Howard Robinson and I both find it problematic to identify a person with a vague object. We agree that all the sensible physical candidates for being a person are vague, and we do not shrink back from the radical conclusion: that we are immaterial thinking things. Although the arguments we give lead from vagueness to immateriality, they are very different in strategy.¹

Part II of Robinson’s From the Knowledge Argument to Mental Substance (2016) contains two main paths that lead from the vagueness of physical objects to the conclusion that I am not one. One important argument has to do with the indeterminacy of identity for physical objects under contrary-to-fact conditions, and the (alleged) determinacy of identity for minds under such conditions. I shall ignore this intriguing argument (found in Chapter 12), and focus on a different one. It is based on considerations that are spread among several chapters; putting the pieces together will take some time and effort. Much as I would welcome additional support for my own conclusion, I discover several plausible ways for materialists to resist Robinson’s argument. Unless I am missing something (which I might be—the book is dense, the arguments complex), materialism faces little danger from this quarter.

I am a great admirer of Robinson’s work, this book included. Here, however, I focus entirely on the part of the book to which I have the most objections. I show my admiration in classic schoolyard fashion—by punching him on the arm as hard as I can, and then running away.

1 Fundamentality, Nonfundamentality, and Vagueness

Robinson does many things in Part II of From the Knowledge Argument to Mental Substance. One of the main threads running through its chapters is support for the premises of the following argument, which shall be my focus: All physical objects (or at least all the ones that are any kind of candidate to be a human being) are “non-fundamental”; they do not belong to the “fundamental level” of reality. Non-fundamental things are inevitably vague in various ways; they are vague in their boundaries, and many of their characteristic properties are vague as well. This vagueness, he says, shows that we must “make a conceptual interpretation of them,” treating them as “artefacts of conceptualisation.” And this in turn precludes our identifying ourselves with any such things. Robinson’s more exact formulations of this argument will be considered shortly. First, I try to clarify what Robinson means by “fundamental” and “fundamental level”; what it is for a thing to be an “artefact of conceptualisation,” or for us to “make a conceptual interpretation of” something; and how vagueness is connected with these two ideas.

Robinson talks of “levels” and “ontologies,” with different ontologies located at or constituting different levels. There is a “fundamental physical level” which he calls “basic physics” (leaving open what form that might take), and non-basic levels, some of which are the subjects of the special sciences.² This talk of levels is flexible; the term “level” (and “ontology”) is open to (at least) two interpretations: it can refer either to a theory (in which case it is a “representational ontology”; that is, “a conceptual picture of the world”) or to the entities that are the subject matter of the theory (Robinson 2016, 168). In the latter use, “the ontology of the theory” simply refers to the ontological commitments of the theory—the things that would have to exist, were the theory true.

Robinson is using “level” in the first way when physics is described as one level, and the special sciences (biology, geology, meteorology, etc.) are said to represent higher levels (see e.g. 2016, 156–157, 220). Different levels will use a different terminology; higher level theories introduce terms not found at lower levels. The concepts these terms express, and the properties to which they refer, can be called “higher level” as well. The most fundamental theory will include terms for the most fundamental or basic properties. The properties of interest to the special sciences will be less fundamental, less basic, than the properties of interest to physics.

Robinson uses “level” in the second, ontological sense when he asks about “the causal efficacy of the non-fundamental levels” (2016, 181), and whether “at the fundamental level,” there are just events in space-time, or enduring objects (2016, 208). If physicalism is true, the level of the “basic ontology” consists only of electrons, quarks, or whatever entities are the subject matter of fundamental physics; the level of biology comprises all the organisms and (at least some of) their parts; the level of geology includes boulders and lava flows, etc.

Although Robinson does not explicitly invoke David Lewis’s notion of a “natural property,” his views about fundamentality of levels, and the association of vagueness with higher levels, can be fairly explicated in terms of naturalness. Lewis uses natural properties for many purposes, but I invoke them here only in their role as resemblance-makers.³

Plato introduced the metaphor of “carving nature at the joints.” Natural properties are posited as the joint-carving ones; each natural property represents a respect in which things can objectively resemble one another. But naturalness is not all-or-nothing. Schemes of classification, and the terms used in scientific theories, may be more or less natural. For example, all mammals resemble one another in certain respects, so being a mammal ensures some degree of objective similarity. But it does not ensure exact similarity with respect to any one precise feature. Being a mammal is much less natural than having a certain height or weight.

Degrees of naturalness suggest the possibility (some would say the necessity) of a lowest level consisting of the most natural properties—what Lewis called the “perfectly natural” properties.⁴ A property is perfectly natural if, and only if, it is responsible for one of the most basic respects in which things can objectively resemble one another—the respects of resemblance that “comprise a minimal basis for characterizing the world completely” (Lewis 1983, 346).

There is a clear connection between vagueness and naturalness. At least some vague terms, such as “bald” or “tall,” are used to ascribe a degree of objective similarity among the things to which they are applied. So there are vague natural properties. But vague terms cannot represent perfectly natural properties. When things resemble one another in virtue of sharing a vague property, they resemble in a less-than-perfectly-precise way. The degree of similarity imposed by the resemblance must, then, be less than perfect.

2 Fundamental Objects, Garden Variety Objects, and “Artefacts of Conceptualisation”

Robinson’s fundamental level, or “basic ontology,” conceived of as a set of entities, consists of the “basic constituents of the world, not constituted by anything else” (2016, 168). Robinson’s “basic constituents” are in the same line of work as Joshua Brown’s “perfectly natural objects.” Brown defines perfectly natural objects as all and only those that possess perfectly natural properties (2016, 260). This raises an interesting question for Brown and Robinson: is it true, as Brown assumes, that no composites—nothing that is constituted by further things—can have perfectly natural properties?

All objects, including the perfectly natural ones, will clearly have some less than perfectly natural properties. Suppose that electrons are perfectly natural objects, and that the precise mass and charge of an electron are perfectly natural properties. An electron also has the property of being less than one kilogram, and the property of being negatively charged—properties that clearly do not ensure a precise degree of resemblance among the things that share them. Since electrons have them, it cannot be that perfectly natural objects have only perfectly natural properties. But Brown takes it to be at least likely that no composites have any perfectly natural properties (2016, 265); and, in particular, that organisms do not have any perfectly natural properties (2016, 259–260). If that is right, and similar things can be said for entities located at all the other levels that Robinson regards as higher (which includes all composites), then Brown’s category of perfectly natural objects lines up nicely with Robinson’s category of fundamental or basic things.

There are puzzles for Brown’s proposal. Are there really no composites with perfectly natural properties? If so, the net mass of a composite, for example, must be less natural than the mass properties of the simple particles that make it up (Brown 2016, 259–260). Alternatively, one might allow for some perfectly natural composites (e.g. the universe as a whole, which may well have a precise finite mass), while denying that the objects of study in the (intuitively) higher-level special sciences have perfectly precise masses and other basic physical properties.⁵ I prefer this more liberal approach: perfectly natural composite physical substances may well exist, but the kinds of things that Robinson regards as higher level entities—most saliently, human bodies and brains—are all vague, and do not have perfectly natural physical properties due to their vagueness.

Robinson will go on to argue that a person must be a perfectly natural object, a conclusion which provides considerable support for substance dualism. After all, the only really plausible physical candidates for being me are not perfectly natural. They are what I elsewhere call “garden variety objects” (GVOs): that is, material objects with spatial boundaries that are defined in terms of detectable physical discontinuities and functional roles that are significant (to us) (Zimmerman 2010, 136–137). To be a reasonable candidate, a GVO should at least include the brain or most of the brain, since it is the organ upon which our minds most directly depend. So, a brain, a nervous system, an entire organism, and perhaps even just one hemisphere of the brain—each is a decent candidate for being me, if I am a GVO. These things are reasonably well-demarcated in terms of physical discontinuity with surrounding matter, and functional unity; and they all include (all or at least half of) my brain. And all such objects are vague in their boundaries, and will not have precise masses, shapes, electrical charges, locations, or any other perfectly natural physical properties.

It is conceivable that I be a physical object that is not of the garden variety. Suppose there were some sort of special physical particle in my brain—either unlike all the others or uniquely located in a physical Cartesian theater (a venue designed for social distancing, with but a single seat). Discovering this particle might make me wonder whether I was, in fact, identical with it, rather than with some larger material object. On the tiny-particle hypothesis, I would be a material object, but I would not be a GVO.⁶ However, we have good reason to doubt the existence of such unique physical things; the brain is made of the same gigantic numbers of a few kinds of fundamental particles as all the other pieces of “middle-sized dry goods” that surround us, and there is no central theater in which one particle could occupy a privileged place. If I am to regard myself as a material object, it had better be a GVO. I will restrict attention to just the most obvious candidate GVOs: namely, brains and entire human organisms (human bodies).

Such things are, Robinson says, non-fundamental, and they have non-fundamental properties. The properties figuring in the special sciences—most relevantly, biology and the human sciences—are vague and therefore far from perfectly natural. And he argues that, since these properties fail to be reducible to non-disjunctive more natural properties, non-fundamental objects and properties must be “perspectival”:

[…] the special sciences are best understood as different perspectives on the physical base, usually with certain interests in mind. They are essentially in the same category as patterns, because, though the concepts they involve are well grounded by the basis physical reality, they do not reflect any reality additional to [the] fundamental physical base, except the interests and other perspectives of the humans who employ them. (2016, 220)

Less-than-perfectly-natural kinds (including brain and organism) and their higher-level properties (their shape, size, chemical make-up, biological properties, and so on) are vague in ways that generate sorites paradoxes. According to Robinson, this is a sign that things falling under such categories and characterized by such properties should not be taken to exist in a “fully realist sense” (2016, 174). The contrast is with existing in a merely “conceptualist” sense. To say that a brain or organism exists in a merely conceptualist sense is to say two things: (i) “[T]he world [is] so organized that it satisfies this concept,” which merely means that paradigmatic cases of someone’s identifying a particular brain or human body manage to track something about the world. We are not just confused when we apply these concepts in some cases but not others. (ii) Nevertheless, “[i]f there were no conceptualisers around (putting God or Divine minds aside),” there would be no brains or organisms or other vague objects—they are mind-dependent “artefacts of conceptualisation” (Robinson 2016, 179). This is what Robinson means by “making a conceptual interpretation” or “making a CI” of a brain, organism, or other object (2016, 178)—it is to affirm their mind-dependence. To say that brains or organisms would be around, with or without conceptualizers, is to give them (or their existence, or their special science properties) a “realist interpretation.”

3 Robinson’s Regress Argument

The pieces are in place, then, for Robinson’s argument that we are not GVOs. He sketches the argument at the end of Chapter 9 (2016, 159) and in Chapter 11 (2016, 179), referring the reader primarily to Chapter 13 (a criticism of Dennett) and the overlapping essay, “Quality, Thought and Consciousness” (2010). (The relevant material in the essay is included in the book, so I will refer just to the book.)

Much of Chapter 13 is aimed specifically at Dennett’s instrumentalism about intentionality. Robinson develops a regress argument against Dennett’s instrumentalism which he eventually extends to reach the conclusion that minds are not composite objects—recall that Robinson, like Brown, takes all composite objects to be problematically vague, like the things I call GVOs; he therefore assumes they all require a “conceptualist interpretation.” As I noted, I am reluctant to say that absolutely no composites can have perfectly natural properties; perhaps the level of “basic physics” (whatever that turns out to be) includes some large things with parts. Nevertheless, the dualist in me would be excited enough by a successful regress argument that rules out all GVOs as candidates for being thinking things. Once they are eliminated, all the alternative candidate physical objects are highly problematic. So I shall treat Robinson’s argument as targeting just GVOs.

The pithiest statement of Robinson’s regress argument for dualism is this:

[I]f all physical composites are artefacts of conceptualisation, and if the human being, brain, mind etc. are physical composites (and they are certainly not physical simples), then they are products of conceptualisation. What is it that does this conceptualising? Not something that only exists conceptually, on pain of a regress. (2016, 179)

Disentangling the argument from the assumption that all physical composites are infected with vagueness, I shall construe it as taking the following form:

Granting that all GVOs are vague, the remaining premises are 2 and 4. In the remainder of the paper, I shall examine the reasons Robinson gives for accepting these two premises. As shall appear, there is much that materialists can say against them.

4 Support for Premise 2

According to Robinson, physicalists must regard not just mental states but all special science properties as “perspectival”—“different perspectives on the physical base, usually with certain interests in mind […] This is a form of interpretationalism, which presupposes a mind picking out the fundamenta that make the higher-order explanations possible.” And so physicalists “cannot avoid assigning an irreducible role to the mind in the creation of the non-basic physical levels” (Robinson 2016, 220). In other words, there would be no GVOs without minds to conceive of them.

This is premise 2, and its plausibility depends upon at least two theses: (a) non-fundamental truths would not be true, were there no minds taking up the perspective required to understand them; and (b) were there no creatures with our perspectives there would be no other minds (e.g. no divine mind) capable of understanding them. If either (a) or (b) is false, premise 2 is in big trouble. I consider each in turn.

In his support of (a), Robinson allows that higher-level predicates (like those of the special sciences) can be appropriately attributed to objects because of “real patterns” that exist “out there.”⁷ But these “non-basic predicates,” because of their vagueness, should be treated in a “conceptualist, rather than a realist, manner. Only basic predicates, and those reducible to basic predicates should be treated in a strictly realist way” (Robinson 2016, 175). If special science (or other higher-level) descriptive terms are not reducible to fundamental physics, they correspond to categories that are picked out from our perspective; which “seems to give the interpreting mind an irreducible role in the creation of these sciences” (2016, 158). This implies, in turn, that the objects that display these patterns only exist “out there” because “they are reified as being of a certain kind by an interpretative act” (2016, 158).

The obvious objection to this line of thought is simple and highly intuitive (and reminiscent of G. E. Moore’s “Refutation of Idealism” 1903): Granted, minds are needed if these “patterns” are to be noticed. But the patterns themselves could exist whether or not they are noticed. The creation of a science—a scientific discipline, with its textbooks and methods—may be impossible without minds. But why think the entities and properties described by the science could not exist without minds (unless the subject matter includes minds)? In other words, why must one “make a conceptualist interpretation” of these entities and properties?

Robinson’s reasons for thinking “that one should treat non-basic predicates in a conceptualist, rather than a realist, manner” (2016, 175) are spread throughout the second half of the book, and they are interrelated in complicated ways. I am not sure that I have fully disentangled them, or seen all the connections among them, but I can discern five distinct lines of argument. (1) The vagueness of the non-basic levels is supposed to show that they are implicitly describing the world in ways that require the existence of concept users. (2) The proper understanding of the autonomy of the special sciences should lead us to interpret biological kinds, for instance, as anthropocentric in ways that require “making a CI” of them (2016, 186–189). (3) The special (physical) sciences are alleged to describe the world in Newtonian ways that are strictly false; and this requires interpreting the terms and ontologies of these sciences conceptualistically (2016, 190). (4) Conceptualism is justified by the fact that it solves problems of constitution (e.g. difficult questions about the relationship between a statue and the clay that constitutes it) and Unger’s problem of the many (2016, 179–180, 190–191). And (5) a sparse theory of universals, such as David Armstrong’s, only treats the basic predicates in a realist way, all others being “understood in the conceptualist way,” which implies making a CI of non-basic things (2016, 175).

Of these, (1) and (5) strike me as the strongest, and I will devote the next two sections to them. As I mention, briefly, at the end of section 5, (2) seems to me to have the strengths and weaknesses of the argument from vagueness; so my response to (1) provides a response to (2). I shall not say much about (3), and (4). In these arguments, Robinson advocates a conceptualist interpretation of certain ways of talking about the world because they seem to be not strictly true. But suppose he is right: chemistry somehow ascribes Newtonian properties to things, descriptions of sculptures imply that the statue and clay are distinct, and platitudes about cats imply that there are many cats right where Tibbles is located. Robinson’s strategy in all these cases is to deny that some higher-level statements are true, though they may be apt or appropriate for certain purposes. But that move is open to anyone bold enough to make it, and does not require the conceptual dependence of the subject matter.⁸ So I find these arguments much less convincing than the more straightforward claim that, since vague language is really implicitly about us, we must “make a CI” of higher-level theories and ontologies.

5 Vagueness Requires “Making a Conceptual Interpretation” of GVOs

I take (1) to be the strongest of the five strategies. Robinson uses the paradoxes of vagueness to forge a link between non-fundamentality and “making a conceptual interpretation” of something. When paradoxes arise, he says our standard practice is to discard non-fundamental talk for more fundamental talk. This practice is supposed to be a sign that vague terms are implicitly about our willingness to ascribe them. Here is a very brief summary of the long, intricate argument of Chapter 10.

How should one respond to the apparent non-bivalence of claims made in vague language, or to the sorites paradoxes generated by vagueness? Not by adopting a non-classical logic, nor by epistemicism, nor by supervaluationism. Look instead to what we actually do: we use normal, bivalent logic with vague terms

until the vagueness becomes salient; then we either contrivedly precisify them for the present purpose, or move to another discourse that is not vague under the relevant circumstances. […] When the vagueness intervenes, the discourse is either modified or suspended, so that normal logic can once again be deployed. (Robinson 2016, 171)

In other words, vague talk is sometimes inadequate to the expression of the facts, in which case it is thrown out in favor of something more precise, and therefore closer to the basic level.

This practice shows that natural languages do not constitute what Robinson calls a “Logical Unity.” The inferential relations among all the propositions expressible in English, for example, cannot be captured in a single formal system because “there appears to be no canonical way of representing the logic of vague predicates” (Robinson 2016, 165).⁹ The truly basic level is presumed to admit of “a characterization […] which is free of inconsistency and which can be regimented according to some canonical form” (and “one might hope that this can be a classical two-valued logic”) (2016, 166). Replacing vague terms with more precise language (because of looming paradox or failures of bivalence) is interpreted by Robinson as admission that vague talk is “not to be taken as realistically” as statements that could be made in the perfectly precise language of the basic ontology. When we speak truly using vague terms, we offer “an ontologically sketchy way of seeing the world”—it may be true, but is at best “a view, an appearance, a kind of secondary quality of the underlying reality” (2016, 168).

Why exactly does Robinson think that the “ontological sketchiness” of the language used to describe some subject matter requires that we give descriptions in that language a “conceptualist” interpretation? Why, that is, should the vagueness of the language lead us to conclude that the propositions expressed using that language are really propositions about the way we view things—therefore implying that persons exist? I find only one explicit source of support for this connection: it is in his treatment of sorites paradoxes, like the paradox of the heap. In order to resist reaching the noxious conclusion at the end of a sorites argument (that a single grain is a heap), one should understand every occurrence in the argument of sentences like “n grains constitute a heap” as meaning the following: “n grains can properly be seen or conceptualized as a heap” (Robinson 2016, 174). Robinson then provides a way of resisting the argument for the conclusion that one grain constitutes a heap, and his strategy turns upon substitution of this psychological description in place of “heap.”

If all vague predicates were, implicitly, about how human beings view things (and if satisfying such predicates implied that some human being exists—an important caveat), then the connection between the vagueness of higher levels and conceptualism about those levels would be reasonably clear. If “heap” means or is otherwise equivalent to “a thing someone conceptualizes as a heap,” then there would be no heaps unless there were minds capable of applying concepts. As a meaning equivalence, this does not seem very plausible; “that is a heap” does not seem, even implicitly, to be a statement about concept-users, and “heaps of sand exist that no one conceptualizes as heaps” has the ring of truth.¹⁰ Moore’s anti-idealist response seems perfectly reasonable here.

It is not even clear that, given Robinson’s proposed meaning for “heap,” there could be no heaps without concept users. If “that is a heap” means “that is something that could properly be seen or conceptualized as a heap (were there creatures like us around to do so),” then heaps exist in worlds without concept users; one need not “make a CI” of heaps. For this meaning equivalence to imply the mind-dependence of heaps, the truth of a statement of the form “such-and-such can properly be conceptualized as a so-and-so” must require the actual existence of someone, somewhere, with the mental equipment to apply the concept of a so-and-so. In other words, the “can” here must not be interpreted in a highly abstract, “in principle” way; this form of words does not mean “were there, perhaps per impossibile, someone around to contemplate the such-and-such, they ought to (or at least are not obliged not to) conceptualize it as a so-and-so.”

For the proposed meaning equivalence to show that vague terms imply the existence of concept users, Robinson would have to assume that “can,” when used in his definitions of vague terms, satisfies a principle along these lines: if it is true, in a possible world, that a such-and-such can be interpreted as a so-and-so, there must be someone who exists in the world in question and who (in some world or other) conceptualizes a such-and-such as a so-and-so. Applying this principle to heaps, one can conclude that it is false that there would have been heaps of minerals and clouds of gas in any universe utterly hostile to living things or other concept users, even if some matter there is arranged in such a way that we would not hesitate to describe it as a heap or cloud.

Although this may sound like the result Robinson wants, it is not quite enough for the conclusion that GVOs like heaps and clouds cannot exist without minds. Perhaps things which actually are heaps could exist without being heaps. That might seem to violate the appealing principle that “heap” is a sortal term, and that such terms pick out the essential kinds to which things belong. But the assumption that “heap” corresponds to an essential sortal kind is not so obvious when one realizes that it is (according to Robinson) a very extrinsic predicate, encoding facts about how human beings would react to something. Maybe “electron” corresponds to an essential kind, but “electron humans may someday detect” does not—some electrons that we may someday detect also exist in worlds without humans, in which they do not satisfy this description. On Robinson’s proposal, “heap” is more similar to “electron humans may someday detect” than it is to “electron.” This provides reason to think that, if Robinson is right, a thing that is actually a heap could exist in worlds without concept users—it just would not then be a heap, though it would be intrinsically just as it is in the actual world.

But how plausible is the original claim about meaning equivalence? What is wrong with saying instead that “is a heap” is equivalent to the alternative I suggested above: “is something that could (perhaps per impossibile) be conceptualized as a heap, were there someone around to contemplate the question”? Robinson’s strategy for resisting the sorites for heaps does not, so far as I can see, depend upon his proposed meaning equivalence as opposed to this alternative suggestion. He points out that “it is a psychological matter whether making some small change affects a subject’s inclination to classify an object in a certain way” (2016, 174). He then describes what would happen in any concrete case involving a single subject being asked repeatedly whether something is a heap, while grains are removed one by one: over the course of many removals, confidence in calling the result a heap will decrease, until at some point the subject will refuse to apply the term. This is meant to falsify one of the crucial premises in a sorites: that removal of a single grain can never make a difference to whether a thing is a heap. But everything he says here could just as well be said if “heap” is understood as “could be seen as a heap.” And the latter interpretation has this going for it: it does not lead to the counterintuitive conclusion that heaps and clouds cannot exist without minds to apply concepts to them.

To sum up then: (1) has not convinced me of the connection between vagueness and “making a CI” of GVOs. A Moorean response to the claim that vague predicates are implicitly about concept users seems plausible to me. And even if Robinson’s strategy for responding to sorites paradoxes is enough to motivate the idea that vague terms are implicitly about concept users, the way in which concept users are involved need not support his conclusion.

These reflections can be applied, mutatis mutandis, to strategy (2). Robinson claims that the anthropocentric interests that lead us to pick out special science kinds like “star,” “planet,” “dog,” and “cell” are baked right into the meaning of these terms in such a way that there would be no planets or cells without concept users to identify them (just as there would be no heaps without heap-identifiers). Similar responses seem to me justified: the identification conditions for these kinds are no part of the meanings; but even if they were, it would not imply that the things that have them co-exist with minds essentially, since these terms should no longer be thought to correspond to essential sortal kinds.

6 Sparse Realism Requires “Making a Conceptual Interpretation” of GVOs

Strategy (5) turns on Robinson’s nominalism about non-fundamental properties. He seems to agree that the metaphysician needs a realist theory of universals for the perfectly natural properties. But he advocates a “sparse” realism, limited to just the perfectly natural. For non-fundamental predicates like “table,” “brain,” “organism,” etc., there need only be corresponding concepts. And concepts are mind-dependent things (Robinson 2016, 159, 183). I conjecture that Robinson further believes that, if the constituent concepts did not exist, the propositions in which such concepts figure would not exist either, and therefore could not be true. (Nor, presumably, could they be false.) This implication of his nominalism is suggested by the following remark about the non-fundamental category of tablehood (I note that this is not just a doctrine about artifactual kinds; Robinson clearly intends his moral to apply to non-fundamental categories from the special sciences, including biology, meteorology, etc.): “If there is no tablehood [i.e. if it is not a universal, but merely a concept], there are no instances of it, and so there are, in the realist sense, no tables” (2016, 183). The last clause—“there are, in the realist sense, no tables”—implies, according to his definition of existing in the “realist sense” (2016, 178), that it would not have been true that there are tables had there been no thinkers to use the concept of a table. Why should tables, organisms, mountain ranges, etc. be able to exist in the actual world, but not in worlds without minds; while quarks, electrons, etc. are not mind-dependent in this way? Robinson here seems to be offering sparse realism as the explanation: the fundamental physical entities are instances of real universals, which are not mind-dependent; while the non-fundamental things merely satisfy concepts, which are mind-dependent.

For this difference to explain why the non-fundamental things are mind-dependent, Robinson must be assuming that the propositions we grasp using these mind-dependent concepts would not be true in worlds without minds. After all, if the propositions about non-fundamental matters existed without us, then some of them, in some mind-free worlds, would have to be true. For example, suppose all life on Earth had been destroyed 700 million years ago, and no minds evolved elsewhere. There would still have been plenty of things we would call organisms, though none with minds. If basic propositions about cell biology would have existed in those circumstances, they would have to be true (after all, they correctly describe the organisms that existed 700 million years ago). So Robinson has to think that these propositions themselves are infected with the mind-dependence of their constituent concepts; and that propositions representing the world entirely in terms of universals are not mind-dependent.

Does the mind-dependence of the concepts we use to refer to and describe GVOs really imply that, were there no minds, there would be no GVOs—e.g.  that there would have been no trees or continents had there been no users of the concept of a tree or continent? Since the conclusion is quite unintuitive, all ways of avoiding it are worth exploring. Fortunately, there are many stations at which one can disembark before reaching Robinson’s destination. The assumption of sparse realism will naturally be questioned by many; as will the mind-dependence of concepts for the non-fundamental. But even if one grants Robinson his preferred metaphysics of properties and concepts, there remain ways to resist the slide from the mind-dependence of the concepts used to think about a certain subject matter to the mind-dependence of the subject matter itself.

Some philosophers of language appeal, in a metaphysically serious way, to propositions as abstract entities expressed by sentences and grasped by means of concepts; and they affirm that all propositions are necessarily existing things. They might, like Frege, take concepts of all sorts to be necessarily existing things as well; but that is not the only way in which one could maintain this view of propositions. Perhaps concepts are like words: words are contingent things that can be used to express propositions that are not themselves about the words used to express them, including propositions that could have been true even if those particular words never existed. Why could concepts not be similarly related to the propositions they enable us to grasp?¹¹

But there are ways to resist (a) even if one denies the necessary existence of propositions. Suppose that it is contingent which propositions exist; that existing propositions must be constructed out of existing materials, and that the non-existence of an individual or a concept prevents the construction of propositions explicitly about that individual or explicitly involving that concept. There remains room to make a distinction between the way in a which we can truly describe a circumstance, given the resources for constructing propositions that actually exist, and the ways in which the circumstance could have been truly described had it been the case. Given Robinson’s assumption of the contingent existence of (many) propositions (e.g. the ones that are not just about necessarily existing universals), there may be propositions available to us that truly describe a non-actual circumstance, but which simply would not have existed had that circumstance obtained. Consider again the world in which the Earth is rendered uninhabitable before sentience evolves. We naturally still want to use plate tectonics to describe that possible history of our planet; there would still have been, for example, continental drift. Had that alternate history occurred, there would have been no propositions about continents and plates, on Robinson’s hypothesis, because no one would have been around to take up the perspective from which plate tectonics can be used as an explanatory theory. Still, one wants to say, our description of this counterfactual circumstance is not false.

Singular propositions about non-existent individuals pose the same puzzle, under the supposition that such propositions are existentially dependent upon their constituents. A world without Julius Caesar can truly be described by us as lacking that very man; but, had that world been actual, no such description would have been possible; the propositions needed to express it would not have existed.¹² Taking this approach to singular propositions has led some philosophers to distinguish between “inner truth” and “outer truth” with respect to a possible world. The inner truths relative to a world are the propositions that would have existed and been true at that world, had it been actual; the outer truths relative to a world include also propositions available to us that (in some hard to specify sense) truly characterize the world, but that would not have been true had that world been actual due to the non-existence of those propositions.¹³ If such moves are needed, and available, for the case of singular propositions, they should suffice to undermine (a) as well.¹⁴

7 The Divine Mind and Assumption (b)

I have surveyed five of Robinson’s reasons for (a): the thesis that, were there no minds around, there would be nothing that satisfies what we mean by “organism” or “brain,” and nothing would have the properties we attribute to such things. I take (1), the argument from vagueness, to be the strongest. But it seems eminently resistible; and the others strike me as less compelling. Insofar as premise 2 depends upon the truth of (a), it appears to be in trouble.

premise 2 depends also upon the plausibility of (b): were there no creatures with our perspectives, there would be no other minds (e.g. no divine mind) capable of understanding propositions about the non-fundamental. This is not at all obvious to those—like myself and, for that matter, Robinson—who believe in something like the God of most monotheistic religions: a being endowed with both intellect and necessary existence. Had God not created us, God still could have done so, and ought therefore to have known what we would have been like, and what sorts of concepts we would have employed. If God cannot know what any non-fundamental phenomena would be like without knowing it in virtue of creating sufficiently intelligent creatures to apply the concepts of the higher level, then God would have to create blindly, to some extent.

The only rationale I can think of for maintaining (b), while accepting the existence of God, would come from emphasis upon the difference between divine and human intellection: perhaps God lacks the ability to think less than perfectly definite thoughts. As John Hawthorne points out, this has radical consequences:

[…] [S]ince our semantic and psychological concepts—means, refers, believes, loves and so on—are vague, we could not on this view coherently think of God as believing that we mean anything, refer to anything, believe anything, or love anything. (2005, 23, n. 12)

I should not like to go so far as that; and so I trust that a perfect being could understand imprecise thoughts. But then (b) looks clearly false, given theism.

All in all, then, I find many reasons to doubt premise 2.

8 Support for Premise 4

But suppose that premise 2 passed muster. Suppose that minds must exist in order for vague objects to exist. Could premise 4 be resisted?

According to premise 4, if the non-fundamental levels—including theories about the behavior of organisms and brains—are to be interpreted conceptualistically, “mind itself cannot be one of those non-basic levels” (Robinson 2016, 220). The reason mind cannot have an irreducible role in creating the levels and also belong to one of the levels is the viciousness of a certain regress.

[I]f all physical composites are artefacts of conceptualisation, and if the human being, brain, mind etc. are physical composites (and they are certainly not physical simples), then they are products of conceptualisation. What is it that does this conceptualising? Not something that only exists conceptually, on pain of a regress […]. (Robinson 2016, 179)

Robinson’s idea here is that something cannot have the power to generate a level—of objects and their distinctive higher-level kinds and properties—while belonging to that very level, and exercising this power in virtue of the higher-level properties appropriate to that level. If thinking things are themselves higher-level kinds, they are mere “patterns” which require “mental activity to reify them.” Reifying oneself would be a problematic kind of bootstrapping, and being reified by other concept users, who are in turn reified by others, etc., would lead to a vicious regress in which no one is reified.

To see whether the regress is truly vicious, I shall explore Robinson’s description of it in greater detail. He claims (2016, 219) that it is the same regress that afflicts Dennett’s “interpretationalism” about minds—the thesis that all minds display intentional states only “instrumentally, i.e. by interpretation” (2016, 213). As shall appear, Robinson’s anti-Dennett regress is not precisely the same as the regress Robinson invokes in the argument under consideration, which I shall call the “anti-GVO-materialism regress.” I will make the case that, although the anti-Dennett regress may be vicious, the anti-GVO-materialism regress is not obviously so.

If thinkers are non-fundamental, the concept of a brain or organism is that of a certain “pattern” in the fundamental physical world; and if the distinctive properties of thinkers are non-fundamental, thinking itself—propositional attitudes and other mental states—must be mere patterns, as well. These biological and psychological categories supervene upon fundamental physics (if the physicalists are right) even if they are not identifiable with something more fundamental. Robinson claims that, since they are supposed to supervene upon the fundamental, they are the kinds of patterns that would not have any distinctive effects, were it not for a mind that recognized them or interpreted them (see 2016, 220); and we should therefore give them all a “conceptualist interpretation.” The “grounds” for a pattern may exist, he says; but if it is not strictly identifiable with something fundamental, it does not automatically exist; there is “the need for mental activity to reify [the pattern] on the basis of those grounds.”

This explains why the mind cannot be just a pattern: it is presupposed by patterns as their co-inventor, together with the grounding. If the mind itself [were] just a pattern, then there would be the kind of regress with which we started our discussion, for it would not be reified unless it were seen as a pattern, and so on. (Robinson 2016, 219)

The anti-Dennett regress and the anti-GVO-materialism regress are, I think, importantly different. The target of the former is the intentionality of a system. Dennett denies that a brain, for instance, can be “intrinsically intentional.” Robinson assumes that there are only two ways to become an intentional system, either by intrinsically being one or by extrinsic interpretation; and that something “cannot have the capacity [to interpret something as an intentional system] solely in virtue of being itself interpreted by something else” (2016, 213). The resulting regress does seem to me to be problematic, given Dennett’s full position. However, it is not quite the same as the regress confronting the defender of a “conceptualising interpretation” of GVOs who takes thinkers themselves to be GVOs. I shall argue that the regress (or bootstrapping circularity) involved in supposing that GVOs generate the concepts on which they existentially depend is not so obviously vicious.

Dennett (on Robinson’s reading) denies the intrinsic intentionality of physical systems for very special reasons—he is averse to the “magic” of intrinsic “aboutness.”¹⁵ “Aboutness” must then come from an extrinsic source. If that source is an interpreter, who must already be able to think about things, it seems we may well be off to the races on a vicious regress. For X to be intentional, someone else S must interpret X’s states; but for S to be able to interpret states, S must itself be an intentional system; and so someone else must interpret S’s states. And so on, either in a regress or a circle, neither of which seems promising. The buck of “bad voodoo magic” is being passed, but it is never transformed into “good physicalist mojo.”

In Robinson’s anti-Dennett regress, X cannot give itself a certain property, but must rely upon something else’s being related to it in a certain way—and in a way that requires that further thing to have the same kind of property, but again only in virtue of someone else standing in the same sort of relation to it, and so on. In Robinson’s anti-GVO-materialism regress, the problem is quite different. A whole family of things, the Xs, (e.g. all organisms) are supposed to depend for their existence upon a further thing, Y (in this case, the concept of an organism). But the existence of Y itself is supposed to depend upon some of the Xs (at least some of them must have the concept). A direct circle of dependence is supposed to be vicious, and turning it into a chain is supposed to generate a vicious regress. There is a similarity between the two regresses, in that each involves the having of concepts; but it does not seem to me to go much deeper than that.

One of the big differences between the regresses is that, as Robinson understands the anti-Dennett regress, each X, in order to qualify as an intentional system, must get its intentionality from something; since it cannot get it from itself, it must get it from a distinct Y which is itself an intentional system interpreting X as intentional. In the anti-GVO-materialism regress, however, what X needs, in order to exist, is the existence of a thing, Y; Y itself is dependent, and (on pain of circular dependence) is supposed not to be able to depend upon X; so it must depend upon something else, some other Z which deploys the concept Y—but Z need not stand in any interesting relation to X. Z does not need to “interpret X as a Y,” or interact in any way with X. If X is a human brain, then what it needs in order to exist is not that any particular thing think about X (or think about the fundamental stuff grounding X) in a certain way, nor that X (or the fundamental stuff grounding X) be in any significant way related to a mind capable of categorizing it; nor that X itself be capable of thinking of itself, or thinking at all. When Robinson says that there would have been no brains or organisms or species had there been no minds to take our sort of perspective on the world, he is not saying that we have to think about each brain or organism or species in order to bring it into existence; he may be an idealist, but he is not what one might call a “Truman show” idealist—someone who disbelieves in all but the GVOs that individual human beings have actually encountered and conceptualized. Robinson nowhere endorses such an extreme position.

Nor need he. The five arguments I surveyed, above, for “making a CI” of organisms (and other higher-level entities) do not support the conclusion that each individual organism (or what-have-you) must be recognized and “thought into existence” by some human concept-user—that each animal must be paraded before some Adam, somewhere, on pain of non-existence. They are arguments that humans must exist somewhere in order for organisms to exist anywhere. Take strategy (1), which requires that vague kinds, like heaps, only exist if there are concept-users who can recognize a heap: so long as there are minds capable of taking our sort of perspective on the world, some matter piled in a heap could be recognized as such, and so is a heap—even if that particular parcel of matter is never in fact recognized as a heap by anyone. Or take strategy (5), which turns on the contingent existence of propositions involving higher-level concepts: so long as the concepts of organisms and brains exist, propositions about organisms and brains are available to be true—including propositions that truly describe organisms and brains no one happens to notice. So all that is required for all the brains and organisms to exist is for someone, somewhere, to have the concept.

The regresses may be different, but they might both be vicious nevertheless. However, there is reason to be suspicious of the form of the anti-GVO-materialism regress, since some respectable metaphysical positions imply that it is benign. The doctrine resembles a less radical and quite popular view about universals—defended, for example, by David Armstrong—according to which there can be no uninstantiated universals.¹⁶ Suppose electronhood is a genuine universal, present in all electrons; and that every electron is essentially an electron. It is at least tempting to say that the electron depends for its existence upon electronhood, since it could not exist without exemplifying it. But, according to Armstrong’s theory, the existence of the electron is also sufficient for the existence of the universal; and, if it were the only electron ever to exist, the existence of the property would depend counterfactually upon the existence of this electron. This sets up exactly the kind of circularity that is supposed, by Robinson, to be vicious and to generate a regress which is itself vicious.

How does Armstrong deal with the apparent two-way dependence of properties on things that have them, and the dependence of things on their essential properties? I cannot find him directly addressing the question in these terms, but there are some suggestive passages, and some obvious moves available. For one thing, the kinds of (alleged) dependency seem quite different; so there may be no circularity at all, or only circularity of a benign sort. For example, he might well say that, although an electron cannot exist without being an electron, that does not mean it depends for its existence upon the property of being an electron. Not everything that must exist, if I exist, is something I am dependent upon; some such things are dependent upon me, but accompany me necessarily (e.g. my unit set, which automatically shows up if I do, but arguably depends upon me, and not the reverse). Armstrong could then say that, necessarily, electronhood exists if any electron exists; but that the property is dependent upon its instances.¹⁷

Alternatively, when a circle of dependence threatens, one may posit shared dependence upon a further thing. Necessary connections exist between the two entities, but these are signs not of a circle of dependence, but of mutual dependence upon something more fundamental yet. Armstrong’s later work on particulars and universals treats both as abstractions dependent upon—or, to use a phrase suggesting both dependence and greater fundamentality, grounded in—things of a further ontological category: states of affairs. Proposing that universals are “state-of-affairs types […] brings out the dependence of universals upon states of affairs. As such, it should at least incline us to accept the primary position of states of affairs and to be sceptical about the reality of uninstantiated universals” (Armstrong 1997, 29).

I shall suggest that the anti-GVO-materialism regress can be defused in the second of these two ways: the existence of both organisms and the concept of an organism are dependent upon more fundamental facts. Take some organized fundamental physical activity that is sufficient for the existence of an organism—at least, activity that is sufficient in a world in which someone, somewhere has the concept of an organism. According to premise 2, the activity in question has the status of (constituting) an organism in virtue of someone, somewhere, having this concept. Now, if this “someone, somewhere” is itself an organism, does it not exist in virtue of its own ability to deploy the concept of an organism? If it is the first and only organism employing the concept, it would seem so. But even if there are many organisms with the concept, take them all as a group; are they not pulling themselves up by their own bootstraps?

Because both concept possession and organism status are—on physicalist assumptions—dependent upon (or grounded in) more fundamental physical facts, this does not seem to me to be anything like the kind of bootstrapping needed to escape the anti-Dennett regress. Why could not both concept possession and organism status be dependent, simultaneously, on the same more fundamental facts?

To undermine premise 4, a critic can maintain that it is intrinsic to the matter making up a properly organized human brain or body that it grounds the existence of a GVO and also that it grounds the exemplification of concepts for GVOs, like the concepts of a brain or animal body. Something’s having such concepts is, on Robinson’s view, necessary for the existence of brains and bodies. But it is the mere existence of the concept of a GVO that is required for there to be truths about that kind of GVO; the possessors of the concept need not be applying it to themselves or other concept-users in order to “bring themselves into existence.” So long as the concept is available and the matter of the world is arranged in a way sufficient for application of the concept, there will be truths about such GVOs, whether or not someone applies them. This seems very different from Dennett’s intentionality regress—no act of recognizing myself as an organism, nor any other interpretative act aimed at myself, is required for the possibility of the truth that I am an organism, only my (or someone else’s) ability to deploy the concept—to take up the human perspective. And this seems sufficiently grounded in the physical—at least on physicalist assumptions about what it is to have a concept.¹⁸

An analogous case of an aesthetic property will help illustrate how such simultaneous grounding can occur. The example is particularly apt, because the aesthetic property is meant to be very much like the concept of an organism, given a conceptualist interpretation of organisms. The property exists contingently; it can, arguably, be exemplified essentially by something, and also depends (at least counterfactually) for its existence upon that very thing; and it also applies to things (namely, passages of literature) that could have existed. In these ways, it is more like organism according to Robinson than electronhood according to Armstrong. Collections of material parts like those in an amoeba could have been arranged just as they are but fail to be an organism due to the non-existence of certain other organisms, namely, human beings. Nothing like that could happen with electronhood; anything intrinsically just like an electron must be an electron.

In “Kafka and His Precursors,” Jorge Luis Borges argues for the contingency of a certain aesthetic property, which could be called Kafkaesque. After detecting Kafka’s “voice, or his practices” (1964, 199) in a number of literary characters, themes, and passages that predate Kafka’s own work, Borges draws a conclusion about the property they have in common:

If I am not mistaken, the heterogeneous pieces I have enumerated resemble Kafka; if I am not mistaken, not all of them resemble each other. This second fact is the more significant. In each of these texts we find Kafka’s idiosyncrasy to a greater or lesser degree, but if Kafka had never written a line, we would not perceive this quality; in other words, it would not exist. (1964, 201)

According to Borges, without Kafka, a host of passages throughout the history of world literature would not have had the property of being Kafkaesque. The words in the passages to which he draws our attention would still have been there, but they would not have been marked by this characteristic. In order for them to resemble one another in this particular way, Kafka (or someone with Kafka’s sensibility) had to write a body of literature sufficient to bring the property into existence.

The relevance of Borges’s theory for premise 4 can be seen by noting that the body of literature that is responsible for bringing this property into existence (The Trial, The Castle, etc.) would itself have the property. Is there anything circular about supposing that what Kafka did—the writing of the words he wrote—both brought passages of literature into existence and created the property of being Kafkaesque which they exemplify? It does not seem so to me—even if those passages (in The Trial, The Castle, etc.) are essentially Kafkaesque, so that they could not have existed without the property.

The example seems perfectly analogous to the supposition that both the existence of a human organism and the existence of the concept of an organism be grounded in further, more fundamental physical facts. The complex activity within a brain is like Kafka’s putting pen to paper, and the concept of a brain or organism is like the property of being Kafkaesque. The structure and functioning of the cells that make up a particular human brain is enough to ensure the existence of certain GVO concepts, like that of a brain or an organism (so long as the brain in question subserves the activity of thinking about brains and organisms). That same arrangement of living cells is sufficient to ensure the existence of a GVO that falls, essentially, under one of those concepts (namely, the concept of a brain). This is no more paradoxical than the idea that Kafka’s writing creates the property of being Kafkaesque and also creates a piece of literature that is essentially Kafkaesque.

I am considering the possibility of resisting premise 4 by agreeing that the existence of organisms and other GVOs depends upon some organism (or brain) using these concepts; call this position “conceptual dependence.” The view gives the concept-using organisms a special role among all the instances of things that fall under the concept: had the concept-users not existed, and no replacement thinkers been introduced to cook up the concepts instead, all the other instances of GVOs would not have existed either, even had the matter that makes them up remained just as it is. I grant that there is something vertiginous about this supposition—and a precisely parallel sense of vertigo is created by Borges’s theory of the Kafkaesque (which is, of course, part of its typically Borgesian charm). It is illuminating to consider how Borges’s aesthetic property can be made to seem less paradoxical: doing so will shed light on how this way of denying premise 4 could be made to seem less strange as well.

It is natural to think of aesthetic properties as intrinsic to the things that have them.¹⁹ If several works of literature resemble one another in a certain intrinsic respect, any one of them could have ceased to be without the others losing the property in virtue of which they resemble—that is part of what it is for the property to be intrinsic. How is it that Kafkaesque is not like this? If it were covertly relational, involving similarity to an aspect of this particular body of work (The Trial, The Castle, etc.), then the dependence would no longer be surprising; the aesthetic property would not be entirely intrinsic after all. It would be partly relational, depending upon both the intrinsic features of various works that have it (their proto-Kafkaesque elements, one might say), and their relation to a property of Kafka’s works that is unique to Kafka (the particular ways in which his work is Kafkaesque). Someone who accepts the doctrine of conceptual dependence (for GVOs) thinks that the matter actually constituting an organism, brain, or other GVO could exist in an intrinsically similar state without constituting such things (so long as the matter does not also constitute a thinker with the concepts in question). A person who holds this view seems to regard being an organism, brain, etc. as covertly relational; it requires not only an intrinsic ground for the application of the concept (matter arranged as it is in living things), but also the existence (somewhere, at some time) of at least one instance of a person utilizing the relevant concept. The contingently existing concept plays the role of the unique property of Kafka’s writing.

I do not find the doctrine of conceptual dependence at all attractive, for reasons that will be obvious from my criticism of premise 2. However, those who accept it are, in effect, saying that being an organism is covertly extrinsic and relational in much the same way as Borges’s Kafkaesque.

I see no inherent instability, then, in holding that an arrangement of matter might be sufficient both to generate a concept of a certain kind and the intrinsic grounds for the application of that concept—in much the same way that Kafka’s words could be arranged in such a way as to generate both a new property shared among many literary works, and also an instance of a work that essentially has that property.

9 Conclusion

As I said at the outset, I agree with Robinson that there is something deeply problematic about supposing oneself to be a vague object. However, Robinson’s attempt to pinpoint the problem will not, I fear, convince many materialists that the problem is real. Most analytic philosophers would doubtless deny premise 2, adopting the Moorean response to the claim that non-fundamental kinds are mind-dependent. The conclusion that a world without minds could not contain mountains or trees is shocking enough to require considerable defense; and, as I showed, there appear to be plausible ways to resist Robinson’s arguments for “making a conceptual interpretation” of the non-fundamental. But, even if these arguments held up, there appear to be plausible lines of resistance to premise 4.

That may not be the end of the story, of course. I treated the five lines of argument in support of premise 2 as independent strands. But I greatly simplified them. They deserve closer attention, and may well have been intended as interwoven parts of a larger argument that I have not fully grasped. I may also have missed some arguments for premise 2 altogether—the book is complex, and its arguments have many moving parts. My criticism of premise 4 is also far from conclusive. (For one thing, in conversation, Robinson has helped me to see a potential response to my attack upon premise 4.²⁰) Often enough, a critique of some philosopher’s argument will seem devastating, until the target of the critique has the chance to respond. I hope Robinson returns to these topics soon, and sheds more light on his proposed path from vagueness to the denial of garden-variety materialism.