2.1 … In outline

Sosa (Reference Sosa2007, Reference Sosa2010, Reference Sosa2011) offers an account of the normativity of performances with an aim. According to Sosa, performances with an aim fall under ‘AAA structure’. This means they can be evaluated in terms of whether they attain the aim (i.e. whether they are successful), whether they are adroit (i.e. whether they manifest competence) and whether they are apt (i.e. whether they are successful because competent). For instance, the aim of a shot in target archery is to hit the target. To say that a shot in target archery falls under AAA structure is to say that it can be evaluated in terms of whether it hits the target, whether it is competent and whether it hits the target because it is competent.

Competences, according to Sosa, are dispositions to perform well. That is to say, competences are dispositions that issue performances that, in turn, are likely to be successful.

Dispositions have three essential components: a constitution component (CO), a condition (or shape) component (SH) and a situation component (SI). For instance, a match's disposition of being flammable has as its constitution component certain physical and chemical properties of the powdery head. These properties constitute the seat of the disposition. Its condition component includes, among other things, being dry, and its situation component includes being in oxygen (Sosa Reference Sosa2010: 465).

Dispositions also ‘correspond to’ trigger-manifestation conditionals. In the case of our match's flammability, this conditional is: if it were struck it would likely light. The idea here seems to be that a disposition corresponds to a trigger-manifestation conditional if and only if that conditional must be true of the object satisfying CO and being in SH and SI. To say that the match's flammability corresponds to the conditional ‘if it were struck it would likely light’ is to say that this conditional must be true of any match the powdery head of which has the relevant physical and chemical properties, which is dry, etc., and which is in oxygen, etc. (Sosa Reference Sosa2010: 466).

Since competences are dispositions, they also feature CO, SH and SI components and correspond to trigger-manifestation conditionals of the form: if the disposition were to issue a performance, the performance would likely be successful. In the case of archery competence, the CO are certain properties of the agent, including more basic motor and cognitive competences, the SH include being awake and sober and the SI include there being enough light and normal winds. Furthermore, since the performances the disposition issues here are shots and successes hits, it corresponds to the following trigger-manifestation conditional: if the agent were to take a shot, he would likely hit the target. So, a competent archer is an agent who has certain properties such that when awake, sober, etc. and there is enough light, winds are normal, etc., he would likely hit the target were he to shoot at it (Sosa Reference Sosa2010: 466).

Sosa also distinguishes between three different levels of competence: first, the constitutional competence which consists of the constitutional component of the agent's disposition (CO); second, the inner competence which combines the constitutional component with the shape component (CO and SH); and, third, the complete competence which, besides the constitutional and the shape components, also includes the situational component (CO, SH and SI). On this picture, it is of course the complete competence that corresponds to the trigger-manifestation conditional (Sosa Reference Sosa2010: 465).

According to Sosa, a performance is successful because competent or apt if and only if its success manifests the agent's competence. With the above account of competence in play, we can say a bit more about what that means. Crucially, aptness is unpacked in terms of the notion of a complete competence: ‘A performance is apt’, Sosa tells us, ‘if and only if its success manifests a complete competence’ (Sosa Reference Sosa2010: 470). Aptness thus requires that the relevant disposition's CO, SH and SI are in place. Suppose, for instance, that an agent is a competent archer in virtue of having certain properties such that when awake, sober, etc. and there is enough light, winds are normal, etc., he would likely hit the target were he to shoot at it. In that case, any shot he fires will be apt only if he is in fact awake, sober, etc. and there is in fact enough light, winds are normal, etc. Any shot he fires while drunk, in darkness or in abnormal winds will not qualify as apt.

Crucially, according to Sosa, beliefs are performances with an aim. More specifically, the aim of belief is truth (Sosa Reference Sosa2011: 1). In consequence, beliefs too fall under AAA structure. They can be evaluated in terms of whether they attain the aim of truth, whether they are adroit or competent, and whether they are apt or true because competent. Unsurprisingly, Sosa's suggestion is that knowledge is apt belief.

This account of competences can be applied to the kinds of competences required for knowing (‘cognitive competences’) in the following way: Cognitive competences are dispositions of agents featuring CO, SH and SI. Since the performances these dispositions issue are beliefs and successes are truths, they correspond to the following kind of trigger-manifestation conditional: if the agent were to believe a proposition within a certain range, he would likely believe truly. For instance, an agent has a cognitive competence to spot barns just in case he has certain properties such that when in suitable SH (awake, sober, etc.) and SI (sufficient light, etc.), if he were to believe that a barn is present, his belief would likely be true. Of course, in case of cognitive competence, too, the three levels of competence (constitutional, inner and complete) can be distinguished. Crucially, for cognitive competences, too, aptness requires that the agent's success manifests a complete competence. In consequence, any belief formed in untoward SH or SI is bound to fall short of knowledge, no matter whether it is adroit and/or turns out to be true.

2.2 … And Fake Barn cases

With Sosa's virtue epistemology in play, let's return to Barns. How can Sosa accommodate the intuition of ignorance in this case? Sosa tackles this question by introducing a distinction between animal knowledge and reflective knowledge. While animal knowledge is first-order apt belief, reflective knowledge is meta-apt belief. Meta-apt belief involves the manifestation of a second-order cognitive competence, which is, in essence, a monitoring competence. More specifically, according to Sosa, the relevant second-order competences are ‘default competences’ in the sense that they assume (by default) that the complete first-order competence is present, unless there are indications to the contrary (Sosa Reference Sosa2010: 473). Like their first-order cousins, second-order competences come in three varieties: constitutional, inner and complete. Here, too, the constitutional competence consists in the seat of the competence (CO) only, the inner combines the seat and the inner component (CO and SH) and the complete competence includes, in addition, the situational component (CO, SH and SI). According to Sosa, the SI of second-order cognitive competences include the following condition:

For a belief to be meta-apt it must manifest a complete second-order competence. A belief will thus be meta-apt only if TTS is satisfied. It is easy to see that, in consequence, a belief can be apt without being meta-apt. This will happen when the agent rises to the level of complete first-order competence but is not so situated that there would be tell-tale signs if he didn't. In that case the agent will acquire animal knowledge, without acquiring reflective knowledge.

According to Sosa, this is exactly what happens in Barns: Sosa accepts that Barnes rises to the level of complete competence at the first order and thus acquires animal knowledge. On the second order, however, things look differently: if Barnes's first-order competence were absent here, this would be because Barnes is looking at a fake barn. In that case, there would be nothing about the situation that would alert Barnes to this fact. That means that Barnes does not satisfy TTS. As a result, Barnes does not rise to the level of complete competence at the second order. Barnes's belief does not qualify as reflective knowledge (Sosa Reference Sosa2010: 473–4).

Crucially, according to Sosa, intuitions about knowledge track the presence and absence of reflective knowledge. Since Barnes does not attain reflective knowledge, the intuition of ignorance in Barns is duly explained.

Before moving on, I would like to discuss one immediate problem that arises for Sosa's account at this stage. To bring it out, consider first the following case:

Consider Barnes's belief about the presence of a barn at the time at which he first forms it (t ₀). Intuitively, at t ₀, Barnes's belief falls short of knowledge. Sosa wants to explain this intuition in terms of the absence of reflective knowledge and that, in turn, in terms of TTS's not being satisfied. Recall that, according to TTS, the SI of a second-order competence will be satisfied only if the agent is so situated that were the first-order competence to be absent, sooner or later there would be tell-tale signs alerting him to this fact. Crucially, there is reason to believe that TTS is satisfied in Barns* at t ₀. After all, if, at t ₀, the first-order competence were absent, this would be because the structure Barnes is looking at is a blue fake barn – or so we may assume. In that case, however, the relevant tell-tale signs do present themselves sooner or later, viz. upon exiting Fake Barn County when Barnes passes by a sign telling him that the blue barn-shaped objects are mere façades. In Barns*, then, Barnes is so situated that if the first-order competence were absent, sooner or later there would be tell-tale signs. TTS turns out to be satisfied. As a result, Sosa's envisaged explanation of the absence of reflective knowledge at t ₀ in terms of TTS fails.

Fortunately, this problem can be solved. Sosa can simply strengthen TTS in the following way. Rather than allowing the tell-tale signs to present themselves sooner or later, Sosa can require that the relevant tell-tale signs present themselves soon enough. If, in addition, it can be argued that in Barns*, the tell-tale signs would not present themselves soon enough, Sosa's explanation of Barnes's lack of knowledge will be reinstated.

Of course, the question that immediately arises at this stage is what counts as soon enough. Curiously, there is reason to believe that Sosa cannot allow the extension of ‘soon enough’ to extend to any times beyond the time of belief-formation. To see this notice that we can easily set up Barns* in such a way that Barnes notices the sign at t ₁, i.e. immediately after forming his belief about the presence of the barn at t ₀. In that case, still Barnes's belief intuitively falls short of knowledge at t ₀. However, if t ₁ falls within the extension of ‘soon enough’, then, at t ₀, Barnes will satisfy even the strengthened version of TTS, for reasons already discussed. Once again, the envisaged explanation of the absence of reflective knowledge at t ₀ in terms of TTS fails.

I take these considerations to indicate that, in order to successfully explain the absence of reflective knowledge in cases like Barns, TTS will have to be restricted to the time of belief-formation. For that reason, in the remainder of the paper I will assume that, properly understood, Sosa's crucial requirement is the following:

3.1 Frankfurt cases

The first problem for Sosa I'd like to discuss concerns the second-order competences required by reflective knowledge. Sosa's account of these competences as competences that respond to the relevant tell-tale signs seems plausible enough. What is less clear to me, however, is that such competences must be manifested in order to acquire knowledge, at least if the SI include TTS*. To see this consider first the following case:

When Detective has completed his case against Killer, intuitively, he knows that Killer did the deed. At the same time, Detective's belief does not qualify as reflective knowledge. If Detective's complete first-order competence had been absent, this would have been because Killer decided not to carry out his plan or didn't leave enough clues. In that case, however, Controller would have stepped in and murdered Victim and/or planted the evidence incriminating Killer. In that case there would not have been tell-tale signs that Detective's first-order competence is absent. Given that absence of first-order competence would not have been accompanied by tell-tale signs, Detective's second-order competence does not satisfy TTS*. By Sosa's lights, his belief does not qualify as reflective knowledge.Footnote ⁵ This is, of course, bad news for Sosa. After all, Detective's belief intuitively qualifies as knowledge. If intuitions about knowledge track the presence and absence of reflective knowledge, it looks as though we have a counterexample to Sosa's account.Footnote ⁶

As I see it the problem here is that the second-order competences are construed in too strong a manner. More specifically, the difficulty arises from TTS*, the requirement that one be so situated that there would be tell-tale signs if one's first-order competence were absent. What Frame highlights is that one can have knowledge even when absence of competence would not have been accompanied by tell-tale signs.

One promising way of avoiding this problem is by weakening TTS*. Instead of requiring that tell-tale signs would be present if the first-order competence were absent, we might think of the second-order competences as competences that are such that if they were to detect tell-tale signs as such, this would duly be taken into account (henceforth ‘TTS**’). Notice that, on TTS**, Detective has the complete second-order competence. After all, if Detective were to detect anything fishy, he would suspend judgment. In that case, Detective does acquire reflective knowledge after all. The counterexample can be avoided.

Unfortunately, this alternative comes at a cost for Sosa. There is now little hope for his intended explanation of the intuition of ignorance in Barns. After all, here too, TTS** is satisfied: if Barnes were to detect tell-tale signs as such, they would be duly taken into account. For instance, were Barnes to look at a broken façade that exposes the structural elements behind it, he would believe that he is not looking at a real barn. In consequence, if TTS* is replaced by the weaker TTS**, Sosa's explanation of the intuition of ignorance in Barns will no longer work.

3.2 Lottery cases

Let's look at yet another case:

Although the probabilistic evidence supporting Raffles's belief is excellent and the belief is true, Raffles does not know that his friend's ticket won't win L.

To the best of my knowledge, Sosa does not discuss cases like Lottery in his published work.Footnote ⁷ That is not to say, however, that they don't pose a problem for him. On the contrary, there is excellent reason to think that Sosa will struggle with these cases. To see this, let's first ask whether Raffles's cognitive success manifests a complete first-order competence here. Before answering this question, let's ask whether Raffles rises to the level of complete competence in this case. Very plausibly, the answer to this question is ‘yes’. After all, Raffles's belief is produced by disposition that in his SH (Raffles is awake, alert, etc.) and SI (the lottery is fair, etc.) corresponds to the trigger-manifestation conditional: if Raffles were to believe that Gambles's ticket will lose, his belief would very likely be true. So the disposition that produces Raffles's belief qualifies as a complete cognitive competence. If so, his cognitive performance (the belief issued by this disposition) manifests a complete cognitive competence in this case. Notice, furthermore, that his cognitive performance is also successful. After all, his belief that Gambles's ticket won't win the lottery is true.

With these points in play, let's return to the question whether Raffles's cognitive success manifests this complete competence in this case. Answering it is actually a bit tricky. The reason for this is that it does not follow from Sosa's account of aptness that cases in which performances are both successful and manifest a complete competence are cases of apt performances, i.e. cases of performances in which the success manifests a complete competence. That said, it is quite hard to see how cases of successful performance that manifest a complete competence could still fail to be cases in which the performance is apt. Think, for instance, of Archer once more. Consider a case in which he is taking a shot in perfect conditions (peak concentration, no winds, etc.) that hits the target. What reason could there possibly be to deny that his shot is apt? I, for one, can't think of any. If so, it is fair to assume that cases of performances that are successful and manifest complete competences are also cases of apt performances. Given Raffles's belief that Gambles's ticket won't win the lottery is both successful (true) and manifests a complete cognitive competence, there is thus reason to believe that it is also apt. In other words, there is reason to believe that it qualifies as animal knowledge.

Does Raffles's true belief also qualify as reflective knowledge? Does it manifest a complete second-order competence? First, note that we may assume that Raffles's belief does manifest the second-order monitoring competence of the sort required for reflective knowledge. By way of evidence, notice that we may assume that Raffles is such that had tell-tale signs been detected by the competence, they would have duly been taken into account. The crucial question, then, is whether Raffles's rises to the level of complete competence at the second-order as well. This question is also tricky to answer. One reason for this is that we need to know whether TTS* is satisfied. And, in order to ascertain this, we need to know what would have been the case if Raffles's complete first-order competence had been absent.

One might think that Raffles's complete first-order competence would have been absent in a case in which Gambles's ticket wins L. Since, in that case, there would not have been the required tell-tale signs, TTS* is not satisfied. Raffles's belief does not manifest a complete second-order competence and thus falls short of reflective knowledge. Unfortunately, there is reason to think that this doesn't work: Raffles's complete first-order competence would not have been absent in a case in which Gambles's ticket wins L. After all, even in that case, Raffles still satisfies the trigger-manifestation conditional: if he were to believe that Gambles's ticket loses, his belief is very likely true. For that reason, the SI of Raffles's first-order competence should include the case in which Gambles's ticket wins. It's just that, in this case, the unlikely possibility of failure materialises.

If the case in which Raffles's first-order competence would have been absent is not one in which Gambles's ticket wins L, presumably it is either a case (i) in which Raffles has no idea that Gambles owns a ticket to begin with or else a case (ii) in which Raffles knows that Gambles does not own a ticket. In both cases, there would be tell-tale signs that he does not have the complete first-order competence (specifically: that he is in SI such that the relevant trigger-manifestation conditional is not true of him). In case (i) the tell-tale sign is the fact that Raffles has no idea that Gambles owns a ticket, in case (ii) it's the fact that Raffles knows that Gambles doesn't own a ticket. There is thus reason to believe that TTS* is satisfied.

This leads us immediately to another difficulty in ascertaining whether Raffles's second-order competence is complete here. Completeness requires that all SH and SI of the second-order competence are satisfied. So far, we have only shown that one such SI, TTS*, is. What remains to be shown is that the other SH and SI are satisfied as well. This task is complicated by the fact that Sosa typically leaves the list of SH and SI open-ended. So how can we possibly hope to answer this question?

The answer is: ‘By investigating trigger-manifestation conditionals.’ If relative to his SH and SI, Raffles satisfies the trigger-manifestation conditional that corresponds to the second-order competence, Raffles rises to the level of complete competence. What then is the trigger-manifestation conditional for Raffles's second-order competences? To answer this question, recall that Sosa construes the second-order competences as default competences that assume (by default) that the complete first-order competence is present. Given that this is so, the trigger-manifestation conditional corresponding to such second-order competences will be conditionals of the form: if – in suitable SH and SI – the second-order competence were to assume that the complete first-order competence is present, then it would likely be present. Crucially, there is excellent reason to believe that, in Lottery, the relevant trigger-manifestation conditional is true of Raffles. After all, we may assume that Raffles is in maximally favourable SH: he is awake, focused, at peak concentration, etc. Moreover, as we have already seen, TTS* is satisfied: Raffles is so situated that if the complete first-order competence were absent, then there would be tell-tale signs. Now consider: if – whilst Raffles is awake, focused, at peak concentration, etc. and so situated that if the complete first-order competence were absent, there would be tell-tale signs – Raffles's second-order competence were to assume that the complete first-order competence is present, it would likely be present. This trigger-manifestation conditional is very plausibly true. Since relative to Raffles's SH and SI, the trigger-manifestation conditional holds, Raffles's second-order competence is complete. In consequence, Raffles attains not only animal knowledge, but also reflective knowledge. Sosa's account makes the wrong predictions in cases like Lottery, too.

3.3 Dreaming scepticism

My last worry with Sosa's account arises from another case which Sosa himself discusses. It takes the following shape:

Sosa grants that, intuitively, Wakes knows that he is awake. Given that Sosa is right in that intuitions about knowledge track the presence and absence of reflective knowledge, Sosa had better establish that Wakes's belief qualifies as reflective knowledge. Now, the question is whether Sosa can secure this result. This task is complicated by the fact that Wakes is in a similar situation to Barnes in that both of them might very easily have acquired false beliefs instead without being in a position to know this. The challenge Sosa thus faces is to show that Wakes's belief qualifies as reflective knowledge without compromising his earlier account of Barnes's ignorance in terms of a lack of reflective knowledge.

Sosa takes on the challenge in the following passage:

The question remains how good Sosa's story is. In order to answer this question, consider the following case:

Let's ask whether Sweets knows that he is about to eat a chocolate. Very plausibly the answer to this question is ‘no’. The case is structurally analogous to Barns. Notice also that, as a result, Sosa had better analyse the case as one in which Sweets has animal knowledge but lacks reflective knowledge.

Of course, so far there is no special problem for Sosa. But now suppose we add:

To begin with, notice that this addition does not alter the intuition that Sweets does not know that he is about to eat a chocolate. With this point in play, here, then, is the crucial question: Does Sweets have his complete second-order competence? Given that dreams provide the required tell-tale signs, it would seem that Sweets is fine on this front. Moreover, given that inability to recognise dreams as such due to diminished competence leaves the second-order competences intact, there are no problems here either. In this respect, Chocolates is just like Dream. If so, and if, in Dream, Wakes retains his second-order competence and acquires reflective knowledge through it, then the same holds for Chocolates.