The Self-Knowledge of Combinatory States

1. SCS Explained & Motivated

Arguably, the most obvious way, sans testimony,⁸ in which one could know they token a particular CS is via SCS. According to the latter, we can know we possess a given CS by knowing we have the distinct states that constitute the latter, and then, on the basis of such knowledge, inferring that we possess the relevant CS. In order to know, for example, that we have a particular hope via SCS, we would first have to know that we have the relevant belief as well as the relevant desire.⁹ If an agent just had knowledge that she believed it’s possible, for instance, that the next supreme court justice nominee be an independent, that would not be enough for her to know she hopes they are. The agent would also need awareness of the requisite desire. Additionally, knowing one has a particular CS via SCS requires that agents also (i) grasp that the constituent states one has knowledge of constitute the relevant CS, and (ii) inferring that they have the relevant CS from grasping this fact.

With respect to (i), note that in order for there to be the appropriate inference made from the constituting states to the fact that one has the relevant CS, the agent must know that the constituting states are constituents of the relevant CS. This is because an agent S who has the relevant belief and desire might not know they have a particular hope if they lack knowledge of the fact that this combination of mental states just is a hope. In such a case, the agent in question wouldn’t make the relevant inference needed to acquire knowledge of the CS in question.¹⁰

Concerning (ii), even if (a) an agent S knows they have the relevant constituent states and (b) also knows these are the types of states involved in having the target CS, it might still be the case that S fails to recognize they have the latter. Perhaps they don’t entertain what is entailed by the fact that (a) and (b) obtain. Or perhaps they infer from (a) and (b) a mistaken conclusion despite knowing what hoping involves. Insofar, then, as one knows they have a given CS via SCS, they must also make the appropriate inference from their knowledge of the fact that they have the relevant constituting states.

Now unless one is a skeptic about self-knowledge in general, I think one should embrace the view that some of us can possess knowledge of our CS’s via SCS. After all, it seems that some of us can know we have the relevant non-combinatory states that constitute a given CS,¹¹ know that those states constitute the latter, and make the appropriate inference. The more precise details of how this occurs will be filled in, in part, by the epistemology one adopts with respect to non-CS’s. So, for example, if one thinks we know our non-combinatory states by following particular epistemic inferential rules, as Byrne (2018) does, then one will think that we can in principle know the states that constitute a CS by following such rules, and then making the appropriate inference from this knowledge base to the conclusion that we have the relevant CS.¹²

However, SCS seems to be a quite demanding way of knowing our CS’s. This is because the process not only requires that we have the requisite knowledge of the constituents of a given CS and that we make the appropriate inference from such a knowledge-base; it also requires that we have the relevant conceptual knowledge about what it is to, e.g., hope. Given this, one might think that in part because most agents appear to lack knowledge of what the nature of hope is, they are not going to possess knowledge via SCS. And if SCS really is the only way we know our CS’s, it would follow that most agents lack knowledge of their CS’s.

One might object to this line-of-reasoning, though, by maintaining that most of the folk can know they have a given CS by relying on implicit, sub-personal knowledge of the nature of e.g., hope. Such a suggestion involves the contention that agents have an unconscious, sub-personal theory about what hoping involves that they rely on in coming to know they have the relevant CS. And in defense of this claim, it might be maintained that most of us rely on implicit theories—theories we are not cognizant of—to help us navigate the world. An implicit theory about what the nature of a given CS is might be one of them. If this is the case, then perhaps the folk do have the conceptual knowledge needed to know their CS’s via SCS.

In response to this suggestion, I am willing to grant for the sake of argument that many of us do have implicit knowledge of the nature of, e.g., hope, knowledge that is sub-personal in nature and that we are unaware of. It is important to note, though, that such implicit knowledge would not be sufficient on its own to afford us with self-knowledge of our CS’s. Insofar as this proposal is in keeping with SCS, we also need to (i) possess self-knowledge of the constituents that make up the relevant CS, (ii) be able to apply the implicit conceptual knowledge in the appropriate way to make the relevant inference to the conclusion that we have a given CS, and (iii) actually make such an inference and form the belief that we have a particular CS. And it seems that (i)–(iii) would have to occur sub-personally as well.

This is because if (i)–(iii) did not occur sub-personally, a number of us would have a more explicit grasp of what hoping involves than it seems reasonable to think we have. In defense of this claim, consider, for example, the possibility that we have non-implicit knowledge of the constituent states—knowledge that at the very least involves awareness of the mental states that we have knowledge of—and yet our grasp of what hoping is, our application of this conceptual knowledge, and the inference we make from the constituent states to the relevant CS is implicit in nature. If this were the case it would entail that while we consciously know we have a given belief and desire, we would unconsciously apply our conceptual knowledge about hoping, infer that we have the relevant hope from our knowledge of the constituent states, and end up with the conscious belief that we have such a hope.

But it seems reasonable to think that a number of us, upon reflection, could fairly easily fill in the missing steps in this process and recognize that we think we have a given hope because we have the relevant belief and desire. This recognition would enable us to fairly easily conclude that hope is a matter of believing and desiring. And if that is the case, then a number of us would have a more reflective, conscious grasp of what the nature of hope is than it seems reasonable to think we in fact have. We have good reason to believe, then, that on this proposal, the steps involved in this process that are supposed to terminate in self-knowledge take place sub-personally; they take place, that is, in a way we are not cognizant of.

Insofar as the steps in this process are all sub-personal in nature, there is a problem for this implicit theory-based proposal—a proposal I call IA-SCS. The problem is that according to proponents of IA-SCS, we would possess implicit self-knowledge of the constituents of a given CS, self-knowledge that on this view serves as the foundation for making the appropriate inferences. I will now argue, though, that the alleged knowledge in question would not in fact amount to knowledge at all.

One major reason for thinking this is the case is that if we have self-knowledge of the constituents of the relevant CS, it seems we would at least be aware of the object of such knowledge. Self-knowledge of M, in other words, suffices for our awareness of M. Not only is this claim intuitively plausible (some would maintain, analytically true¹³), the most widely-discussed accounts of self-knowledge in the literature entail it. Indeed, approaches as markedly distinct as constitutivist views of self-knowledge, neo-expressivist views of self-knowledge, and inner sense theories of self-knowledge all have the result that when we possess self-knowledge of M we are aware of M.¹⁴ In fact, a number of theories of self-knowledge rely on the claim that we have awareness of the mental state we have knowledge of to explain how we have such knowledge.¹⁵

On this proposal, though, no such awareness accompanies our self-knowledge of the state in question. If it did, then all the major steps in the process would not be implicit, and we would, as defended above, have a more reflective, explicit understanding of the nature of CS’s than most of us in fact do. IA-SCS, then, posits a type of “self-knowledge,” that runs counter to what having self-knowledge involves. We therefore have reason to be skeptical of IA-SCS.

I have just defended the view that self-knowledge of a mental state M entails awareness of M. However, even if this claim is mistaken, it is important to note that in order for IA-SCS to be plausible, it would have to be the case that when most of us possess self-knowledge of the constituents of a given CS, we typically lack awareness of these constituent states. In other words, it would have to be the case that even if self-knowledge of M does not entail awareness of M, knowing that we have the relevant constituent states very rarely makes us aware of these states. This is because if we were on occasion aware that we have the constituents of a given CS, it would be fairly undemanding for us to infer what the nature of a given CS is (by filling in the missing steps of the process). So even if self-knowledge of M doesn’t guarantee that we have awareness of M, whatever explanation a proponent of IA-SCS posits for how we know our constituent states would have to have the result that we very rarely possess such awareness. And it thus seems reasonable to think that whatever means are posited by proponents of IA-SCS for how we know these constituent states is going to be an implausible one. Such self-knowledge would at least make us aware on occasion of the given mental state in question.

An alternative way of developing the same problem for IA-SCS is to note that in order to provide a fully satisfactory explanation of how we know our CS’s, a proponent of this view needs to explain how it is we know the relevant non-CS’s that compose a given CS. The epistemology adopted, as discussed above, can be filled in, in various ways. But whatever epistemology is embraced, it would need to be the case that however we know these states during this process, we would at least rarely be aware of the object of our knowledge. Again, the reason is that if we were aware of these states, it would be fairly easy, it seems, to simply fill in the missing links of the process and arrive at the conclusion that we believe we have a given hope because we have the given belief and desire. And then we would have a more explicit, conscious grasp of what the nature of hope is than most of us do. That being the case, it seems reasonable to hold that any plausible epistemology of these non-CS states would have the result that we are at least occasionally cognizant of the fact that we have the non-CS states in question. Given this, we have good reason not to posit IA-SCS as the explanation of how the folk possess knowledge of their CS’s.¹⁶

The upshot of this discussion is that if SCS is the only way of knowing our CS’s then most of us lack self-knowledge of our CS’s. And insofar as we want to avoid skepticism about such self-knowledge, we should, I think, be inclined to consider alternative, non-SCS ways of knowing our CS’s.

The following puzzle, consisting of three jointly inconsistent propositions, is a precise way of understanding the problem:

1. SCS is the only non-testimonial way to know our CS’s.

2. If SCS is the only non-testimonial way to know our CS’s, then most agents fail to possess knowledge of most (if not all of)¹⁷ their CS’s.

3. Most agents do not fail to possess knowledge of most (if not all) of their CSs.

I just offered a defense of (ii) of this puzzle. And while I will not unqualifiedly defend (iii) in this paper, I take it that (iii) is plausible (and almost all anti-skeptics about self-knowledge will want to embrace it). That leaves (i).

The latter is the part of the puzzle I am most inclined to reject and I suspect I am not alone here.¹⁸ Fortunately, I think we have good reason to reject (i). More cautiously, we have good reason to reject the latter if particular extrospectionist views of self-knowledge are plausible with respect to non-CS’s. I turn now to defending this claim.

2. A Non-SCS Extrospectionist Epistemology of Combinatory States

In recent work, Byrne (2018) defends the view that we know facts about our minds, not by looking inward, as the introspectionist has it, but rather by “looking outward” to facts about the world.¹⁹ More specifically, Byrne holds that we know the contents of our minds by following²⁰ particular epistemic rules that involve making inferences from propositions about the world to propositions about the contents of our mental states. So, for instance, Byrne thinks that in the typical case we know what we believe by conforming to the following rule:

BEL: If p, believe that you believe that p.

To determine whether we believe a particular proposition, then, we need to consider whether the target proposition, p, is true. The latter is a fact about the world, not about our minds. BEL, according to Byrne, affords us with “safe” beliefs or beliefs that “could not easily have been false” (110) because following BEL involves believing that one believes that p because one recognizes that the antecedent obtains. But if one recognizes and hence knows that p, then one believes that p. So, following BEL ensures that one’s belief that one believes that p will be true.

Additionally, Byrne thinks we know we token particular desires by conforming to the following rule:

DES: If Φ-ing is a desirable option, believe that you want to Φ.

Insofar as there is an intimate connection between Φ-ing being desirable²¹ and one wanting to Φ, and Byrne thinks there is, then DES, like BEL, will be a safe rule.

Now Byrne could claim one knows their hopes or their intentions, if these states are combinations of beliefs and desires, by first arriving at knowledge of one’s beliefs and desires via following BEL and DES and then making the appropriate inference. In other words, Byrne could invoke a BEL-DES inspired-version of SCS to offer an epistemology of these states. But there is good reason to think Byrne can offer epistemic rules that are in keeping with his program that afford us with non-SCS self-knowledge of these states.²² Before offering one such rule, I want to make explicit a particular feature of Byrne’s account of self-knowledge that helps explain why his approach to self-knowledge is particularly well-suited to avoid SCS. The feature in question involves the fact that, on this approach, one needs to know very little about the nature of belief and desire to possess self-knowledge of these states.

Consider Byrne’s epistemology of desire first. According to the latter, the very most one needs to know in order to arrive at knowledge of her desires is that there is an intimate connection between desirability and desiring. Because Byrne understands desirability in terms of features that make a given phenomenon good,²³ it seems fairly clear that one can have little grasp of the full nature of this attitude and yet possess knowledge that they have a given desire by following DES. For while there might be an intimate connection between desirability and desiring, x being desirable is not the same as x being desired; nor, as Byrne himself would acknowledge, is desirability a necessary feature of desiring. The upshot is that following Byrne’s DES requires very little awareness on the part of the knower about what the nature of desire is.

Now it might be pointed out that some embrace the view that to desire x just is to judge that x is good or desirable.²⁴ But even if this view of desire is correct, it is important to note that the agent who follows DES need not connect up judging that something is desirable with desiring. This is because on Byrne’s view one does not need to make an inference from (i) I judge that Φ-ing is a desirable option to (ii) I want to Φ. Such an inference is cumbersome and unnecessary. Instead, all one needs to do is make the inference from Φ-ing being a desirable option to wanting to Φ. It follows then that insofar as agents are consciously following DES, they just need to grasp that desirability is connected up with desiring, not that judgments about desirability are connected up with their having a particular desire.

The above discussion segues into a further important point about agents’ awareness of following rules like DES. Given the thoroughly externalist framework Byrne is operating with, it seems clear that agents do not need to consciously grasp that they are following one of Byrne’s rules in order to possess self-knowledge by way of doing so. The process can instead be a matter of brute causality where an agent’s judgment that Φ-ing is desirable causes them to believe they have a given desire. Part of this process, it can be held, could take place sub-personally and without the agent’s grasp that they are following a particular rule. For our purposes, the important point to note is that agents do not need to consciously adopt and conform to Byrne’s inferential rules in order to come to possess self-knowledge of the relevant state via following them.

A number of the comments made above about DES apply to BEL as well. Concerning the latter, an agent does not need to know that there is an intimate connection between judging that p and believing that p. All they need to be aware of, if they are consciously following the rule, is that there is an intimate connection between p being true and believing that p. Moreover, one’s judgment that p can, I think, cause one to believe that they believe that p where such brute causality takes place in part, at the sub-personal level. Given this discussion, we are now in a position to see why Byrne’s account is well-suited for avoiding to SCS.

Let’s begin with a rule Byrne could invoke for how we know our hopes. Recall that the standard view has it that hoping that some outcome E occurs is a matter of wanting E to occur and believing that E occurring is possible (but has yet to obtain). If this view of hoping is correct, then Byrne could posit the following rule for how we know our hopes:

HOP: If Φ could happen and it would be a good thing if it does, believe you hope that Φ.

Note first about HOP that if one judges that Φ could happen, it’s reasonable to think such a person believes that Φ is possible.²⁵ Moreover, if Byrne is correct about the connection between judgments concerning desirability and desiring, then if one judges that it is a good thing, or desirable, that Φ occurs,²⁶ then it’s reasonable to think they also have the desire that Φ. Lastly, one who judges that the antecedent of HOP obtains, I think, could reasonably be taken to believe that Φ has yet to occur. After all, the implication of one who asserted the antecedent of the above rule is that Φ has yet to occur, and one conforming to HOP would, in the typical case I think, embrace this implication. If the above is close to the mark, then HOP, it seems, is going to lead to safe beliefs about what one hopes.

One might wonder, however, why we couldn’t adopt a simpler, more elegant rule concerning how we know our hopes such as the following:

HOP*: If it is a good thing for Φ to occur in the future, believe you hope that Φ.

The reason HOP* is flawed is because there are a number of cases in which one could judge that it is a good thing for Φ to occur in the future without believing that Φ occurring in the future is possible. One might, for example, judge that it would be a good thing for our bodies to be resurrected at the time of death and survive in an afterlife without believing it to be possible that this could occur. Given this, there are a number of cases where one could conform to HOP* without arriving at knowledge that they have a particular hope (as opposed to a particular wish).

Although more can be discussed with respect to HOP²⁷ what I want to argue next is that insofar as this rule does afford us with knowledge of our hopes, it can also afford us with non-SCS knowledge of the latter. The previous discussion of BEL and DES already indicates why. One can, it seems, judge that Φ is a good thing without recognizing that one of the constituents of hoping is the desire that Φ. Moreover, one can judge that Φ is a good thing without being aware that such a judgment makes it likely that one has a particular desire. All that needs to be grasped, if one is consciously following HOP, is that there is an intimate connection between Φ obtaining being a good thing and hoping that Φ. In other words, one does not need to know that desire is one of the constituents of hopes in order to know their hopes. Moreover, one does not need to know that hoping involves the belief that Φ is possible but has yet to occur, in order to know their hopes. All one has to be aware of is that it being possible that Φ occurs is connected up with hoping. We have, here, then, a non-SCS extrospectionist explanation of how we know our hopes.

It might be objected, however, that while it might be true that one can follow HOP without being aware that hopes are a combination of beliefs and desires, it’s reasonable to think that one would only conform to this rule by having a more sophisticated understanding of what hoping involves than we commonly think the folk possess. In particular, one might object that the folk would not consciously follow HOP unless they thought that hoping that Φ requires believing that Φ is possible. And this is too much to expect the hoi polloi to know about hoping.

In response to this concern, it doesn’t strike me as unreasonable to think agents without a thorough background concerning the nature of attitudes implicitly grasp that for one to hope that Φ, Φ must at least be possible. In defense of this claim, consider the following example: Sam tells Isabella that she hopes to one day travel back in time to explore ancient Greece. If Isabella were aware that Sam believes that time travel is not possible, it wouldn’t be unreasonable to expect her to respond to the strangeness of Sam’s assertion, if not outright deny that Sam lacks the hope in question. Examples like this one indicate that it is not implausible to think that folk without a nuanced understanding of what hopes are could nevertheless follow HOP.²⁸

Onto intentions. Unlike hopes, Byrne does offer an epistemology of intentions. But because he holds the view that intentions are not a combination of beliefs and desires, and specifically that one can have an intention without having a relevant desire,²⁹ Byrne’s proposed inferential rule for intentions is one that cannot reasonably be embraced by friends of the view that intentions are CS’s. More specifically, Byrne’s epistemology of intentions is not one that can be reasonably embraced by proponents of the view that intentions are a combination of the belief that one will bring about a particular state of affairs A and the desire that A obtain.³⁰ Indeed, the rule Byrne proposes for how we know our intentions is the following:

INT: If you will Φ, believe that you intend to Φ.

If intentions involve wanting Φ to obtain, as proponents of the view that intentions are a combination of beliefs and desires maintain, then INT will be an unsafe rule. After all, it is quite common for us to believe that we will do x without wanting to do x. I might, for instance, believe that I will miss a free throw without wanting to miss it. Likewise, a public speaker might believe that they will flub a line without wanting to flub a line. Therefore, if, pace Byrne, intentions are a combination of the relevant beliefs and desires, we need an alternative rule to explain how we know these states.

I suggest replacing INT with the following rule:

INT-CS: If Φ will be brought about by myself in the future, then believe I intend to Ψ.

Read ‘Φ’ in the antecedent as a placeholder for an event like one’s lawn being mown and ‘Ψ› as an action that is supposed to be relevant to the bringing about of the event referenced in the antecedent (e.g., the mowing of a lawn).

Next note that if we accept, as a number of those who embrace the view that intentions are CS’s do, that one cannot bring about a particular state of affairs via one of their actions without having the relevant desire, then insofar as one follows INT-CS, it will be the case that one judges accurately that they have the relevant desire. This is because one judges accurately that they will bring about the state of affairs in question and they couldn’t do this without having the relevant desire. Additionally, when one conforms to INT-CS, one will have the relevant belief required in order to have the relevant intention. This is the case because in conforming to INT-CS one judges accurately that they will bring about a particular event.³¹ Thus, assuming that a Humean approach to action³² is close to the mark, it will be the case that when INT-CS is followed, the person will have the relevant belief and desire required to have an intention.

It should be acknowledged, though, that there are cases in which one would not be inclined to follow INT-CS despite seeming to have the relevant intention and we need an explanation of why this is the case. Phil Connors, the protagonist of the film Groundhog Day, might believe that he will bring it about that he experiences February 2^nd at least one more time by prematurely ending his experience of the current day, without it being the case that Connors would infer that he intends to perform some relevant action—e.g., jumping off a high-rise building—to bring this event about.³³ Relatedly, Jonathan Bennett’s (1981) tactical bomber might believe that he will cause the death of civilians in bombing a factory, but not be inclined to infer on the basis of this judgment that he intends to perform some action that will bring about these deaths. These agents, it might be thought, would judge that Φ will be brought about by them in the future, but are not inclined to believe on the basis of this judgment that they intend to Ψ. And we need an explanation for why this is the case. Given that such cases also pose a challenge to INT, Byrne too needs to explain what is going on in them.³⁴

He attempts to handle the latter by invoking a defeater rule designed to explain why the above agents wouldn’t follow INT in the relevant circumstances. The rule in question involves the contention that an agent who has strong observational evidence that they will do a particular thing A will not follow INT and infer that they have the intention to do A.³⁵ To use an example discussed by Byrne, an agent who has strong evidence that they will wear down a pair of running shoes, evidence that consists in knowing they will be wearing those shoes in a marathon tomorrow, will not in turn infer that they intend to do this. I am skeptical, however, that this defeater rule as it applies to INT does adequately account for these problem cases.³⁶ Given this, as well as the fact that we are working with a different account of intention than Byrne is, I think we should consider embracing a different defeater rule for INT-CS, one that is in keeping with Byrne’s approach. And fortunately, I think there is one in the offing.

The rule in question is that in cases in which an agent does not judge Φ to be desirable, they will refrain from inferring that they intend to Ψ. This rule conforms with Byrne’s program given that he thinks there is a strong connection between Φ being desirable and Φ being desired.³⁷ If there is such a connection, then it seems reasonable to hold that those who don’t view Φ as being desirable will be disinclined to follow INT-CS. And I think this is indeed what we find in the cases of Phil Connors and Bennett’s pilot. Neither of them finds the relevant Φ desirable, and this seems to explain why they wouldn’t infer that they intend to Ψ. Connors, for example, doesn’t find it desirable to relive February 2^nd, and so he likely would not infer that he intends to bring this event about by performing the relevant action.

This explanation of why those in circumstances like Connors’ and Bennett’s pilot would not follow INT-CS, though, only solves part of the problem. This is because Connors and the pilot are going to perform the relevant actions in question and appear to intend to perform them. Connors, for instance, does intend to jump off the high-rise building. But I just argued that he wouldn’t infer that he has the relevant intention if he doesn’t find reliving February 2^nd over again to be desirable. Is it the case then, that Connors simply won’t be able to know that he has the relevant intention to jump off this building via INT-CS? The answer I think is ‘no.’ To see why we have to get clear on what the event in the antecedent is. In the case of Connors, if the event is something like the end of his existence on this present day, then Connors might very well find this event desirable. Given this, insofar as he judges that he will end his existence on this current day, he is likely to infer that he will jump off the high-rise building to bring this event about. Connors, then, can know via INT-CS—assuming Byrne’s approach to self-knowledge is correct—that he has the relevant intention.

The above might seem like a somewhat awkward, ad hoc way of dealing with these cases. But it is important to note that oftentimes we intend to perform actions without desiring some of their likely consequences, and without intending to bring them about. Connors might very well intend to end his existence on the day in question without also desiring one of the likely consequences of doing so—viz., having to relive February 2^nd again. Bennett’s tactical bomber might very well intend to bomb the factory without desiring the deaths of the civilians that are likely to result from him doing so. Both of them, in other words, do not intend to bring about some of the likely consequences of their actions. I don’t think that my explanation of these cases, then, is awkward or ad hoc. Our actional lives are messy; the above account accommodates such messiness. I think, then, that these cases fail to pose a serious problem for INT-CS even if they do pose a problem for INT.

Note next that as with the following of HOP, one can follow INT-CS without being aware that intentions are a combination of beliefs and desires. All they would have to know is that insofar as some event is likely to be brought about by them, they should infer that they have the intention to perform an action relevant to bringing about that event. An agent would not need to recognize, for instance, that performing some action requires having a particular desire. Nor would they consciously need to know that intentions involve particular beliefs about what one will do. It follows that if conforming to INT-CS does afford us with knowledge of our intentions, then we could know we token these states in a non-SCS manner.³⁸

The results of this section concerning hopes and intentions are significant. As discussed above, it is plausible that many agents, not just the folk, would lack knowledge of their CS’s if SCS were the only way they could know they have these states. If agents can know they token a given hope or intention by following the above rules, then we have an explanation of how agents can know their hopes and intentions that is non-SCS in nature.

It seems, however, that the non-SCS epistemology of hopes and intentions advanced here will only be plausible if Byrne’s approach to self-knowledge in general is plausible. We have been assuming that it is. But needless to say, there have been a number of criticisms leveled against it.³⁹ Some of these criticisms involve concerns about the very features of Byrne’s view that make it amenable to delivering a non-SCS epistemology of CS’s, (e.g., the view explaining at least some of our self-knowledge as being the result of brutely and unconsciously following inferential rules most of us are not aware of). I will not dwell on these criticisms here. What I want to instead defend next, is that if one rejects the extrospectionist approach to self-knowledge Byrne embraces in favor of its most natural opposition—an introspectionist approach to self-knowledge—it is going to be much more challenging to make sense of how we can come to possess non-SCS knowledge of our CS’s. I turn to that task now.

3. Introspecting Combinatory States

Unlike extrospectionist approaches to self-knowledge, introspective approaches to self-knowledge construe knowing the contents of our minds as involving a process of “looking inward.” Removal of the scare quotes typically involves invocation of phenomena like observation, attention, and consciousness, which, with proper clarification, are supposed to explain how we have a more direct grasp of the contents of our minds than accounts like Byrne’s afford. More specifically, according to a number of introspective accounts of self-knowledge, we either directly “grasp” the mental state M itself or essential features of M via a process of introspection. But it is precisely because the introspectionist construes the introspective process as involving agents standing in a more direct relationship to what they know that it becomes more difficult to see how we can have non-SCS knowledge of our CS’s. In order to see why, it will help to look at several specific introspective accounts of self-knowledge. Let’s begin with so-called inner sense accounts of self-knowledge.

These accounts, defended by D.M. Armstrong (1968), William Lycan (1996) and more recently Alvin Goldman (2006), entail that we achieve self-knowledge via an observational causal process similar to the type of visual process many think we undergo to arrive at knowledge of the external world. On a fairly standard understanding of the view, this causal process involves becoming aware of a mental state M via attention mechanisms in the brain “scanning” M, an awareness that in turn causes one to judge that she is in M.⁴⁰ So, an agent might form the belief that she has a particular desire by first becoming aware of this attitude via a scanning process in the brain. This view of the mechanics of self-knowledge is typically coupled with a version of epistemic externalism, (e.g., a version of reliabilism), to explain how we possess knowledge that we have, for instance, a particular desire.⁴¹

Although truncated, the above remarks suffice, I believe, to see why proponents of these approaches have a difficult time offering a plausible non-SCS account of our CS’s. Recall first that CS’s such as hopes are, on the standard views, combinations of beliefs and desires. Given this, to offer a non-SCS inner sense explanation of our self-knowledge of hopes, the inner sense theorist would have to posit an attention mechanism that makes us aware of the fact that we have a given hope H, without also making us aware of the fact that we are tokening the relevant belief B and desire D.⁴² But positing an attention mechanism, or scanning process, that makes us aware of H without making us aware of B and D seems problematic. This is because in the case in which we just possess B but not D, our attention mechanisms, if reliable, would simply make us aware of B. Why should we think, then, that when D, a distinct state, is tokened, a different attention mechanism⁴³ “kicks in” and makes us aware of the fact that we have H, all while bypassing making us aware of B and D? First, positing such a sophisticated, (some would say magical), faculty is uneconomical. It is uneconomical because the inner sense theorist could explain our self-knowledge of CS’s by simply positing attention scanners that afford us with knowledge of the constituents of such states, knowledge that serves as the base for making the appropriate inference.

Second, such a view would also make self-knowledge of our CS’s a very facile process that seems to belie the fact that knowing we token one of these states takes a modicum of cognitive effort on our part. Knowing what we hope, if we had such mechanisms, it seems, would be as direct and easy as knowing we are in pain. But given that the latter is a simple sensation and the former is complex state composed of (arguably) dispositional states, it seems implausible to think the self-knowledge of these two distinct types of states would be in these respects so similar. Therefore, while it is, at the end of the day, an empirical question whether there are such attention mechanisms, it would be surprising, I think, for us to discover them. The upshot of this discussion, then, is that it is going to be surprising if there is an inner sense-friendly, plausible non-SCS account of the self-knowledge of combinatory states in the offing.⁴⁴

A similar conclusion can be reached with respect to a markedly different introspective approach to self-knowledge—viz. Laurence Bonjour’s (2003) internalist account. Bonjour’s approach is grounded in the claim that some attitudes are conscious states. The consciousness Bonjour thinks occurrent attitudes such as beliefs and desires possess is awareness consciousness or what Lycan (1996) calls state/event consciousness.⁴⁵ Bonjour holds that not only do we have awareness of our occurrent attitudes, but that we have a “built-in constitutive awareness” (hereafter, “BCA”) of these states. What Bonjour means by this is that the awareness one has of (conscious) attitudes is awareness that: (i) is not achieved by a second-order mental act that represents the first-order state and (ii) in part makes the attitude the state it is.⁴⁶ This BCA, according to Bonjour, allows us to be aware of both the content of the state in question as well as the attitudinal mode that we adopt towards that content.⁴⁷

Note that on Bonjour’s view, because the awareness of the content and attitudinal mode we adopt to that content is intrinsic and (at least partly constitutive) of conscious attitudes, we have such consciousness solely in virtue of tokening such states. In other words, to possess the state suffices for having such awareness because the state wouldn’t be the state it is without this built-in constitutive awareness.⁴⁸ Moreover, Bonjour holds that because such awareness is intrinsic and constitutive, it is not itself in need of justification, nor, according to him, does it admit of any. In fact, he claims that given that we have this constitutive awareness, our awareness of the content of such attitudes is “strictly infallible” (2003: 64). However, this does not mean he thinks such awareness on its own suffices for possessing self-knowledge. Knowledge entails belief, and the awareness is, according to him, not a belief.⁴⁹ Still, Bonjour argues that having this awareness puts us in a prime position to know (or at least be justified in believing) that we have a particular attitude with a specific content.⁵⁰

BAC does so, according to him, by serving as the justification for a second-order belief that describes the first-order attitude in question. Bonjour summarizes his approach to the self-knowledge of attitudes, focusing on belief, as follows:

It is by appeal to [the] intrinsic, constitutive awareness of the propositional and assertive content of the first-level belief or thought that the second-level meta-belief can be justified. Such a constitutive awareness of content seems obviously enough to constitute in and by itself, at least if other things are equal, a reason for thinking that the second-level belief that I have an occurrent belief with that very content is true… The way in which this works… is that the meta-belief [second-order belief] is a description of the very content involved in the constitutive awareness of content so that by consciously having that constitutive awareness, I am in an ideal position to judge whether or not this description is true. (2003: 64)

On Bonjour’s internalist approach, then, when an agent forms a second-order judgment about her first-order attitude, she is able to determine that this second-order judgment is accurate. She accomplishes this by comparing her second-order judgment with the infallible awareness she has of the content of the first-order state. The built-in constitutive awareness of conscious attitudes, then, puts her in an ideal position to determine that the second-order description of that belief is accurate.

Given such a view of the epistemology of our attitudes, we can see why Bonjour’s account comports best with SCS. Recall that on Bonjour’s approach, we simply cannot possess occurrent beliefs and desires without being aware of them. But if that’s the case, then insofar as hopes H just are a combination of beliefs B and desires D, then in knowing we have H we are also going to be aware we possess B and D. And such awareness of B and D on its own is not going to afford us with knowledge of the fact that we possess H without our knowing that B and D just are constituents of H. If we do recognize this fact, then we could use such recognition to make the appropriate inference to the conclusion that we have H. But this would just be a Bonjour-friendly SCS way of knowing we token H.

The only way I can see for how SCS, on this approach, can be avoided, is by positing that H has its own separate BAC that enables one who has this state to be aware of this fact, and hence bypass the need to rely on the awareness of B and D to know they have H. But given that H just is B and D, positing such awareness is problematic. This is because such awareness, on Bonjour’s view, would have to consist in awareness of the content of the attitude and the attitudinal mode. But the attitudinal mode of hoping isn’t some separate attitudinal mode from believing and desiring (on the standard view). The attitudinal mode of hoping just is a combination of believing and desiring. And awareness of this attitudinal mode would, it seems, return us to the above SCS-based story of how we know our hopes given Bonjour’s epistemology. Even more so than the inner sense approach discussed above, then, Bonjour’s epistemology of the attitudes is going to sit neatly in the SCS camp.

We have examined two introspective accounts of self-knowledge and seen the challenges they face in avoiding SCS. There is, however, an alternative introspectionist approach besides the two approaches just discussed, an approach grounded in phenomenal consciousness or the felt quality some mental states are thought to possess. Indeed, in recent work on the self-knowledge of attitudes, I have defended the view that it’s possible for some people to possess introspective knowledge of their attitudes via attending to the phenomenology some attitudes possess.⁵¹ I will here provide only a summary of it, one that will allow us to determine whether such an account has the resources to avoid SCS.

The epistemology of attitudes in question involves a process I call phenomenal simulation. The process involves an agent entertaining a given mental representation⁵² and in the case of desires, focusing on whether she has a felt attraction to it. Such attraction, as I have argued elsewhere, is the phenomenology that is partly constitutive of desiring and enables us to type identify these mental states.⁵³ When an agent is attracted to the content in question, that just is what is involved in having a particular occurrent desire. So, the process of phenomenal simulation involves, in some cases, the creation of a desire, a creation that puts the agent in a prime position to know they token such a state. But since one can produce an occurrent attitude without knowing they possess it, phenomenal simulation is only part of the story of how we know such attitudes.

We come to possess knowledge of the output of phenomenal simulation by scrupulously attending to it in such a way that we are able to form judgments about the states in question that are partially constituted by those very states. Such a process involves “bracketing out” relational information about such phenomenology and focusing on the latter itself to the extent that the way the phenomenal state appears to the agent is the way it is.⁵⁴ Moreover, because the judgment itself is partially constituted by that which the judgment is about, the agent is going to accurately conceptualize and classify the first-order phenomenology as the phenomenology it is. And given an awareness of the content in question, she is going to form an accurate judgment of it.⁵⁵ This type of reliability, however, on the view I favor, does not suffice for knowledge—what does is having direct awareness of one’s belief that they have a particular mental state, direct awareness of the latter itself and direct awareness of the correspondence that obtains between the two.⁵⁶

Given the above outline, we are now in a position to determine whether such a view, if tenable with respect to non-CS’s such as desires and beliefs, enables us to avoid SCS. As with the above two approaches, the view, admittedly, seems to sit more comfortably in the SCS camp. There are, however, at least two ways such a view could avoid SCS. The first way depends on there being the relevant type of combinatory phenomenology. The latter is my term for the combination of two or more distinct phenomenal properties that are both tokened at the same time by an agent and can thus be thought of as constituting⁵⁷ one phenomenal state of the agent at a given time. There is nothing strange, I think, about combinatory phenomenology in and of itself. Anyone who thinks that we can instantiate multiple, distinct phenomenal properties at a given point in time should think that there is such phenomenology.⁵⁸ Next note that the felt quality of one’s total phenomenal state that contains attraction, and say, the phenomenology of judging, is going to feel differently than the phenomenology that just contains attraction.

Given the existence of such combinatory phenomenology, we can ask the following: is it possible for (a) one to grasp a phenomenal state constituted in part by the phenomenology of attraction and judging, (b) not be aware (or at least not be attentive to the fact) that such phenomenology is the phenomenology of desiring and judging, but (c) know that such phenomenology is what hoping feels like? One reason we might think that the answer is ‘yes’ is because there are numerous examples in which we appear to be aware of x, unaware or not attentive to the fact that x is composed of y, z, etc. and yet still know that the thing in question is x. For example, it seems that Larry can know that the dish in front of him is moussaka without knowing or attending to all the ingredients of the dish.⁵⁹ Similarly, I can know, it seems, that a painting is purple without knowing or attending to the fact that purple is a combination of red and blue. Admittedly, the above examples are not cases in which phenomenology is directly involved. But they do, I think, lend support to the view that (a)-(c) can simultaneously obtain. And if they can, then we should think that the epistemology of attitudes I favor has the resources to provide a non-SCS way of knowing our CS’s.

A further reason for thinking that (a)–(c) can be realized together is the fact that a phenomenal state composed of the phenomenology of desiring and judging is going to feel markedly distinct from a phenomenal state that is composed of just the phenomenology of judging or just the phenomenology of desiring. Given this, it seems possible for an agent to attend to the feeling of hoping without being attentive to the fact that the phenomenology in question is composed of the phenomenology of judging and desiring. Something analogous, it seems, might occur when one grasps that the olfactory sensation they are tokening is a sensation of, say, smelling cinnamon, even if one is not aware of the fact that such a sensation just is a combination of multiple different experiences.⁶⁰

In addition to the above suggestion there is one other more controversial way for how the epistemology of attitudes I favor can avoid SCS. The way in question involves the possibility that some CS’s are partially constituted⁶¹ by additional phenomenology than the phenomenology we are assuming that, for instance, hoping possesses. Some, such as Luc Bovens (1999), have proposed that hoping is constituted by not just a belief and a desire, but a third constituent as well.⁶² If this further component itself possesses phenomenology, phenomenology that is indicative of the attitude in question, it might be possible to grasp that one tokens the attitude in question on the basis of just some of the combinatory phenomenology in question. In order for this suggestion to be plausible, it would, I think, have to be the case that whatever additional phenomenology hoping might in part be constituted by, it must be a distinctive enough phenomenology to enable the agent to bypass having to be aware of some other phenomenal elements that we are assuming are constitutive of hoping.⁶³

Admittedly, it is controversial whether any attitude possesses the kind of type-individuating phenomenology that on my view states like (occurrent) desire and judgment do, let alone that CS’s have an additional type of phenomenology. But insofar as some CS’s do possess the latter, this opens up an alternative way of making sense of how we can introspectively know our combinatory states in a non-SCS manner. In order to fully defend this alternative non-SCS means, of course, we would need to engage in a more thorough, contentious debate about the nature of CS’s. I will not be engaging in such a debate here. What I hope to have accomplished instead is the more modest goal of underscoring this possibility, and more generally, demonstrating the importance of having a fully sufficient account of the nature of states like hoping.

Conclusion

Above, I worked to establish a foundation for an epistemology of combinatory states. I did so by outlining the most obvious way in which we could know these states, a way I denoted the self-knowledge of constituting states means or SCS. I in turn pointed up the limitations of SCS and explored alternative non-SCS ways of knowing our CS’s. The latter was done by considering whether proponents of extrospective approaches and introspective approaches can offer a plausible non-SCS account of how we know we have a given combinatory state. I maintained that a particular extrospective approach to non-CS states—Byrne’s (2018) account—if tenable with respect to non-CS states, has the resources to provide us with a non-SCS way of knowing our combinatory states. It is much more difficult, I argued, to make sense of how particular introspectionist approaches can avoid SCS. I contended, however, that given an epistemology of the attitudes grounded in phenomenology, there are ways in which the introspectionist might be able to avoid having to posit SCS explanations of our self-knowledge of these states as well.

Needless to say, I take this discussion to merely scratch the surface of work to be done on the epistemology of CS’s. The discussion centered on only two broad approaches to self-knowledge where only a few specific approaches within the broader genera were investigated. Accounts that do not fit neatly into either camp were left undiscussed. There is fertile ground, then, for future work on CS’s. Work, I think, that should be done. For if there is one thing I hope has been demonstrated by the above discussion it is that providing an epistemology of combinatory states is more interesting and demanding than one might initially be inclined to suspect. Interestingness and challengingness aside, a more robust explanation of how we know what types of persons we are might very well depend on it.