1. Conceivability Arguments, Properties, and their Powers: The Current State of Play

The most popular version of the necessitarian position maintains that the essences of spatiotemporally instantiated (or ‘natural’) properties are determined by the powers that they confer upon their bearers; according to this view, all natural properties are intrinsically powerful. (For the record, Ellis [Reference Ellis2001] and Molnar [Reference Molnar2003] maintain that while some natural properties are intrinsically powerful, others are not.) I will follow the trend of referring to this position as dispositionalism. Within this camp there are different accounts of the exact relationship between a property's powers and its essence. Some (Shoemaker Reference Shoemaker and van Inwagen1980, Reference Shoemaker1998; Bird Reference Bird2007) maintain that a property's powers exhaust its essence. Others (Martin Reference Martin1997, Reference Martin2008; Heil Reference Heil2003, Reference Heil2005; and Strawson Reference Strawson2008) maintain that although a property is essentially linked to the powers it confers, its essence is not exhausted by those powers (for my take on this issue, see Schroer Reference Schroer2010).

A common objection to dispositionalism is that since we can conceive of properties as failing to confer at least some of the powers that they confer in the actual world, we have reason for thinking the connection between properties and their powers is contingent and not necessary as the dispositionalists maintain. I do not want to spend much time exploring the various senses in which a scenario could be said to be ‘conceivable’ and identifying which of those senses is relevant to the conceivability/possibility inference (attempts to execute such a project can be found in Yablo Reference Yablo1993 and Chalmers Reference Chalmers, Gendler and Hawthorne2002.) Instead, I will adopt what Gendler and Hawthorne (Reference Gendler, Hawthorne, Gendler and Hawthorne2002: 9) refer to as the ‘placeholder’ use of the term, which involves using ‘conceivability’ in the specific sense that is relevant to acquiring evidence for the possibility of some scenario while not specifying what the actual act of representing that scenario is like. (Terminological note: I will be using ‘conceivability’ and ‘imaginability’ interchangeably in this paper, so please extend the above qualification concerning my use of the former term to the latter term as well.)

With this qualification in place, let us take a closer look at the conceivability argument against dispositionalism. In exploring this objection, it will be useful to work from some concrete examples. Consider, for instance, the microphysical property that makes salt soluble in water. (By describing it as ‘microphysical’, I am treating this property as a first-order property. Some would instead analyze the solubility of salt as a second-order, functional property; see, for instance, Prior, Pargetter, and Jackson [Reference Prior, Pargetter and Jackson1982]. I ignore this complication in what follows.) Let us call this property S. It seems we can conceive of a possible world in which an object instantiates S and yet fails to dissolve when placed in water. More carefully, it seems we can conceive of a possible world where (1) the stimulus conditions relevant to S manifesting its disposition to make its bearer dissolve in water obtain, (2) there are no ‘finks’, ‘antidotes’, or anything else that could interfere with S’s manifestation (see Martin Reference Martin1994 and Bird Reference Bird1998), and yet (3) the object that instantiates S fails to dissolve when placed in water. What is true of S seems true of other properties and their powers as well. For instance, it seems we can conceive of a possible world where an object instantiates sphericity and yet fails to roll when placed upon an incline despite being in conditions where it should do so. It seems we can conceive of a possible world where a moving object with mass does not transfer its momentum to a second object that it collides with despite being in conditions where it should do so. And so on.

Given the assumption that conceivability is a prima facie guide to possibility, the conceivability of the above-described scenarios provides us with prima facie evidence for thinking that these properties do not have their powers built into their essence, pace dispositionalism. In response to this challenge, some suggest that it is a mistake to view conceivability of any stripe as a prima facie guide to possibility. For instance, both Ellis (Reference Ellis2001) and Molnar (Reference Molnar2003) attempt to shake our confidence about the general inference from conceivability to possibility by appealing to purported counterexamples to that inference. This response comes with a high price, however, in that it throws all conceivability arguments, and not just those directed at dispositionalism, under the bus.

A more popular dispositionalist response is to accept that conceivability is a prima facie guide to possibility while also maintaining that contingentists have made a mistake in describing what they are imagining—in performing the act of imagination, contingentists have acquired evidence for the genuine possibility of some scenario, just not the particular scenario that they were trying to imagine. (For examples of this second strategy, see Shoemaker Reference Shoemaker1998; Ellis Reference Ellis2001; Heil Reference Heil2003; and Bird Reference Bird2007. The core idea of undermining problematic conceivability claims by way of redescription comes from Kripke [Reference Kripke1980] who uses it to undermine conceivability objections to a posteriori identities such as ‘water = H₂O’.) When contingentists assert that they can imagine an object that instantiates S that does not dissolve when placed in water, for instance, dispositionalists deploying this ‘redescription strategy’ will claim that what the contingentists are actually imagining is an object that instantiates an S-like property not dissolving in water. Or that what they are actually imagining is an object that instantiates S that fails to dissolve when placed in a water-like liquid. The important thing, from the perspective of the dispositionalists, is that neither of these possibilities threatens the claim that, due to the essence of the properties involved, salt must dissolve when placed in water. (Notice that according to the redescription strategy there is still a type of contingency in play: namely, a contingency about which properties are instantiated by a given object in a particular possible world. This kind of contingency, however, is no threat to dispositionalism. Additional discussion of this point can be found in Mumford [Reference Mumford1998] and Heil [Reference Heil2003].)

As you might expect, this line of defense has not shaken the confidence of contingentists much. Here is what Jonathan Schaffer has to say about it:

Let us dig a bit deeper into this standoff. In doing so, I want to focus on an issue that, while having been extensively explored in the philosophy of mind, has not been discussed much in this context: namely, the nature of the concepts that are used in performing these acts of imagination or conception. In particular, I want to focus on the cognitive contents of these concepts. By ‘cognitive content’ I mean the kind of content that determines how a subject thinks about (or ‘grasps’) the referent of a concept when using that concept. Understood this way, cognitive content grounds whatever a priori beliefs the subject has about a referent when thinking about it using the concept in question. The cognitive content of the everyday person's concept of water, for instance, is along the lines of ‘a clear, tasteless, drinkable liquid in my environment’. The cognitive content of a concept is what determines the epistemic possibilities for the subject with regard to the referent of that concept. For example, if your concept of water carries the cognitive content ‘a clear, tasteless liquid’, you would not be a position, using that concept, to distinguish epistemically the actual referent of the concept, H₂O, from other clear, tasteless liquids (such as XYZ).

With the notion of ‘cognitive content’ in hand, let us take a second look at the redescription strategy. This strategy claims that contingentists are making a mistake that involves confusing one type of property for another but superficially similar type of property. Shoemaker (Reference Shoemaker1998), for instance, claims that in performing the relevant acts of imagination the contingentist grasps the property in question in virtue of the effects that it has on her sensory system (or some observable instrument) while not grasping some of the other powers of that property. Shoemaker goes on to claim that these particular powers—the powers that allow the contingentist to ‘grasp’ the property in question—need not be unique to a single property-type; other property-types, including other possible property-types that are not instantiated in this world, may also confer the same powers. Since the mode of presentation through which the subject grasps a given property-type need not be uniquely associated with that property-type, Shoemaker says the contingentist cannot be sure that she is imagining something about that specific property-type and not accidentally imagining something about some other property-type that happens to confer the same power.

As we have seen, contingentists are not moved by such attempts at redescription. They will reject the claim that in performing these acts of imagination they grasp the relevant properties in virtue of the cognitive contents of their concepts characterizing powers that are not uniquely associated with the property-types in question. Given this, it is natural to wonder how the contingentists’ alternative account of how we grasp these properties will go. I do not know of any contingentist who has explicitly addressed this question, but I think we can lay down the general shape of what their rival account should look like. To start with, they should reject the claim that the concepts used in performing these acts of imagination characterize their referents—i.e., various property-types—exclusively in terms of some of the powers that those properties confer upon their bearers. In rejecting this claim, contingentists can allow that these concepts provide some characterization of their referents in terms of the powers they confer. The important thing, however, is that contingentists should maintain that these concepts provide at least some characterization of their referents that is not in terms of the powers that those properties confer. Let us call the latter kind of characterization a non-powers characterization of a property.

The reason a non-powers characterization should be part of the contingentists’ account is that it would allow them to resist the redescription strategy by putting them in a position to claim that the cognitive contents of the relevant concepts uniquely identify their referents. Recall that, according to contingentists, properties have transworld identity conditions that involve something other than their powers. (For discussions of what this ‘something else’ might be, see Locke [Reference Locke2012] and Smith [Reference Smith2016].) If the concepts we use in imagining the relevant scenarios provide a characterization of the non-power essences of the relevant property-types, then those concepts will uniquely identify those property-types. By advancing such a position, contingentists would be in an excellent position to reject the suggestion that they have made the mistake that the redescription strategy is accusing them of.

Although I am a dispositionalist, I find myself siding with contingentists on the issue of whether I have the ability to imagine these scenarios in a way that involves concepts with cognitive contents that uniquely identify their referents. I think there is a way of imagining various things about property S, for instance, where there is something in my grasp of this property that assures me that I really am imagining something about an object that instantiates property S and am not (accidentally) imagining something about an object that instantiates an S-like property instead. In the next section, I will draw upon an idea from the literature on the conceivability arguments leveled against physicalism that will allow me to grant contingentists this description of what I am imagining while still avoiding the conclusion that such a scenario is genuinely possible.

2. A New Dispositionalist Response to Conceivability Arguments

Contemporary physicalists are divided in terms of how they respond to conceivability arguments leveled against their position: some argue that the relevant scenarios (inverted spectra, zombies, etc.) are not actually conceivable, while others grant the conceivability of these scenarios but deny that they are genuinely possible. Following the lead of Chalmers (Reference Chalmers1996), we can refer to the former as type-A materialists and the latter as type-B materialists. My interest resides with an early, influential type-B materialist position advanced by Brian Loar (Reference Loar, Tomberlin, Block, Flanagan and Güzeldere1990). As I interpret Loar's account, the reason that inverted spectra, zombies, etc. are conceivable is that our ‘phenomenal concepts’—i.e., the primary concepts we use to think about phenomenal properties in a direct and unmediated fashion—provide us with a characterization of their referents that is too thin to allow us to grasp fully their physical (or functional) nature. This, in turn, means that our grasp of these properties is too thin to reveal the metaphysical impossibility of such imagined scenarios. Our introspective grasp of phenomenal red, for example, is underpinned by a type-demonstrative concept that characterizes its referent as ‘that type of quality’ and not in terms of its physical (or functional) nature. As a result of the thinness of our grasp of this property, our ability to imagine various scenarios involving it is not compelling evidence for thinking that those scenarios are genuinely possible. My failure to uncover a problem when imagining an inverted spectrum scenario involving a phenomenal property that is merely characterized as ‘that type of property’, for instance, is not compelling evidence for thinking that such a scenario is genuinely possible; even if this property is physical or functional in nature, I will still be able to imagine it behaving contrary to its essence using the type-demonstrative concept described above.

Loar acts as though phenomenal concepts are anomalous relative to most of our other concepts in terms of the thinness of their cognitive content. A different route towards the same overall strategy of undermining conceivability arguments against physicalism would be to claim that phenomenal concepts are not unusual in this regard. Joseph Levine (Reference Levine2001: 84), for instance, claims that almost all of our concepts provide thin characterizations of their referents; even the characterization provided by our regular concept of water is ‘a “thin” one that merely labels a phenomenon/substance in the world’. A similar position is occupied by Block and Stalnaker (Reference Block and Stalnaker1999). As you will see, I will follow Loar by focusing specifically on type-demonstrative concepts.

Loar's account has had a substantial impact upon the literature within the philosophy of mind and has spawned a number of follow-up accounts that build on (and attempt to improve upon) the core idea of his original position. What I am interested in, however, is whether the strategy of appealing to type-demonstrative concepts can be used to undermine conceivability arguments against other philosophical positions, including dispositionalism. As far as I can tell, there is nothing in this general strategy, which we can call the type-demonstrative strategy, that restricts its application in principle to just the conceivability arguments that have been leveled against physicalism.

Part of the plausibility of Loar's particular appeal to the type-demonstrative strategy is that our primary grasp of phenomenal properties is through introspection, and Loar makes a compelling case for the existence of a class of type-demonstrative concepts that can be deployed through introspection. He claims that these type-demonstrative concepts are ‘recognitional’ in that they are grounded in various recognitional abilities that type experiences in virtue of their phenomenal properties—just as we can form a conception of ‘those hedges’ in virtue of making use of various perceptual recognitional abilities, we can form a conception of ‘that kind of experience’ by making use of various introspective recognitional abilities (Reference Loar, Tomberlin, Block, Flanagan and Güzeldere1990: 601).

Where Loar focuses on phenomenal properties, I want to use the type-demonstrative strategy to undermine contingentist conceivability arguments aimed at any and all (natural) properties. This shift from a limited range of properties that we grasp (primarily) through introspection to a far wider class of properties raises some questions: How do we form type-demonstrative concepts of all these properties? Are these type-demonstrative concepts supposed to be the ‘primary’ way we have for thinking about all of these properties? If they are not, then what reason is there for thinking that these concepts (and not some other concepts) are deployed when we try to imagine what contingentists claim to be imagining? And what, precisely, are the cognitive contents of these type-demonstrative concepts?

Obviously, I cannot follow Loar's lead in claiming that the relevant type-demonstrative concepts are all recognitional concepts that are applied introspectively to one's own experiences. But this is not a substantial problem, for there are plenty of other ways that subjects could come to possess type-demonstrative concepts of various properties besides introspection. A more serious question facing my account stems from the fact that we also possess non-type-demonstrative concepts of many of these same properties. A student who is sitting in a chemistry class and thinking about the microphysical property that is responsible for salt's solubility (S), for example, is probably not grasping that property through a type-demonstrative concept that carries a thin cognitive content. (Perhaps she grasps it through a concept that characterizes S in terms of its molecular structure or in terms of some of its causal powers.) Given this, what reason is there for thinking that when this same student is sitting in her philosophy class, imagining salt that does not dissolve when placed in water, she is now using a type-demonstrative concept to grasp that same property? In absence of an answer to this question, my appeal to the type-demonstrative strategy may look unmotivated.

As a first step in responding to this question, I want to concede that there are ways of performing acts of imagination targeting property S that do not involve grasping that property through a type-demonstrative concept. Consider, for example, the act of imagination that dispositionalists who adopt the redescription strategy take themselves to be performing. They will claim that when they seemingly imagine something that instantiates S failing to dissolve in water, what they are actually imagining is something that instantiates an S-like property that fails to dissolve in water. The natural thing to say about this particular act of imagination is that the subject is grasping a property through a non-demonstrative concept that carries a cognitive content that, as it turns out, does not uniquely identify S. (In fact, in this particular scenario the concept does not even pick out S; instead, it picks out an S-like property.) I am happy to grant that in cases of imagination like this, there is no reason to think that a type-demonstrative concept is involved.

However, what I am interested in is the act of the imagination that the contingentist takes herself to be performing. Through the cognitive contents of the concepts involved, this act of imagination is supposed to provide the subject with a positive reason for thinking that she is imagining what she thinks she is; it is supposed to provide her with some form of reassurance for thinking she really is imagining something about S and not (accidentally) imagining something about an S-like property instead. It is just in these cases of imagination that I want to maintain that the subject is grasping the relevant property via a type-demonstrative concept.

So how is it that a type-demonstrative concept can come to be involved in cases of imagination like this? When I want to I imagine a piece of salt that does not dissolve in water in the manner that the contingentist does—i.e., when I want to imagine this scenario in a way where something in my grasp of property S indicates to me that I have not accidentally switched to imagining an object that instantiates an S-like property instead—here is what I do: I start off by thinking about property S as it is in the actual world, using whatever concept I normally use when thinking about S. (This initial concept of S may itself be demonstrative or not. If this concept is descriptive, it may characterize S through a contingent mode of presentation or not. It does not matter.) Once I have this property placed before my mind, I mentally ‘tag’ it using a type-demonstrative concept—I think of it as ‘that type of property’, ‘that type of property I was just thinking about’, etc. (It would be nice to have an account of the specific details of how such type-demonstratives concepts are able to ‘tag’ properties in the manner that I describe. I take it, however, that the phenomenon I describe has enough intuitive plausibility that I can survive without actually providing such an account. Loar also does not offer an account of the details of how the introspectively applied recognitional concepts he posits are able to ‘tag’ the phenomenal properties they do [see Loar Reference Loar, Tomberlin, Block, Flanagan and Güzeldere1990: 601].) As a result of this demonstrative act of mental tagging, I come to have a cognitive grasp of whatever property I was originally thinking about that is now both direct and which does not involves a mode of presentation that is only contingently associated with that property. This type-demonstrative concept, in turn, is what I subsequently use in performing the act of imagination.

According to this account, the general shape of the cognitive content of my act of imagination, when I try to imagine things the way contingentists want me to, takes following form:

The presence of the type-demonstrative in this content is what provides me with a straightforward reason for thinking that I am still thinking about the same type of property (S) that I was originally thinking about before I undertook the act of imagination. It is what gives me a reason for thinking that, in performing the act of imagination, I have not accidentally switched from thinking about the property I was originally thinking about, namely, S, to some other S-like property instead.

Notice that this account of how type-demonstrative concepts enter into the relevant acts of imagination is pretty general; the account can be applied to any property that other concepts, including nondemonstrative concepts, can (initially) place before the mind. This means my account can handle the full range of properties that contingentists wish to apply their conceivability arguments to. This account can also allow for some variation within the richness of the cognitive content of the relevant type-demonstratives concepts. In some cases, the cognitive contents of these concepts might be as bare as ‘that property’ or ‘that property I was just thinking about’. In other cases, there could be additional descriptive elements contained within the cognitive contents of these concepts: they might characterize their referents as ‘that experiential property’, ‘that physical property’, ‘that perceivable property’, etc.

The account thus has flexibility both in terms of the properties to which it can be applied and in terms of how our type-demonstrative concepts characterize those properties. To undermine conceivability arguments against dispositionalism, all that the account really requires is that (1) these concepts are demonstratively formed on the basis of mentally tagging the property prior to the act of imagination, and (2) the cognitive contents of these type-demonstrative concepts are not detailed enough to put the subject in a position to appreciate the metaphysical impossibility of various (imagined) scenarios that involve that type of property failing to confer some of the powers that it confers in the actual world. Condition (1) is important to the account because it explains how the cognitive contents of these concepts provide subjects with a positive reason for thinking that, in performing the act of imagination, they have not accidentally switched which properties they are thinking about. Condition (2), in turn, is important because it explains why these acts of imagination do not ultimately provide subjects with compelling evidence for thinking that what they have imagined is genuinely possible.

This new strategy for responding to conceivability arguments does some interesting things for the dispositionalist. To start with, it puts the dispositionalist in a position to grant the contingentist her preferred description of what she is imagining—she can grant that the contingentist is imagining an object that instantiates property S that fails to dissolve when placed in water and no longer needs to claim that the contingentist has made a mistake that involves confusing an S-like property for property S in her act of imagination. What is more, in virtue of characterizing its referent as ‘that type of property’, a type-demonstrative concept provides a non-powers characterization of its referent, which is also something the contingentist should be drawn to. Relatedly, a type-demonstrative concept does not pick out its referent by describing it in terms of something that some other property-type could also share—instead, it picks out its referent directly and not through a contingent mode of presentation of that referent. This means that, despite being thin, a type-demonstrative characterization still distinguishes its referent relative to all other property-types and thereby uniquely identifies it. Despite granting the contingentist all of these claims, however, the type-demonstrative strategy still blunts the inference from conceivability to possibility. If in imagining an object that instantiates property S that does not dissolve in water, the subject's grasp of S is mediated through a type-demonstrative concept with an appropriately thin cognitive content, this act of imagination does not provide compelling evidence for thinking that such a scenario is genuinely possible.

The obvious reply for contingentists to make to all of this is to claim that they do not use type-demonstrative concepts in performing the relevant acts of imagination. Instead, they use some other kind of concept that, like a type-demonstrative concept, is capable of providing a non-powers characterization of its referent that uniquely identifies that referent relative to other property-types but, unlike a type-demonstrative concept, does not undermine the inference from conceivability to possibility. A similar reply has been given in response to Loar's use of the type-demonstrative strategy in philosophy of mind: a number of people have asserted that when imagining inverted spectra, zombies, etc., they are not actually using the kind of introspectively applied type-demonstrative concepts that Loar says they are. In particular, these folks argue that the concepts Loar describes are not viable candidates for our phenomenal concepts by arguing that there is an aspect of the cognitive contents of our phenomenal concepts—their substantial or ‘thick’ nature—that cannot be accommodated by an account that identifies those concepts with type-demonstrative concepts that carry thin cognitive contents. In the next section, I will review this complaint against Loar's original use of the type-demonstrative strategy and explore whether the contingentist will be able to make a similar complaint against my use of that strategy.

3. When We Imagine Things how the Contingentist Wants us to, is our Grasp of the Relevant Properties Thick or Thin?

One common complaint leveled against Loar—a complaint leveled by dualists and even some physicalists—is that his account fails to do justice to the substantial grasp we seem to have of phenomenal properties in virtue of thinking about them using phenomenal concepts. Here, for example, is Janet Levin's statement of the problem.

David Chalmers and Joseph Levine also emphasize how substantial our grasp of phenomenal properties seems to be in discussing (and critiquing) Loar's account.

There are several things that I would like to draw your attention to with regard to these claims. First, this thick, substantial grasp is supposed to be of the essence of the phenomenal property in question. (For a discussion of the senses in which this grasp could be described as ‘thick’, ‘substantial’, or ‘determinate’, see Veillet [Reference Veillet2005].) My phenomenal concept of phenomenal red, for instance, is supposed to provide me with substantial knowledge of the essence of that property. It is also worth noting that the phenomenal properties themselves are thought to have thick (or substantial) metaphysical natures. This means is that phenomenal concepts are supposed to provide an epistemically thick grasp of phenomenal properties that possess metaphysically thick natures.

The objection against Loar is that there is a significant mismatch between what he says the cognitive contents of our phenomenal concepts should be like and what our a priori access to those cognitive contents actually reveals about them. Could a similar objection be made against my use of the type-demonstrative strategy? To find out, let us explore the prospects of an objection against my use of the type-demonstrative strategy that, like the objection leveled against Loar, asserts that the cognitive contents of the relevant concepts provide a thick or substantial characterization of the essences of the properties they refer to—essences that, according to the contingentist, involve something other than the powers that are associated with those properties.

As it turns out, a number of contingentists are committed to claims that would make it challenging for them to echo the complaint made against Loar in this new context. One such claim is that properties have unknowable non-power essences. A surprising number of contingentists are drawn to this claim. For instance, Frank Jackson (Reference Jackson1998), Rae Langton (Reference Langton1998), and David Lewis (Reference Lewis, Braddon-Mitchell and Nola2009) all embrace it. (Ellis [Reference Ellis2001] and Molnar [Reference Molnar2003] also consider, albeit in less detail, the possibility that properties have unknowable non-power essences.) It is not important, given our purposes, to spend much time examining (or evaluating) the arguments that contingentists have offered for thinking that properties have unknowable non-power essences. (Typically, these arguments focus on the difficulty of discriminating epistemically between distinct non-power property-types that happen to confer the same powers as one another; for discussions of this style of argument, see Schaffer [Reference Schaffer2005] and Lewis [Reference Lewis, Braddon-Mitchell and Nola2009].) What is important for us is the manner in which these contingentists think that the non-power essences of properties are unknowable. They do not claim that these properties are unknowable in virtue of our concepts providing a substantial characterization of their non-power essences that, for whatever reason, fails to qualify as knowledge. Instead, these properties are claimed to be unknowable in virtue of our concepts providing no positive characterization whatsoever of what their non-power essences are like. As Frank Jackson (Reference Jackson1998: 24) puts it, the nature of these properties is ‘irretrievably beyond our reach’.

This claim, popular with a number of contingentists, has significant implications for what those contingentists could say about their cognitive grasp of the properties that they are imagining things about. Recall that we are interpreting contingentists as maintaining that, when they perform the relevant acts of imagination, they grasp the relevant properties in virtue of grasping their non-power essences. From this assumption and the claim that properties have unknowable non-power essences, it follows that these contingentists will be committed to the idea that their concepts provide an extremely thin characterization of the properties in question. It would seem that the best the cognitive contents of those concepts could do is characterize various property-tokens of the same property-type as having the same non-power essence as one another while providing no positive characterization of what that non-power essence is like.

If this is correct, it means that these contingentists are committed to an account of the cognitive contents of the concepts used in performing the relevant acts of imagination that is not far off from what my type-demonstrative account says these cognitive contents are like. Both accounts will end up claiming that these concepts characterize a given property-token as having the same essence (or as not having the same essence) as various other property-tokens while not providing any additional information about what that essence is like. Of course, my account and their account will disagree about why our concepts provide such thin characterizations of their referents: I think it is because the concepts are type-demonstrative in nature (they label a given property along the lines of ‘that type of property’), whereas they think it is because the properties in question have unknowable non-power essences. What is important for our purposes, however, is simply that an a priori analysis of the cognitive contents of the relevant concepts does not provide a reason for favoring their account over my own. The complaint leveled against Loar's use of the type-demonstrative strategy cannot be echoed here.

A similar situation arises if we consider another claim that is popular with contingentists. To introduce this second claim, all we have to do is shift from what David Lewis says about our knowledge of nondispositional properties (which we just discussed) to what he says about the metaphysical nature (or essence) of these properties. With regard to such properties, Lewis (Reference Lewis1986: 205) asserts that there ‘isn't much’ to their intrinsic natures (which I will view as being another way of saying there is not much to their non-power essences). Similar characterizations of the intrinsic non-power nature of properties are not hard to find: David Armstrong (Reference Armstrong1997), Robert Black (Reference Black2000), and Alexander Bird (Reference Bird2007) all think that if nondispositional properties exist, then they are best viewed as being ‘quiddities’ with extremely thin metaphysical natures. Armstrong (Reference Armstrong1997: 168), for instance, says that such properties (which he calls ‘universals’) are ‘merely numerically different from every other universal of that same –adicity’ while Black (Reference Black2000: 91) claims that ‘Just about all there is to a Humean fundamental quality is its identity with itself and its distinctness from other qualities’. (To be clear, both Black and Bird argue, in different ways, that nondispositional properties being quiddities is a reason to think that such properties do not exist. In contrast, I am happy to grant that properties could exist as quiddities. My interest resides in the implications that the claim that properties are quiddities has for what contingentists can say about how their concepts characterize the essences of those properties.)

Just like the first claim we examined, this second claim has significant ramifications for what the contingentists drawn to it can say about the cognitive contents of our concepts of the relevant properties. Unlike the previous group of contingentists, who claimed that the non-power essences of properties are unknowable, this group can claim that these non-power essences are knowable. They can even claim that our concepts provide a complete grasp (i.e., a complete characterization) of the non-power essences of the properties in question. But since they also claim that there just is not that much to the non-power essences of these properties, the cognitive contents of their concepts of these properties will still look a lot like the cognitive contents of type-demonstrative concepts of properties that thinly characterize their referents. If the metaphysical natures of these properties really are as thin as these contingentists say they are, then even concepts that provide a ‘thick’ epistemic grasp of these natures will still characterize various property-tokens of the same property-type as having the same non-power essence as one another while providing no positive characterizations of what that essence is like. To put the point another way, a complete (or ‘thick’) characterization of a metaphysically thin property can end up looking a lot like the thin characterization of a metaphysically thick property provided by a type-demonstrative concept.

Let us summarize the discussion so far. As we have seen, a common complaint against Loar's original use of the type-demonstrative strategy is that there is a substantial mismatch between what his account says the relevant cognitive contents should be like and what our a priori access to these cognitive contents actually reveals about them. Due to their commitments to other claims, however, many contingentists will be unable to echo this complaint against my use of type-demonstrative concepts. But perhaps there is another line of attack that my opponents could take in an effort to bring the complaint against Loar to bear in this new context. Earlier, I characterized dispositionalism as being a thesis about all natural properties, and in my examples of conceivability arguments I have focused on both shape (sphericality) and a microstructural property (S). It might be objected that when it comes to conceivability arguments and the richness of our ‘grasp’ of various properties, there is a significant difference between fundamental and nonfundamental natural properties. Consider, for instance, a hypothetical contingentist who maintains that while we have a thin grasp of fundamental properties, we can have a much thicker grasp of nonfundamental properties. If this were true, then it would be implausible for me to claim that we are using type-demonstrative concepts that provide thin characterizations of their referents when imagining the relevant things about nonfundamental natural properties.

There are a couple of things to say in response to this potential challenge. First, although I have characterized dispositionalism as applying to all natural properties, others have characterized it as applying only to fundamental natural properties (see, for instance, Bird Reference Bird2007). If I were to adopt this narrower characterization of dispositionalism, then the ability to imagine various scenarios involving nonfundamental properties would be irrelevant. That said, the literature on conceivability arguments against dispositionalism tends to focus on properties that are nonfundamental, so I will stick with the expanded characterization of dispositionalism, where the thesis applies to both fundamental and nonfundamental properties, and seek another way to respond to the current challenge.

As a first salvo, it is worth asking how psychologically plausible it is to claim that we imagine various things about, say, salt's solubility by imaging things about its molecular structure. But even if we imagine things about S in virtue of imagining things about its molecular structure (and we imagine things about sphericality in virtue of imagining things about its geometrical structure), that fact would not automatically give the contingentist what she needs in order to create trouble for my account. More specifically, it would not give the contingentist what she needs if our grasp of these various kinds of structure is ultimately mediated through our grasp of the powers associated with them. Suppose, for instance, that your grasp of the geometrical structure of sphericality is mediated by a cognitive content that says that objects with this property will pass through an appropriately sized circular hole, will leave a certain type of indentation if pressed into the sand, will typically cause human beings to have a certain type of visual experience, and will result in a subject's arriving at a specific number (zero) when counting the corners of the object in question. (As Molnar [Reference Molnar2003: 171] points out, shape necessitates the extent of contact-without-overlap that can occur between objects. This, in turn, is what necessitates the first two of the powers listed above. The classic source for the point about counting corners is Mellor [Reference Mellor1974]; an updated defense of that same idea can be found in Bird [Reference Bird2007].) If this is how we grasp the geometrical structure of sphericality, then our concept is not offering a non-powers characterization of that property. Of course, this does not establish that it is impossible for someone to have a substantial non-powers grasp of the internal geometrical structure of sphericality (or of S’s internal molecular structure). It does, however, show that just because someone grasps a property in virtue of grasping its geometrical (or molecular) structure does not entail that they have a non-powers grasp of that property.

4. Another Potential Defense of Dispositionalism from Conceivability Arguments

Let us consider another potential strategy that dispositionalists could adopt in response to conceivability arguments, a strategy inspired by a recent argument advanced by Carruth (Reference Carruth2016). Carruth uses a particular dispositionalist account of properties—the account defended by C. B. Martin and John Heil (among others)—in an effort to undermine David Chalmers's zombie argument against physicalism. As we saw in section 1, Martin and Heil occupy a dispositionalist position where a property is essentially linked to the powers it confers, but its essence is not exhausted by those powers. According to Martin and Heil, properties also have a ‘qualitative’ nature and the dispositionality and the qualitativity of a given property are equally fundamental to it; in fact, they go so far as to maintain that there is a ‘surprising identity’ between a property's dispositionality and its qualitativity. Although there is a lot more that could be said about this conception of properties, what is important for our purposes is how Carruth uses it in an attempt to undermine David Chalmers's zombie argument against physicalism.

According to that argument, zombies—i.e., beings that are physical duplicates of us but lack any phenomenal experiences—are conceivable. Carruth's focus is on what Chalmers (Reference Chalmers1996, Reference Chalmers, Gendler and Hawthorne2002, Reference Chalmers and Chalmers2010) dubs the ‘negative conceivability’ of this scenario, which involves the failure of a priori reasoning to reveal a contradiction within it. Following Chalmers (Reference Chalmers, Gendler and Hawthorne2002, Reference Chalmers and Chalmers2010), we can distinguish between a scenario's being prima facie negatively conceivable (i.e., its appearing negatively conceivable upon first glance), its being secunda facie negatively conceivable (i.e., its continuing to appear negatively conceivable after being subjected to sustained rational scrutiny), and its being ideally negatively conceivable (i.e., its being negatively conceivable after ideal rational reflection). Clearly, ideal conceivability is the gold standard for any conceivability argument. But since it is a standard that is challenging to meet, Chalmers (Reference Chalmers, Gendler and Hawthorne2002) posits that if a scenario is secunda facie negatively conceivable that counts as excellent evidence for thinking that it is also ideally negatively conceivable.

Carruth claims that the zombie scenario is not secunda facie negatively conceivable. More carefully, he claims that this scenario will not be secunda facie conceivable for anyone who is committed to Martin's and Heil's theory of properties. Why not? Because to subject the zombie scenario to rational scrutiny, part of what one needs to do is subject the notion of a ‘physical duplicate’ to rational scrutiny. Given that this notion is understood in terms of the instantiation of exactly similar physical properties, the rational scrutiny of the zombie scenario requires, among other things, that one possesses a positive conception of the nature of properties; in absence of possessing such a conception, one would not be in a position to assess rationally whether the notions of a physical duplicate and of a zombie are genuinely free of contradiction. But when using the Martin/Heil theory of properties to assess the relevant notions of properties, duplication, and zombies, a contradiction will be quickly uncovered for according to that theory no property can confer dispositionality upon its bearers without also conferring qualitativity upon them.

Arguably, the strategy adopted by Carruth to undermine Chalmers's zombie argument could also be adopted by the dispositionalist to defend her theory against conceivability arguments. Just like Carruth, our (hypothetical) dispositionalist could concede that the scenarios described in this paper are prima facie conceivable but then point out that, to reach the level of secunda facie conceivability, these imagined scenarios must survive sustained rational scrutiny. If in rationally scrutinizing those scenarios our dispositionalist is permitted to bring her preferred theory of properties to bear upon them, a contradiction will quickly be uncovered. As a result, these scenarios will not be secunda facie conceivable, and the challenge to dispositionalism will be blunted.

There are a number of things that could be said both about Carruth's strategy for using the Martin/Heil theory to undermine Chalmers's zombie argument and about the prospects of extending that strategy to the conceivability arguments discussed in this paper. As Carruth notes, there are different versions of Chalmers's argument, some of which do not invoke the secunda facie negative conceivability of zombies. It is unclear whether Carruth's approach can also undermine those other versions of the argument. In addition, there is way of reading Carruth where it is the zombie argument's reliance upon the notion of ‘duplication’ that requires one to adopt a theory of properties in rationally scrutinizing the zombie scenario. Since there is no notion of duplication at work in the conceivability arguments that are the focus of this paper, this is a potential disanalogy that could impact the ability to carry Carruth's strategy over to this other set of conceivability arguments. There is also the question of whether one really needs a detailed theory of properties—like the Martin/Heil theory—to ‘assess rationally’ the basic notion of duplication at work in the zombie argument.

The main concern I want to raise, however, is that a Carruth-inspired strategy for undermining conceivability arguments works a little too well. More specifically, the concern is that the same basic strategy could be used to save positions from conceivability arguments that, frankly, they should not be saved from. As Stoljar (Reference Stoljar2006) notes, if we are not going to be utter skeptics about the link between conceivability and possibility then we ought to be wary of a strategy that is meant to save a particular philosophical position from a particular conceivability argument that is capable of also ‘saving’ other philosophical positions from conceivability arguments that they should not be saved from.

To borrow one of Stoljar's examples, consider the ‘perfect actor’ conceivability argument against behaviorism. This argument begins with the claim that a ‘perfect actor duplicate’—someone behaviorally identical to you yet experientially different on the inside—is conceivable. From this, we are to infer that such a being is possible, which means that behaviorism is false. I suspect that most people who are not skeptics about the conceivability-possibility link would agree with Stoljar that this argument effectively establishes the possibility of a perfect actor and thereby provides a legitimate objection to the behaviorists’ attempt to identify mental properties with behavioral properties. Given this, we can follow Stoljar's lead and use the perfect actor argument as a litmus test for various strategies for undermining conceivability arguments: if a given strategy for undermining a conceivability argument against some position would also be capable of saving behaviorism from the perfect actor argument, then there is reason to be suspicious of that strategy. In short, there is reason to think that the strategy works a little too well. As homage to Stoljar, I will call this Stoljar's test. (Stoljar [Reference Stoljar2006: 54] refers to it as a ‘condition of adequacy’ that any method of defusing conceivability arguments must meet.)

To see if the strategy used by Carruth to undermine the zombie argument passes the test, consider a behaviorist who, in response to the perfect actor argument, claims that while such a scenario is prima facie negatively conceivable it is not secunda facie negatively conceivable. More specifically, our behaviorist claims that it is not secunda facie conceivable because in order to subject the perfect-actor-scenario to rational scrutiny, part of what one needs to do is subject the notion of a ‘behavioral duplicate’ to rational scrutiny. Given that this notion is understood in terms of the instantiation of exactly similar behavioral properties, rational scrutiny of the perfect-actor-scenario requires, among other things, that one applies a positive conception of the nature of properties to the scenario. However, according to the theory of (experiential) properties embraced by behaviorists, there will be a contradiction in the notion of a being that is behaviorally identical to you while being experientially different. So the scenario is not secunda facie after all!

As far as I can tell, the only difference between what this (hypothetical) behaviorist says in response to the perfect actor argument and what Carruth says in response to the zombie argument involves the theory of properties that they each bring to bear in rationally scrutinizing the relevant notion of a ‘duplicate’. If the former is legitimated in claiming that the zombie scenario is not secunda facie conceivable, then it seems the latter should be equally legitimated in claiming that the perfect actor scenario is not secunda facie conceivable. But this is an unacceptable outcome, for it saves behaviorism from the perfect actor argument. And for this reason, I submit that Carruth's strategy should be rejected. It fails Stoljar's test.

Those familiar with Stoljar's thoughts on this front will have noted a tension in my appeal to his work: namely, Stoljar uses the same ideas to argue against attempts to blunt conceivability arguments that appeal to type-demonstrative concepts! (See, for instance, Stoljar Reference Stoljar2006: 193–94.) Not surprisingly, I do not think Stoljar's argument works against the type-demonstrative strategy. Although I lack the space to go into much detail about it here, I think that any version of the type-demonstrative strategy that is capable of saving behaviorism from the perfect actor argument would be forced to make the cognitive content of the relevant concepts implausibly thin. More specifically, it would have to posit type-demonstrative concepts that fail to characterize their referents as being internal properties. I agree that such an account should be dismissed for its failure to pass Stoljar's test. A plausible account of these type-demonstrative concepts, however, will ‘thicken’ them with a bit more descriptive content—at the very least, the content of these type-demonstrative concepts will need to characterize experiential properties as being ‘that type of internal property’. An appeal to type-demonstrative concepts that carry this ‘thicker’ content would not be capable of saving Behaviorism from the perfect actor argument and thus would be capable of passing Stoljar's test. So while Stoljar is right to be wary of some versions of the type-demonstrative strategy, there are other, more plausible versions that would not save behaviorism from the perfect actor objection.