Explanatory vs. Defensive reasons

In this post I want to approach the topic of the previous post from a different angle. I raised two questions about the U&C study: whether people believe the comparative ratings (Question 1), and what inference, if any, leads them to their ultimate verdict (Question 2). Either question would be natural to raise on its own. But besides having this in common, the questions are related in a deeper way. Here I want to explain how I take them to be related.

When we were discussing Question 2, we explored different sets of claims that participants might have taken to support the verdict that Michael would make a good police chief. We ended up with roughly two types of proposals.

At one extreme, the inference operated over propositions that the participants wouldn’t easily recognize as the ones they took to support the verdict. Nick’s inference from paradigms was like that. So was the inference from ‘Men make good police chiefs’, and ‘Michael is a man’, to ‘Michael makes good police chief’.

At the other extreme, the inference operated only over propositions that you could read off from the ratings that the participants were asked to make. That was the inference that I (perhaps wrongly) took John to emphasize (in connection with Matt Boyle) when he suggested that if participants believe that Michael would make a good police chief, then there must be something they would take to be a reason for this belief.

So there are the psychological factors that seem to most directly explain why the participants reach the verdict, and then there are things people offer as reasons for their verdict. In short, explanatory reasons come apart from defensive reasons. They come apart in Uhlmann and Cohen’s study, and in many other studies and everyday situations as well. The defensive reasons don’t control the verdict, and to that extent they’re not explanatory. The explanatory reasons aren’t ones that participants take themselves to be inferring from, and for that reason they wouldn’t show up in their own explicit defense of the verdict.

Some philosophers refuse to use the label ‘reasons’ for factors that are inaccessible to the subject in the way that the explanatory reasons are. I’ve been using “reasons” without that restriction.

Each kind of reason brings into focus a different dimension of what it might be to ‘take some considerations to support a conclusion’. That’s why cases where they come apart challenge us to refine our account of the kinds of inferences that bear on the rationality or irrationality of the subject.

Frege (in his Logic) characterized inference like this:

“To make a judgment because we are cognisant of other truths as providing a justification for it is known as inferring.”

As various philosophers have pointed out (e.g. Boghossian 2013, Chudnoff 2013), Frege’s insight remains if you expand the relata of inference to include more than judgments (e.g., supposition. Or for that matter, perceptual experiences). And it remains even if you relax the factive dimension of being ‘cognisant of providing justification’, so that you can infer Q from P even if P doesn’t really support Q, but you take it to. Frege’s main idea about inference  survives these adjustments. His idea is that you infer Q from P, only if you in some way take P to support Q.

Explanatory and defensive reasons bring out two kinds of ‘taking’. The explanatory reasons bring out the aspect of taking when your belief is sensitive to a consideration. The defensive reasons seem to be an obvious form of taking, since they’re offered in defense of the conclusion.

It is often said that merely explanatory reasons are outside the subject’s ken, whereas defensive reasons are inside the subject’s ken. This is a mistake. The difference is about accessibility to introspection, without much reflection. The explanatory reasons are control a person’s cognitive life.

Why does it matter whether participants believe the comparative ratings?

A chunk of the literature on inference is devoted to illuminating what kind of psychological states are involved in ‘taking’ some considerations to support a conclusion. If participants don’t believe the comparative ratings in the U&C experiment, then can they take them to support the verdict?

Well, the comparative ratings belong to the participants’ defensive reasons for their ultimate verdict. And didn’t I just write that there are two dimensions of taking – roughly: controlling (done by explanatory reasons) and reciting (enabled by defensive reasons)?

But here I think we need to distinguish between two kinds of defensive reasons: the ones you believe, and the ones you don’t believe but merely recite.

If you believe the comparative ratings, and offer them as defensive reasons, then there’s an obvious sense in which you take them to support the ultimate verdict. Your belief in the comparative ratings is unstable in the ways we’ve discussed, but perhaps belief can tolerate that kind of instability.

What if you don’t believe the comparative ratings? What if the shiftiness brought out by the experiment precludes you from believing them? Then it seems wrong to say that you are taking them to support the conclusion. (I’m assuming that there isn’t any other belief-like relationship you have to the propositions, such as supposition, intuition, or perceptual experience, that would make those propositions available as inputs to the inference). If you don’t believe that Socrates can fly, then you won’t take the proposition that Socrates can fly to support the verdict that Men can fly, with help from the proposition that Socrates is a man. You might recognize the corresponding argument as valid, but that’s not enough for you to use it in drawing an inference.

So the question about whether you believe the comparative ratings bears directly on whether it figures in an inference, when inference is constrained by Frege’s insight.

We could speak loosely and relax the notion of inference by letting go of Frege’s insight. That would leave us with something like superficial inference. Reciting things you don’t believe in defense of a verdict is a way of making noises that are typical of inference without actually making an inference using the propositions expressed by those noises (did you ever try to sing the prosody of philosophical argument to a toddler?). The participants may still be reaching the verdict via inference. But we should look elsewhere to find out which inference they’re making.

If the comparative ratings aren’t believed and don’t figure in any inference to the ultimate verdict, but the verdict is reached by inferences, then that leaves the stereotype (perhaps in the unpacked version Nick outlined) as inputs to the inference.

Some readers of the last post resisted the idea that the verdict was reached via inference at all. The factors leading to belief struck them as the wrong kind of thing to be involved in inference. In the next post, I want to discuss the difference between inputs to inference (such as the beliefs we’ve been exploring), and factors that regulate which inferences you make. If you thought that the stereotypes in the U&C case were not belief-like enough to be inputs to inference, they might still regulate which inferences a subject makes. In the next post I want to discuss that kind of regulative role by focusing on fears and preferences.


  1. Zoe Jenkin

    Hi Susanna, and thanks for the second post. I’m wondering whether there is any role at all for defensive reasons in the sort of taking that is relevant to inference. If we’re thinking of inference as a sort of mental action (a type of transition between cognitive states or sets of cognitive states), then it seems like the beliefs that you’re sensitive to during that actual transition have to be what’s relevant to the nature of the transition-not any reconstruction that you might give later on. So I’d think that it was only explanatory reasons that were relevant for the sort of taking that’s involved in inference, even if you do happen to believe the defensive reasons that you’d recite when asked. Does that seem right to you? If so, then it seems like the only relevance of defensive reasons to the nature of the inference is as a highly defeasible guide to what the right explanatory reasons might be.

    • Susanna Siegel

      thanks zoe, yes that seems right. i take neil’s point that some beliefs about qualifications (though not about comparative rankings) are explanatory rather than defensive. but yes, the issue for identifying the inference is which reasons are explanatory. and more broadly, the issue for epistemic evaluation is which factors are explanatory – whether those factors are components of an inference, or factors that regulate the inference.


  2. Hi Susanna,

    I think you’re right that it’s a stretch to see the comparative ratings as part of what leads the subjects to their initial evaluations, and also that it’s an open question whether they really believe the comparative ratings when they appeal to them as defensive reasons.

    I’m not convinced, though, that the comparative ratings can’t be explanatory, or even that they can’t figure as premises in an inference in some loose sense, just because they aren’t sincerely believed.

    Part of what motivates me to resist this is Tamar Gendler’s work on pretense: put very roughly, she argues that there are certain attitudes that are a form of “make-believe”, in that they are subjectively belief-like, and motivating in the manner of beliefs, but in some way don’t amount to the real article — perhaps because, as I would like to think of it, they fail in some deep way to be guided by the aim of truth. If there are such attitudes, and I would think that confabulatory post-hoc rationalizations sound like very good candidates, then they might play a role in our cognitive economy that’s similar to that of real beliefs, including enabling a certain kind of (perhaps itself pretend) inference that has a role in explaining why we hold onto certain judgments.

    What do you think of this? Is it compatible with what you are arguing? I guess it’s similar in some ways to the “superficial inference” picture in your third-to-last paragraph, but the “recitation” in question proceeds silently to oneself, and enables self-deception rather than the deception of others.

  3. Also, one more thought. A big reason why some of us are attracted to the idea that supporting reasons must be self-conscious, or at least potentially so, is that there are so many psychological and non-psychological factors that bear causally on our beliefs without counting as our reasons for them in any robust sense (cf. the literature on basing in epistemology), that the only way to identify which ones really are the reasons might be to query the subject herself. As I suggested just above, I think there’s lots of room for error and self-deception here, but still it’s hard to see how someone could count as believing P because Q while simply failing to know that she was reasoning in this way. I suppose this may come down to the loose vs. strict conceptions of being a reason that you distinguish early on, but it seems worth emphasizing why some of us incline toward strictness.

  4. Susanna,

    Thanks again for another great post. I like your distinction between explanatory and defensive reasons and I appreciate you, once again, laying out a helpful structure of possibilities.

    John has explained why some lean towards treating reasons as self-conscious and he has pushed on the distinction between defensive and explanatory reasons. I’d like to weigh in on this.


    John (et al): Indeed, there are many causal factors influencing our beliefs and if only self-conscious factors can count as “robust reasons,” then self-reports might be our best (only?) hope at learning about one’s reasons. But surely there also non-conscious causal factors influencing our self-reports about our reasons for believing. If there are, then I am not sure how appealing to self-reports overcomes the problem of unconscious causal factors.

    Perhaps I am missing something?

    But let’s suppose the self-reports somehow overcome the problem. Even so, we have to admit that self-reports are sometimes unreliable — maybe systematically so — so it seems that we need a reliable method of dissociating pseudo reporting (e.g., post hoc confabulation, etc.) from veridical reporting. Without this ability to dissociate the two, then it seems like two options remain:

    (1) Principled reasons to categorically accept/deny that self-reports about supporting reasons are veridical.
    (2) A middle view.

    Does anyone offer anything like option 1?

    Since I am unaware of a viable Option 1, I will mention a middle view. This could be an optimistic policy such as a veridical-until-otherwise-demonstrated policy. That is, self-reports about reasons could be treated as veridical until other evidence conflicts with their being veridical. In the U&C study, it seems like there is evidence that conflicts with the veridicality of some of the participants’ self-reported reasons, so — on this middle view — we could reject or discount the veridicality of the self-reports. This middle view will depend heavily on what counts as evidence and how we treat self-reports in the case of conflicting evidence (i.e., do we flatly reject them or just discount them?).

    Do people have thoughts about this middle view or other middle views?


    John, you have offered the possibility that even confabulated reasons could play a supporting role — i.e., even defensive reasons can be explanatory. This seems possible (and this might even be supported by empirical work; I don’t know). But even if this is the case, then it still seems that there would still be an important difference between the inaugural explanatory reasons and post hoc confabulated explanatory reasons. One difference might be this: inferences can do without the latter, but not the former.

    I wonder if people would accept this difference?

    • Hi Nick,

      These are all great points. For me, the reason to take self-consciousness seriously is not primarily to identify the causes of a belief, but rather to identify which of these causes are *reasons* in a robust sense. That is, I accept that beliefs are influenced by lots of (psychological and non-psychological, rational and non-rational, conscious and non-conscious) causal factors, but don’t think it’s possible to explain what makes such a factor the reason / evidence / justification / etc. for a belief except in terms of how the subject herself thinks.

      I think you’re right, though, that this doesn’t mean subjects are infallible on these matters! But my view, which as you’d expect is informed by Moran’s conception of first-personal authority, is that in order for a person to be wrong about the reasons (in the strict sense) for her beliefs, a special sort of explanation is needed, such as some sort of self-deception, repression, or alienation. That’s what I mean by saying subjects can’t “simply” be wrong about the reasons for their beliefs: since a subject’s self-conscious take on her reasons is partly constitutive of what those reasons are, the only way for the two to come apart is through a deep breakdown in her rationality. And if this kind of thing were normal rather than exceptional, then we’d no longer be justified in thinking of ourselves as rational agents *at all*.

  5. Susanna Siegel

    hi john, thanks. since ‘inference’ can label many different things, i’ve been regimenting the notion in a way that helps us draw some further distinctions that matter for identifying different types of epistemic evaluability.

    the regimented notion i’ve been working with respects Frege’s insight. as Zoe pointed out, the comparative rankings don’t play that role.

    when we focus on the role of the defensive reasons, do those bear on the subject’s rationality? yes. you can be a better or worse liar, the kid can give more or less intelligent rationalizations for why they won’t clean up their room. (‘i have to practice my violin’ is pretty good, ‘my tongue hurts’ is worse). My point was that when we evaluate defensive reasons, we are not evaluating the thinker’s inference in the regimented sense.

    i like the way Gendler has helped us think about pretense and the wide range of belief-like states in the mind. I disagree with her idea that some of these belief-like states are ‘a-rational’.

    in your last comment you said:
    “A big reason why some of us are attracted to the idea that supporting reasons must be self-conscious, or at least potentially so, is that there are so many psychological and non-psychological factors that bear causally on our beliefs without counting as our reasons for them in any robust sense (cf. the literature on basing in epistemology), that the only way to identify which ones really are the reasons might be to query the subject herself.”

    i think the question “Which psychological factors really are the reasons?” does not help us inquire into the epistemically evaluable parts of the subject’s mind, until we specify what kinds of reasons we are talking about. when you ask that question, what are you holding fixed about what counts as ‘the real reasons’? some philosophers hold fixed that reasons are things the subject is potentially self-aware of. if we want a way to describe how the subject views the epistemic support relations between her beliefs (or her actions), then this is a useful notion of reason. but it won’t track the psychological person level factors that explain how the subject arrives at a belief. and it won’t track the inputs to the subject’s inference, in the regimented sense of ‘inference’ that comes from Frege (as Zoe pointed out).


    • Hi Susanna,

      What about the idea that talk of reasons in this regimented sense picks out what we are searching for when we ask things like “Why should I believe that?” or “Is this a good reason for so-and-so’s belief?”, etc. I.e. reasons in this sense are bound up in our normative evaluation of beliefs as justified or unjustified, as knowledge or mere opinions, and so on. I admit that it takes further argument to show that there is a connection between reasons in this strict sense and a subject’s self-consciousness. And as I said in my reply to Nick, I agree that there are many factors other than our reasons-in-this-strict-sense that explain how our beliefs are formed and sustained. But still this strict concept of reasons seems really important, and it would be more than a bit disturbing to find out that the reasons that justify our beliefs, etc. don’t play any distinctive role in our cognitive economy.

  6. Susanna Siegel

    hi john and nick,

    john has brought up the idea that defensive reasons might be explain the persistence of a belief, even if they do not explain why the belief was formed in the first place. (nick called this ‘supporting’ vs ‘inaugural’ reasons).

    i think it’s an important project in the philosophy and psychology of belief to identify the conditions under which non-inaugural reasons explain the persistence of belief, and the conditions under which they’re not even doing that.

    it seems useful to distinguish between a de dicto supporting role and a de re supporting role. if defensive reasons play a de dicto supporting role, then the belief would not persist without some set of defensive reasons – but no specific set is such that it is needed. in that case, the defensive reasons one actually has provide support for the belief de dicto, but not de re.

    the U&C experiment, we can see that the comparative ratings don’t play a de re role, because you can switch them and still have the belief. do they play a de dicto supporting role? i don’t think the experiment speaks to that question. perhaps they don’t even play a de dicto supporting role.

    it seems plausible that some defensive reasons play a de dicto supporting role, while others don’t even do that. i suspect that we can’t read backwards from a defensive reason to its supporting role.

    i find it helpful to compare the case of individual beliefs to the case of cultural assumptions. think about chattel slavery. slavery tells a lie: the lie of inequality of human worth. it is such a big lie, that it has to be justified. and it is, by theories about natural cognitive inequalities. here’s a case where there’s something analogous to defensive reasons for belief. the analog of defensive reasons is something like ideology, and the analog of belief is what i called a ‘cultural assumption’. the cultural assumption may not begin its life as an assumption, it may begin its life as a set of social practices, and the assumptions get articulated afterward. we have to relax our standards of precision a bit to talk about this phenomena but that’s because of the nature of the phenomena.

    could there be chattel slavery without justifications for it that masquerade as descriptive truths about reality? (compare: the normative claim that animals should not run the farm is backed up by descriptive claim that animals are incapable for running the farm. [for vivid discussion and challenges, see the book for toddlers “Click Clack Moo”, where cows type messages on typewriters issuing various demands to their farmer.]) many social arrangements would not be tenable without something like defensive reasons.

    some beliefs might operate in the same way. that’s a question for the psychology and philosophy of belief: when are defensive reasons needed for support, and when aren’t they.

    there’s a separate question, which is whether it is in the nature of belief for there to be reasons that the subject offers for the belief. john in an earlier comment denied this. (that surprised me. i think some people who think there have to be self-conscious (or potentially self-cs) reasons for belief think that that is a fact about the nature of belief. it’s a bit like a doxastic analog of the Guise of the Good hypothesis.


    • Okay, one more reply (for now! — I said last week that I had a billion questions, so there is a long way to go).

      The distinction between de re and de dicto supporting roles is really helpful, Susanna, and I agree that in any case where a belief (or belief-like attitude) that plays only a de dicto supporting role in supporting a belief, this attitude should be counted as a reason for it only in a very weak sense, as the “deeper” explanation for the belief should appeal to something else. Indeed, perhaps this would be a case of co-variance without real causal dependence (here I am thinking of John Campbell’s work, e.g. https://socrates.berkeley.edu/~jcampbel/documents/Causation.pdf): A (the belief) and B (the defensive reasons) seem connected, but only because there is some C (the underlying cultural assumption, or whatever) that is the common cause of both of them. So the appearance of A’s depending on B will be illusory.

      Regarding beliefs without reasons, the sort of case I had in mind was one where a certain belief is in some sense epistemically “basic”, i.e. held because it seems true or obvious, but without further justification for it. As I think I suggested, a similar phenomenon in the practical domain might be when a person does something “for no particular reason”, though not without seeing it as somehow good to do (thus it’s compatible with the Guise of the Good).

Comments are closed.

Back to Top
%d bloggers like this: