1. AN ACCOUNT OF TRUST

It seems intuitively right to say that each one of us has a natural and unavoidable trust in one's own cognitive faculties, call this one's epistemic self-trust.Footnote ³ But what precisely does this consist of? This question falls apart into two sub-questions: what exactly is trust, and what exactly is it about one's everyday life that manifests epistemic self-trust?

Trust is an important concept and there is a large literature discussing its nature and why it matters.Footnote ⁴ One aim in promulgating a definition of trust would be pursuit and capture of that elusive phantom, a determinate ordinary language concept. A less treacherous, more manageably constrained aim is to propose a definition aimed at capturing a theoretically important concept for our concerns – which will include both epistemology and normative interpersonal relations. The remarks now made are in pursuit of this second aim.

We can agree with Zagzebski (EA: 36) that one core notion of trust is of a 3-place relation: A trusts X to Phi. X here may or may not be another person, an agent, and phi-ing is something that X is able to ‘do’ in a very broad sense. This doing may or may not be an intentional action; it may just be a matter of an item exemplifying a certain property – for instance, a wooden beam not breaking when a heavy load is placed on it. Zagzebski characterises trust as involving 3 components:

In the literature on trust it is debated whether or not trust entails belief, as LZ (i) posits (see Jones Reference Jones1996). Whether or not this is so, I think it is perspicuous in describing the phenomena to identify as the core phenomenon that of:

Reliance is part of what is involved in LZ (iii), once one adds that whether X phi-s is of concern to A, that is impacts on outcomes of concern to A. Thus, for instance, I rely on the fact that my car has enough fuel to get me to my destination when I do not stop to refuel at the last petrol station. A's reliance on X to phi is consistent with A's believing that X will phi, but does not entail it.Footnote ⁶ In the present example, I may have considered whether I have enough fuel and decided affirmatively; or I may simply not have considered the matter. I would resist the proposal that the mere fact of someone's reliance on the fact that X will phi entails that she believes, or has any personal-level mental attitude to, this proposition. (Someone who has, due to haste and preoccupation with other matters, forgotten to check whether there is enough fuel may have no doxastic attitude on the matter; it may be that once considered, she would realize it entirely possible there is not.)

One could define ‘trust’ so that all cases of reliance by definition count as trust. But there are other important pulls on the notion of trust. For instance, as Baier observes (Reference Baier1995a), one can be disappointed and indeed dismayed when an item one relies on fails to perform; but only when trust is involved can one coherently feel betrayed. I suggest that we should distinguish between simple reliance and trust-based reliance. The latter is the sub-kind of reliance that is necessarily on another agent, a person, and occurs when one relies on another person to do something due to her epistemic and/or character virtues and her competences – for instance, when one trusts another person to pay back by the agreed time money one lends to her, and so does not require her first to sign a legally binding contract to do so.Footnote ⁷ Such trust-based reliance on another person is often referred to in everyday talk as ‘trusting’ her (‘Did you make her sign a contract?’ – ‘No, I trust her to pay the money back.’).

In general it is not a good idea to rely on a fact unless one knows or at least has good evidence that it obtains. But sometimes one may be forced into relying on some fact in the absence of this. When one does so, one makes a leap of epistemic faith in one's reliance. One strand in ordinary usage of ‘trust’ has it that the term is aptly used regarding what we have called trust-based reliance – reliance by one person on another to perform out of good character – only when the truster's expectation of performance is not grounded in evidence possessed by her of the trustee's relevant character virtues, and instead her reliance rests solely on epistemic faith. But I agree with Zagzebski, regarding trust-based reliance, that one trusts a person no less, when one has strong empirical grounds to believe in her trustworthiness (EA: 42). As with our money-lending case above, the hallmark of trust-based reliance is not that it is empirically ungrounded – it would be foolish indeed to lend money to someone expecting its prompt return without evidence of her competence in managing her financial affairs and of her good will – but the fact that one relies on the trustee's good character as the mechanism ensuring she effects the performance on which one relies, and does not seek to put in place some other constraint. The term ‘trust’ is apt even for such empirically grounded trust-based reliance, since good character and relevant competences can never guarantee as a matter of strict law that the person will perform. Even the most virtuous are subject to temptation, and the most competent may slip up sometimes or succumb to a bad-luck environment.

Simple reliance on some non-agent to perform in some respect, when it has no evidential basis and rests on epistemic faith, merits on this different basis the title of trust, and our everyday term ‘trust’ is often used also in this kind of circumstance.Footnote ⁸ Empirically ungrounded simple reliance, then, is one species of trust.Footnote ⁹ It does not entail belief, or any personal-level attitude, to the proposition that the item relied on to perform will do so. In this species of trust we have a core component of LZ (iii), but not LZ (i).

2. AN ACCOUNT OF EPISTEMIC SELF-TRUST

We are now in a position to characterize the phenomenon of epistemic self-trust. I think the core phenomenon is this: Each one of us in one's everyday life relies on one's core package of cognitive faculties – perception, proprioception, memory, intellectual intuition and introspection – reliably to deliver one true beliefs. Once one appreciates that this reliance is, perforce, epistemically ungrounded (see section 3 below), it is apt to call this ungrounded reliance epistemic self-trust – instancing our second species of trust just described at the end of the previous section. Thus we can formulate:

This core phenomenon of epistemic self-trust does not entail that the epistemically self-trusting subject has any doxastic attitude to a general proposition to the effect ‘my faculties reliably furnish me with beliefs that are true’ – so LZ(i) is not part of the account of epistemic self-trust here proposed. A subject who lacks the conceptual resources to frame such an attitude can exhibit epistemic self-trust. Even if she has these resources, the fact of her epistemic self-trust does not entail that she has reflected to form such a general reflective attitude to her own faculties; and the fact of her ongoing ungrounded reliance on her faculties does not in itself constitute her having such a general belief or other attitude. (Of course, once she raises the question: ‘Are my faculties reliable?’ she will perforce form some doxastic attitude and, if it is not belief that they are so, she will be in a situation of cognitive dissonance – unavoidably relying on her faculties while having an attitude to the proposition that they are reliable other than confident belief in this.)

Some further explanatory comments on EST are in order. First, EST presupposes that it is of concern to one to acquire beliefs by methods that reliably yield truth. This is entailed, once it is conceded one has an interest in acquiring true beliefs. But there are tricky issues here. First, there is the question of why one has an interest in acquiring true beliefs. On one view, this is only because they are instrumentally useful in helping one to survive and flourish. This view faces problems explaining why it is exactly truth that is needed here, rather than a slightly different target: some false beliefs might help us flourish better, so long as belief-forming methods that yield only the instrumentally useful falsehoods, not damaging ones, can be honed. In fact I think one can make an argument that we have an interest in our faculties being truth-yielding as an ultimate, not merely an instrumental goal. I say why below. But even if truth, not utility for survival, is established to be the goal, there are issues about the apt trade-off between the dual ills of missing out on true beliefs through excessive caution, and risking false beliefs due to excessive boldness. Depending on precisely what property it is in one's best interest that one's cognitive faculties possess in this respect, there will be variant characterizations of precisely what fact one is relying on. I shall not attempt to fine-tune this matter now, leaving it open what precise property in the ball-park of truth-conduciveness we rely on our faculties possessing. My discussion concerns this property of one's cognitive faculties – delivering true beliefs with the apt trade-off between risk of falsehood versus risk of missing out on available true belief – whatever exactly that proves to be.

Why is it that each one of us has an interest in acquiring true beliefs, hence is correctly described as relying on one's faculties reliably to deliver these? Zagzebski ascribes this interest to the fact that, as she claims, each one of us possesses a humanly ‘natural desire for truth’ (EA: Ch. 2.2). If indeed each one of us did have a personal-level propositional attitude with content: I want that I amass true beliefs, this would ensure this interest. But I find no reason to suppose all adult humans have a self-regarding desire with this general content. As noted just above, the core phenomenon of epistemic self-trust as ungrounded reliance specified in EST does not entail that the relying subject has the conceptual resources to frame for herself a propositional attitude to her own belief-providing cognitive resources; and even if she is conceptually equipped to do so, she may have been too busy engaging with the world to bother with such general epistemic self-reflection and setting of goals. Ground-level reliance, the core of epistemic self-trust – forming beliefs about one's environment through perception, retaining and accessing them in memory, making simple inferences – comes before, and independently of, epistemic reflective self-awareness. So the ubiquity of epistemic self-trust does not show that such a desire with general self-regarding content is universal, and this seems unlikely.

But a personal-level propositional attitude with content: ‘I want that I amass true beliefs’ does not seem to be what Zagzebski has in mind when she writes of the ‘natural desire for truth’. What then does she mean? Explaining what she means by saying we all have a natural desire for truth Zagzebski glosses her claim thus: ‘The natural desire for truth is the natural desire to have our questions answered. We want to know who or what something is, or whether something is the case, or how something works … it is natural to ask questions and to want answers to them … we want the right answers … [this is] another way of saying that we want the truth’ (EA: 34). This quote indicates that Zagzebski's posit is not of a self-regarding propositional attitude; the sense in which, she maintains, we naturally desire truth is that in many individual cases we ask a question, and want a correct answer to it.

I propose that a good way to spell out Zagzebski's thesis of the ‘natural desire for truth’ is: we humans are naturally inquisitive, we have intellectual curiosity. That is to say, we are naturally disposed to engage in a certain activity, enquiry – seeking to find things out about the world. This means, as Zagzebski says, that we formulate questions and want answers, and want correct – that is, true – answers. The sense in which it is a constitutive feature of normal humans that they ‘desire truth’ is not that one has a propositional attitude to a self-regarding generalization: ‘I want that I have lots of true beliefs’ – that is sophisticated reflection that surely is not universal. The sense in which it is endemic to all humans that one desires truth is one that does not require any self-reflective powers. Enquiry includes, at one extreme, specialized scientific and scholarly research, and at the other simply looking around and attending to one's surroundings. Humans, we can agree with Zagzebski, are constitutively disposed towards it.

Moreover, it is important to note, wanting true answers is not an additional contingent feature of our enquiries. It is intrinsic to what enquiry is, that it aims at truth. Finding out what is so regarding some matter just is finding out what is true regarding the matter. The fact that ‘true’ conforms to the schema: P if and only if it is true that P ensures this identity, and hence the fact that truth is the internal aim of enquiry. I think Zagzebski's thesis of the humanly universal natural desire for truth, when so explicated, is correct: humans are constitutively disposed to engage from time to time in enquiry.

Since enquiry has truth as its internal aim, this suffices to provide an answer to our question: why does each one of us have an interest in acquiring true beliefs? Each one of us does so, because she is from time to time an enquirer – each one of us is constitutively disposed to engage in an activity that has truth as its internal aim. This is why each human has an interest in acquiring true beliefs as an intrinsic goal, not merely an instrumental one. This fact about each one of us vindicates the presupposition of EST that in employing our faculties to acquire beliefs we rely on them reliably to yield us beliefs that are true.

Moreover, the fact that each one of us is from time to time an enquirer also ensures that we each do and must perforce employ our proper and only belief-yielding resources, our various cognitive faculties, to furnish us with beliefs;Footnote ¹⁰ relying on them as we do so reliably to furnish ones that are true. So: one's ‘natural desire for truth’, explicated as one's disposition from time to time to enquire, entails first, that one from time to time employs one's cognitive faculties (perception, memory etc.) to acquire beliefs; and second, that in doing so one relies on them to furnish one with true beliefs. This vindicates Zagzebski's thesis that the humanly natural desire for truth, so explicated, ipso facto lands each one of us in a situation of epistemic self-trust, when this is explicated as EST: ‘I relied on my faculties to answer my questions … trusting my faculties was already part of the prereflective self’ (EA: 36).

Enquiry is an intentional activity. Even if one is too lazy to do much by way of active finding out about one's environment, one still cannot help but acquire beliefs via perception and retain and access them in memory. We can see this relatively involuntary and passive acquisition of belief as also intrinsically aimed at truth, since it takes place alongside voluntary enquiry, and by means of the very same cognitive faculties employed in enquiry.

Zagzebski (EA: Ch. 2.2) equates the phenomenon of epistemic self-trust with the claimed fact of our natural desire for truth, together with our ‘natural belief that this desire is satisfiable’ – which is equivalent to belief in ‘the suitability of my faculties for getting truth’ (EA: 36). She writes: ‘I trust my car to get me to campus. …When the trust in question is epistemic trust in myself, I trust my epistemic faculties to get me to the truth. Trusting myself in this sense is like trusting my car …’ (EA: 36). And she takes trust to involve all of LZ (i) – LZ (iii). But the core phenomenon of epistemic self-trust captured in EST does not include belief in the suitability of one's faculties for getting truth, i.e. their reliability; and I have argued that the fact that one relies on one's faculties to yield truth does not ipso facto entail that one is correctly ascribed belief or any other attitude to their doing so. Contrary to Zagzebski's thesis, the core phenomenon of epistemic self-trust does not involve or imply any belief in the reliability of one's faculties. Having general meta-beliefs of any kind about one's own cognitive powers is a further self-reflective layer, which is not ensured merely by the fact of first-level reliance – forming beliefs from perception, laying down and later retrieving memories, making simple inferences, and so forth. My disagreement with Zagzebski over whether epistemic self-trust of itself involves or entails a self-regarding belief in the reliability of one's faculties does not affect the cogency of the various arguments for its legitimacy she makes, which are considered in my remaining sections.Footnote ¹¹ What matters for this in the characterization of epistemic self-trust is her thesis, with which I agree, that one's first-level reliance on one's faculties reliably to yield one true beliefs is both natural and practically unavoidable, as well as being epistemically ungrounded, resting only on epistemic faith.

3. THE EPISTEMIC UNGROUNDEDNESS OF OUR EPISTEMIC SELF-TRUST

In section 1 I gave an account of two different types of circumstance in which the notion of trust is aptly applied, and in section 2 I offered an account of epistemic self-trust. I agreed with Zagzebski that we inevitably employ our own cognitive faculties to gain beliefs in the activity of enquiry, to which we are constitutively disposed, as well as forming beliefs more passively from perception and memory, and that in doing these things we rely on our faculties to yield us true beliefs. This reliance is – as will be emphasized shortly – epistemically ungrounded, resting only on epistemic faith, and thus, I suggested, deserves the title epistemic self-trust. I identified this ground-level reliance on one's faculties to yield one true beliefs as the core of one's epistemic self-trust. I disagreed with Zagzebski's account of trust, and specifically of epistemic self-trust, in that I did not include a reflective belief in the reliability of one's faculties as a constitutive part of epistemic self-trust; and I suggested it is not empirically plausible that each one of us has formed such a reflective belief about one's own faculties. I rejected the idea that the mere fact of one's ground-level reliance constitutes possession of such a belief. In the remainder of this discussion I now turn to explicate and evaluate Zagzebski's several arguments for the rationality and epistemic legitimacy of our epistemic self-trust.

Let us quickly agree with Zagzebski that while it is perhaps not absolutely and utterly impossible to prescind from one's epistemic self-trust at least to some extent, this is not a live practical possibility. We saw above that engaging in enquiry necessitates epistemic self-trust, and agreed that it is part of human nature to enquire. Moreover we form beliefs continually from perception and memory in a manner that is not under conscious control, even if we were to refrain from engaging in optional information-gathering activity involving search behaviour. So we can accept and proceed on the premiss that one's epistemic self-trust is unavoidable.

Let us also quickly agree with Zagzebski and others that, when each human comes to reflect on her epistemic situation, and the fact of her ground-level reliance on her cognitive faculties, she finds herself in the predicament made notorious as the ‘Cartesian Circle’ – there is no possibility of achieving an epistemically non-circular certification of her core cognitive faculties as reliable, since she will necessarily be employing them in making the argument for their reliability.Footnote ¹² Once an epistemic subject has realized this situation through reflection, she sees that her reliance on her faculties is based on a leap of epistemic faith. It involves ‘trust’ in one of the two senses identified in section 1, namely empirically ungrounded reliance resting only on epistemic faith.Footnote ¹³ Once it is realized that one's inevitable continual ground-floor reliance on one's cognitive faculties is epistemically ungrounded, the question of how it is nonetheless rational and epistemically legitimateFootnote ¹⁴ presents itself as urgent. (One may have no doubt that one's ungrounded reliance on one's faculties is epistemically rational, and still want an explanation of why it is so. Justification that is explanatory, though not intended to be suasive – not intended to persuade someone not already convinced of the legitimacy of what the explanation exhibits as justified – is central in epistemology.) In the remainder of this discussion I offer an explicatory exegesis and critique of Zagzebski's several arguments for the legitimacy of epistemic self-trust.

4. ZAGZEBSKI'S CONCLUDING CASE FOR THE LEGITIMACY OF EPISTEMIC SELF-TRUST

Zagzebski argues that epistemic self-trust is ‘rational on reflection’ – which in her internalist framework amounts to its being epistemically legitimate. In fact there is more than one argument to be found in EA Ch. 2 for this conclusion. Zagzebski's final conclusion of the chapter goes thus:

I find this statement rhetorically persuasive despite being unsure exactly what it means. There are at least three different arguments suggested by it – perhaps the hint of each one of them endows the statement with its persuasive ring. One argument LZCONCL suggests is a Transcendental Argument for Self-Trust:

This argument is only intimated in LZ CONCL, since the second premiss – there are reasons for belief – is not stated; it is perhaps conversationally implied by the whole, or implicit it in the final sentence. I find TAST plausible. But one in genuine doubt about the epistemic legitimacy of self-trust would not be willing to accept the premiss that there are reasons (i.e. legitimate reasons) for belief as a given. This does not prevent TAST serving as an explanatory justification for epistemic self-trust. Alternatively one can detect in LZCONCL what we may call an Analytic Argument for Self-Trust:

The final sentence of LZCONCL hints at this argument reminiscent of Strawson's ‘analytic’ argument for the epistemic rationality of induction (Strawson Reference Strawson1952). Somewhat differently again one might discern in LZCONCL a Kantian-style argument for the epistemic legitimacy of self-trust as a necessary condition for the possibility of rational thought:

I think an elaboration of KAST could make a convincing case for the legitimacy of self-trust, more so than the rather thin unconvincing AAST.

LZCONCL and Zagzebski's discussion generally have to be understood in light of her idiosyncratic conception of what ‘reasons for belief’ are (EA: Ch. 2.5). In theorizing epistemic justifiedness of belief, I think clarity requires that one keep firmly in mind the distinction between grounds for belief, versus conditions for justifiedness of belief. Grounds for beliefs are other contentful mental states of the subject reflectively accessible to her – other beliefs, or perhaps conscious experiences; they might also be called evidence for her belief possessed by the subject. Conditions for justifiedness of belief comprise any conditions an epistemologist may cite in giving an account of what renders a belief justified. Conditions of justifiedness are never themselves grounds or evidence; amongst them will be on occasion the fact that the subject possesses appropriate grounds or evidence for her belief. I shall not here try to give an accurate account of exactly how Zagzebski understands ‘reasons for belief’. But what she says about self-trust does make it seem that she includes as such ‘reasons’ both grounds, and other conditions of justifiedness. When she says that ‘self-trust is a reason for belief’ (EA: 50) I think the only way to make sense of what she has in mind is to take her to mean by this that self-trust is a condition of justifiedness for belief: it is a condition that can contribute to an account (given by the epistemologist) of what renders beliefs warranted, epistemically legitimate. This does not mean that self-trust is part of a subject's grounds or evidence for her beliefs – a mental item reflectively accessible to the subject, that she is able to cite in evidence for her belief if challenged.Footnote ¹⁶ Another important observation about Zagzebski's case for the epistemic legitimacy of self-trust is its resolute pure internalism. Self-trust is legitimate if and only if it survives the reflective scrutiny of the subject; if it is ‘rational on reflection’ (EA: Ch. 2.4). One might instead think that its legitimacy can only be adequately explained, given the inevitable circularity of its reflective endorsement, with a dose of externalism – as is argued in Van Cleve (Reference Van Cleve1979): its legitimacy, if it is such, turns at least in part on the fact that our cognitive faculties reliably yield truth in our actual or normal environment, not on what beliefs about this matter reflection may furnish us with.

LZCONCL is Zagzebski's last word in EA Ch. 2 on why epistemic self-trust is legitimate, but it is not her only word. I'll finish by looking closely at another argument to vindicate self-trust found earlier in the chapter. I call it the Meaty Argument, since it rests on substantial theses about the nature of rationality whose truth is not evident.

5. ZAGZEBKSI'S MEATY ARGUMENT FOR EPISTEMIC SELF-TRUST

Zagzebki's main argument for the epistemic legitimacy of epistemic self-trust is offered in EA Ch. 2.4, but this builds on materials from EA Ch. 2.1. In EA Ch. 2.4, after noting the fact of epistemic circularity in any vindication of one's faculties as reliable, Zagzebski comments: ‘Of course, if someone thinks it is not rational to have a belief without non-circular justification, then self-trust is not rational; but then none of our beliefs is rational. However, I see no reason to think that that is what rationality is’ (EA: 43). The conception of human rationality she instead develops in EA Ch. 2.1 and then applies in the rest of the chapter can be summarized thus:

The Meaty Argument for the epistemic rationality of epistemic self-trust is given in EA Ch. 2.4. It draws on RDE. Zagzebski writes:

She then argues, invoking RDE, that the most epistemically rational response by the reflecting subject to the dissonance she feels due to her awareness of conflict between her inevitable epistemic self-trust, and its ungroundedness, is to resolve in favour of trust: to somehow form beliefFootnote ¹⁷ in the reliability of her faculties, thereby eliminating the conflict and so the dissonance. (Zagzbebski includes also the third component of trust on her own account of it, LZ (ii) – a feeling of confidence in one's faculties, or at least absence of feelings of doubt about them.) RDE underwrites this argument, once it is conceded that the option of ending dissonance by ceasing to rely on one's faculties is not feasible. (This is not exactly how Zagzebski spins the argument however – see below.) RDE does so, that is, so long as it is understood in a strong interpretation: that it is always rationally permissible to respond to dissonance by eliminating it. It seems Zagzebski holds this strong version of RDE. She writes: ‘It is rational to resolve dissonance between the components of the self …’ (EA: 44). Her case for the epistemic legitimacy of self-trust is summarized by her thus:

She adds that resolving the dissonance in favour of trust is more rational that doing so in favour of Pyrhonnian scepticism, since it ‘involves less of a departure from what is natural’ (EA: 45). As a further rider to the argument she suggests resolving for self-trust is rational because it puts an end to reflection, and excessive reflection is not rational (EA: 45).

We can formulate this Meaty Argument thus:

This Meaty Argument relies entirely on RDE. Should we agree with RDE? How much exactly of the basis Zagzebksi offers for RDE do we need to accept, to accept the Meaty Argument?

I think the claim that resolving for fully-believing self-trust – belief, affect and treatment – is rational because it puts an end both to excessive reflection, and to the sense of unease one is prone to in light of one's awareness of the ungroundedness of one's inevitable ground-floor epistemic self-reliance, is plausible. One might call this thinned-down variant of the Meaty Argument the Pragmatic Argument for the legitimacy of epistemic self-trust. The Pragmatic Argument, plus TAST and KAST, together add up to a strong cumulative overall case for the rationality-on-reflection and corresponding epistemic legitimacy of self-trust.Footnote ¹⁸ But I have significant worries about RDE. In my final section I scrutinize it more closely. I argue that as a general conception of human rationality it has serious flaws: it both posits too much, and achieves too little.

6. SCRUTINIZING RDE

RDE is not explicitly formulated as a principle by Zagzebski, but its components are laid out sequentially in EA Ch. 2.1. Zagzebski first observes that ‘Beliefs, emotions, desires and decisions can conflict with one another’ (EA: 29). She goes on to make three foundational claims, premisses that together support RDE as conclusion. First, she posits that

In clarification of Diss One Zagzebski tells us that ‘It seems to me that the experience of dissonance is basic. It cannot be explained or analysed in terms of some other experience.’ She does however admit that ‘I do not mean that conflict is defined by the experience of dissonance … but we detect conflict through the experience of dissonance’ (EA: 30). Dissonance occurs as a felt psychological phenomenon, Zagzebski tells us, as a response to one's having two mental states that conflict. It is resolved, that is it disappears, when one of the conflicting states is eliminated – one gets rid of one of the offending states from one's psychic economy, and thereby ‘psychic harmony is restored’ (EA: 30). The second key premiss is that

As Zagzebski puts it: ‘Many times when there is dissonance, the self automatically adjusts by giving up one of the states that conflict … we typically feel a need to resolve it [dissonance] by giving up one of the components of the self that creates the dissonance …’ (EA: 30).

Diss One and Diss Two are empirical claims about how we are psychologically. In themselves they have no normative upshot. To get RDE Zagzebski adds a third crucial premiss with normative content and so power:

Elaborating on NN Zagzebski tells us: ‘there is a close connection between the way the self naturally operates and what the self ought to do. This means there is a connection between the natural and the normative, in particular, between the self as it naturally operates and the way it should operate’ (EA: 32–3).

A consideration she offers in support of NN is that ‘what we do automatically gives us our initial standard of rationality, a standard for what it is to make the adjustment in the self correctly' (EA: 32); ‘the awareness of dissonance resolved without effort gives us our initial model of what rationality is’ (EA: 30).

Diss One, Diss Two and NN together take one a fair way down the road to establishing RDE: if it is natural to end dissonance by eliminating one of two conflicting states causing it, and to be rational is to do what is natural, then surely it is rational to end dissonance by eliminating one of two conflicting states causing it? I will argue shortly that they cannot take one all the way to establishing a strong version of RDE. But first we ask: Are Zagzebski's three premisses in her case for RDE ones we should accept?

We must first note that Diss One is not presented as offering a psychologistic reduction of what it is for two mental states to conflict. Dissonance, Zagzebski tells us, is sensed as a feeling response to conflict perceived between two mental states, it does not constitute it. So one is free to supplement Diss One with one's preferred non-psychologistic account of the logical and evidential relations that obtain between the contents of contentful mental states, together with one's preferred account of how humans are able to latch onto these – perhaps through a faculty of intellectual intuition. Ending of dissonance by eliminating one of a pair of conflicting mental states is offered by Zagzebski not as an account of what logical and evidential relations are; but as an account of what it is for a self able to apprehend these logical and evidential relations to rationally regulate itself. This, I propose, is how we should understand RDE: as a normative principle of rational self-regulation. As such it may seem plausible. But one must be careful. If one has, for instance, a fresh perception and a previously held belief with inconsistent contents, it is a compelling idea that one is rationally required to regulate one's mental life by eliminating one of these two – either drop the belief, or reject the perception as non-veridical. The reason this is so is because one aims in one's belief-formation at truth, and given the inconsistency of belief and perception at most one of them can be veridical. But the same sort of truth-regarding norm requiring elimination of one of the conflicting states is simply not in play, when one has both a desire, and beliefs showing this desire is very unlikely ever to be fulfilled. One might be happier if one could somehow lose the unfulfillable desire, or alternatively lose one from the beliefs that together show its unattainability; but there is no demand imposed by a truth-guided consistency requirement on one's mental states that one do so. Similar remarks apply in relation to emotions. I think that Zagzebski's Diss Two, and RDE, wrongly conflate two very different phenomena: the requirement of epistemic rationality not to persist in holding inconsistent beliefs or belief-like states; and a supposed non-truth-related adaptive psychological process that restores harmony in the psyche, by either getting rid of unfulfillable desires, or lulling one into accepting non-evidenced beliefs. We should beware of accepting RDE too readily, due to its superficial similarity to a very different principle concerning solely truth-directed states that has an entirely different, and more secure, rationale.Footnote ¹⁹ Extending this theme I suggest below that RDE and Zagzebski's case for it posit a natural and normative unity within distinct adjustment mechanisms in the psyche that is simply not there.

Given the crucial role that Diss One plays in Zagzebski's case for RDE and the Meaty Argument that depends on it, I do not think that Zagzebski tells us nearly enough about the posited psychological phenomenon of dissonance. For starters, we are not told whether it is a sensation that tends to accompany awareness of conflict, or is more like a perception – intrinsically representing logical incompatibility between the contents of two of one's contentful states. It seems as though, like Hume ([1748] Reference Hume, Niddich and Selby-Bigge1975) when he introduces his distinction between impressions and ideas in terms of vivacity, Zagzebski believes she need only gesture towards a familiar phenomenon that the reader will immediately recognize. This reader did not experience such instant familiarity with the alleged phenomenon. There are, certainly, sensations of frustration, longing and the like when one has an unsatisfied desire, especially if it is strong and one recognizes it is unlikely ever to be fulfilled. There is nothing psychologically similar to this when one notes that a fresh perception is inconsistent with a previous belief, and immediately drops the belief. Even if the belief is deeply entrenched, and abandoning it as it were takes some effort, introspection in this reader's case does not discover anything I would be inclined to call dissonance; and certainly nothing instantiating a primitive phenomenal kind common to this case, and the keen yearning of an unfulfilled longing.

But suppose we allow Zagzebski both Diss One and Diss Two, and also some version of NN. Does this suffice to establish RDE? To establish RDE from this basis we need first, a strong version of Diss Two, that posits the natural response to conflict is always, not only sometimes, to eliminate one of the conflicting states – it is never natural to live with the conflict. Second, NN says that being rational is doing what comes naturally, but with some improvements. So to establish RDE one would have to show it is never an improvement on what comes naturally – viz. eliminating one of two conflicting states – to instead maintain both states, and the accompanying dissonance. NN itself does not and cannot show this. Nearly all the hard work in applying NN will be in determining in what respects we need to improve on how we naturally operate.Footnote ²⁰ I claim below that in some cases the epistemically mandatory response to dissonance is to maintain the conflicting states.

It gets worse. RDE, even if it were established, does not tell one how to rationally regulate one's epistemic-cum-affective life. When one has a pair of mental states M1 and M2 that conflict, there are three possible responses: to maintain both, living with the supposed ‘dissonance’ this involves; to eliminate M1; or to eliminate M2. A weak version of RDE that says only that it is always rationally permissible to eliminate one of a pair of conflicting mental states tells us nothing about what to do. A stronger version saying it is rationally mandatory to eliminate one tells us something; but it still does not tell us which of two conflicting states one must eliminate. NN and Diss Two, which tells us it is natural to eliminate one of a pair of conflicting states, thereby restoring harmony in the self, cannot help here.

The queries I have brought up about Zagzebski's case for RDE show serious flaws in the general account of human rational self-regulation she develops in EA Ch. 2. I do not think these general worries significantly undermine the case specifically for the rationality-on-reflection of epistemic self-trust that she offers as an application of this general account of rational self-regulation. Regarding epistemic self-trust there is a strong pragmatic case against living with dissonance; and ending dissonance by ceasing to rely on one's faculties is not a live option. But I think that as a general account of rational self-regulation RDE both posits too much – the premisses Diss One and a strong version of Diss Two are not empirically plausible; and does not achieve enough, since RDE cannot show which of two conflicting states one should eliminate, or even whether one should do so in every case.

More broadly, I take issue with Zagzebski's attempt to provide a unitary account of rational self-regulation in terms of maintaining harmony in the self by eliminating dissonance. There is no such unity in the maintenance of harmony in the self. On the contrary: there is on the one hand epistemic rationality, regulation of one's truth-directed mental states by the goal of maintaining consistency. And there are on the other hand various human adaptive psychological mechanisms that restore harmony in the self. These are not part of epistemic rationality. Thus, for instance, humans to some extent instantiate the mechanism of ‘sour grapes’ where one accommodates to thinking an unattainable good is not really so desirable; and exemplify wishful-thinking and self-deception more generally. One may have beliefs that show a deep desire one has is unlikely ever to be fulfilled – say, one's desire to end torture in the world, and one's beliefs about how various institutions actually operate. One can restore harmony in the self either by eliminating the desire; or by somehow lulling oneself into various false beliefs. But this is a case where, intuitively, the requirement of epistemic and broader rationality is to maintain both beliefs, and desire, and to live with the consequent dissonance.Footnote ²¹ Epistemic rationality and psychological mechanisms evolved to maintain psychic comfort come apart in this, and many similar cases. If it is natural in such a case to restore comfort by eliminating one of the conflicting states, then what is natural is contrary to what is epistemically rational. So a viable application of NN must concede that this is a case where doing better than what we do naturally means not doing the natural thing at all. Zagzebski's unitary global account of rational self-regulation as a single universally applicable mechanism of restoring harmony through eliminating dissonance wrongly precludes exhibiting the genuine conflicts there are between the demands of epistemic rationality based in truth on one hand, and psychological comfort on the other, in this and many other cases. Sometimes the mandatory epistemically rational response to conflict in one's states is to maintain them all, and live with the disharmony in one's self.Footnote ²²