking what the term rabbit means but asking to which object it is being applied. There is a semantic argument occasionally given for thinking that realists are mistaken when they claim that we empirically learn what the persistence conditions for Fs are. It goes like this: “We cannot be talking or thinking about Fs at all, unless we first have an accurate understanding of what the properties of Fs are—and, particularly, what the persistence conditions for Fs are.” This relies on a deeply internalist view of semantics. Realists will instead endorse an externalist picture. My argument in this paper in no way relies on this semantic claim. Rather, I am arguing over how we can discover essential properties of objects and setting aside how the semantics would occur. (Many thanks to Crawford Elder and Donald Joy for help on this point.)


Note that I am not suggesting that persistence and numerical sameness across time are necessarily the same concept. Persistence is usually understood as certain facts about what an object can and cannot survive, whereas numerical sameness across time is simply our ability or inability to track the same object from one moment to the next. I am arguing that once we place an object in kind K we then have more thoroughly elaborated persistence conditions for object X. But there exists an object X to place into kind K prior to its placement therein. This object must have some persistence—at least insofar as we can even conceive of it as there in front of us. This “simple persistence” may just be the same thing, then, as numerical sameness across time for this object (logically) prior to our placing it in a kind K. Elder distinguishes “numerical sameness across time” and “sameness in kind” for an object in Familiar Objects and Their Shadows. But, to use his locution, I argue that he misses as a matter of logical priority that (it seems) sameness in kind for an object cannot be accessed until numerical sameness across time for an object is first accessed; or at least we should ask the question of if it could be. Perhaps it can, but there seems to be no good reason to assume so.


For certainly, if we start to enumerate properties and property changes, over the short-term and long-term (as Elder claims we do), we already assume that there is an object existing over which these other properties can take hold. As Rea puts nicely: “If there are no facts at all about what sorts of changes a putative thing X can and cannot survive, then there is no such thing as X.” (Rea 2002, 82.) (Note: All uses of “first” and other temporally loaded terms are, of course, to be understood as purely logical or epistemic priority ordering; not actual temporal event sequences.)


This is my paraphrase of the kinds of explanation a realist might give. Elder writes something very similar in Familiar Objects and Their Shadows, 2010b. Emphasis mine.


Rea, personal correspondence, March 1st, 2009. Emphasis Rea’s.


In an earlier work (that was elaborated upon and developed in World Without Design) we find some of the roots of Rea’s position here, “Kind concepts are the tools by which we forge the link between essential properties and ways of arranging matter. It is traditionally assumed that there is both a metaphysical and an epistemological connection between kind-membership and persistence conditions… persistence conditions are grounded in essential properties, there are also such connections between kind-membership and essential properties.” (Rea 2000, 115.).


In correspondence with Rea on this point, he agrees that my argument will hold in either case. In the symposium to Rea’s World Without Design in Philo, Andrew Melnyk argues that there’s a general mistake made by Rea in connecting persistence to essential properties, but it does not concern the epistemic priority relationship as my challenge lays out here. See Melnyk 2005 and Rea 2005 for discussion.


Thanks to an anonymous Philosophia reviewer for this potential response.


Notice, again, Sidelle and the conventionalists would here say that we can carve an object out of sense data—and that this is precisely what we do. Such carving, wherein we impose essential properties onto objects in the natural world, is not an option, of course, for the realist.


See S. Schwartz 2009.


Again, RMO not committed to naturalism has some potential explanations for how this could be done because the mechanism need not be empirical. See note 4 above.



I am deeply indebted to Crawford Elder for extensive help and advice on this paper. Thanks are also due to Michael Rea for helpful conversations and permission to quote our personal correspondence. Finally, I am grateful to Jesse Mulder, Donald Joy, Micah Newman, Alexis Elder, Levente Szentkirályi, and Abbilynn Strawser for helpful conversations, editorial advice, and reviews of this paper.

Copyright information

© Springer Science+Business Media B.V. 2010