2. THE ROLE OF THE COGNITIVE IN EPISTEMIC THEORIZING

Unsurprisingly, theorizing in epistemology has been largely blind to REC (since prominent discussions of REC are recent). This is not to say that theorizing has been largely blind to cognitive kinds simpliciter, some of which might be instanced in radically extended ways. Indeed, the opposite is true. Epistemologists often take the cognitive to play a central role in their theories. The point is simply that epistemologists haven't formulated these theories in such a way that it would matter whether cognition is radically extended or not. In the past, theories of epistemic rationality and justification have been fleshed out in terms of cognitive kinds like belief, experience, perception, memory, inference, (cognitive) dispositions to believe, etc. Theories of knowledge have typically used the truth conditions of beliefs – a cognitive feature – to specify the factivity and anti-luck requirements for knowledge.Footnote ⁷ Also, it is commonly supposed that truth conditions of beliefs dictate something at least fairly central to their epistemology: a kind of objective success (truth) and a kind of objective failure (falsehood).Footnote ⁸ However, when epistemologists have invoked the cognitive in their theories, there's no evidence that it has been with any consideration for REC in particular. Certainly, invocations of the cognitive in epistemology haven't included any restrictions as to whether the cognitive in question is radically extended or not. For instance, epistemologists haven't tended to say things like “Beliefs can be properly inferred from other beliefs unless the latter beliefs are radically extended” or “Knowledge is factive and safe unless it is radically extended knowledge” or “If ducks quack, a belief that ducks quack is objectively successful even if not justified – unless, of course, the belief is radically extended.”

It is natural to be suspicious of blanket generalizations about practice in epistemology, but it is fairly easy to check that, at least up until very recently, the idea of REC hasn't made an appearance in standard anthologies in (“analytic”) epistemology.Footnote ⁹ Nevertheless, given the importance of the point for latter discussion, it may be worth surveying some examples to see that, in fact, prominent theories in epistemology are formulated so as to be indifferent to the possibility of REC. I offer five:

Other theories of justification in the literature are also indifferent to the possibility of REC:

The neutrality of epistemology with respect to REC does not end with theories of justification:

While an exhaustive survey is not possible, (1)–(5) provide enough of a sampling to support the claim that theorizing in (at least “analytic”) contemporary epistemology hasn't generally allocated any significance to radically extended cognition. The radical extension of cognition hasn't been taken to explain the presence or absence of epistemic properties.

The question now arises, however, whether this blindness to REC is problematic. Epistemic statuses are, like all normative or value-laden statuses, subject to supervenience constraints.Footnote ¹³ There cannot be differences in epistemic status without underlying (mental, functional, or material) differences. Epistemic assessments do not cohere if they attribute epistemic differences without compensatory differences of other kinds. One might propose that the difference between cognition that is and isn't radically extended just is the kind of difference that could ground epistemic differences (even though epistemologists haven't taken them to). The distinction between cognition that is and isn't radically extended might be theoretically important for epistemology in just this way. But, how plausible is this proposal?

Not very, I think. While I don't have any knockdown arguments, it strikes me that the burden of proof rests on any proponents who, in effect, favor special clauses in epistemological theories to cover cases of REC. The basic line among believers in REC is that the distinction between those processes, states, and character that are radically extended and those that aren't doesn't matter for the philosophy of cognition.Footnote ¹⁴ However, if this distinction doesn't matter for cognition, it's hard to see why it should matter for the assessment of cognition as such – and assessment of certain (doxastic) kinds of cognitive as such is, of course, precisely what epistemic assessment is.

Thomas Jefferson – or more specifically, how he was – is arguably relevant to the epistemic assessment of beliefs about Thomas Jefferson (e.g. whether they meet epistemic standards for success) at least partly as a consequence of the fact that beliefs about Thomas Jefferson only qualify as beliefs of this kind due to their relation with Thomas Jefferson.Footnote ¹⁵ This illustrates the general principle that if Y is relevant to whether X is cognitive in a particular way, then Y may well be relevant to whether X is, in its role as cognitive in that particular way, good, bad, etc. However, a qualified converse of this principle is also plausible. If Y isn't relevant to whether X is cognitive in a particular way, it is not at all clear why Y should be relevant to whether X is, in its role as cognitive in that particular way, good, bad, etc. By way of example, beliefs about Australia do not qualify as beliefs of this kind due to their relation with Thomas Jefferson. In specific circumstances, facts about Thomas Jefferson might be relevant as such to epistemic assessment of beliefs about Australia, but a special explanation is presumably required as to why – e.g. an explanation of appropriate relations between Thomas Jefferson and Australia. Certainly, it is not at all clear why facts about Thomas Jefferson would be generally relevant as such to epistemic assessment of beliefs about Australia. Consider a further analogy. It is not at all surprising that relations between Jim and Jane matter for the evaluation of Jim qua father if Jim is a father in virtue of his relation to Jane (because Jim is the father of Jane). But, if Jim is not a father in virtue of his relation to Jane, it is not at all clear why the relations between Jim and Jane should matter for assessing Jim as a father. Of course, they might matter, but it is reasonable for us to expect an explanation of the relevance. Similarly, in the case of REC, being radically extended isn't (so we are told) relevant to being cognitive per se, so it is not at all clear why radical extension should generally be relevant to epistemic assessment of the cognitive, i.e. assessment of the cognitive as such.

Let me stress further that I am not making a definitive claim about the relevance of REC to epistemology. I am merely indicating that it is reasonable to expect an explanation as to why radical extension should be relevant per se to the epistemic when it isn't to the cognitive. In the absence of this explanation, I contend it is appropriate to proceed on the assumption that the radically extended distinction – i.e. the distinction between those processes, states, and character that are radically extended and those that aren't – isn't one that should appear in theories in epistemology. If this assumption is true, then the blindness that epistemologists have shown to the possibility of REC isn't obviously problematic. Epistemologists haven't been saying things like “Knowledge is factive and safe unless it is radically extended knowledge,” but they shouldn't have been. Rather, they should have been formulating their theories using cognitive kinds that are neutral with respect to radical extension – just as they have.

It's worth noting that any defense of epistemic parity principlesFootnote ¹⁶ipso facto goes to support the idea that (at least certain) epistemological theories should not include carve-out clauses for REC. For example, Carter and Pritchard (forthcoming) defend the following principle:

Any theory of defeasible warrant that included a carve-out clause for radically extended cognitive processes would violate this principle because it would allow for a case where the presence of defeasible warrant turned on whether the REC carve-out clause applied. So, if this epistemic parity principle is true, theories of defeasible warrant should not include carve-out clauses for REC of this sort.

Even still, the failure of certain epistemic parity principles need not undermine the main claim of this section. For instance, Carter and Pritchard (forthcoming) argue against this principle:

However, this principle fails to hold for general considerations that do not stem from the possibility of REC. Any difference in feature, F, between two beliefs, B and B*, can make for a difference in knowledge between B and B* if the agents are in a position to rationally suspect that beliefs with feature F are generally or locally unreliable. Thus, the fact that a belief was formed by a REC process could, in principle, go towards explaining why the belief isn't knowledge because an agent might be in a position to rationally suspect that beliefs formed by REC process are generally unreliable – or, at least, unreliable in the present circumstances. Nevertheless, no special REC carve-out clauses are required in the theory of knowledge since this last point can be fully generalized so as to abstract away from REC in particular. Consequently, the main claim of this section – that radical extension is not relevant per se to the epistemic – goes untouched.

3. RADICALLY EXTENDED COGNITION IN COUNTEREXAMPLES

If epistemologists shouldn't be carving out exceptions for the epistemic assessment of REC, it becomes hard to see how REC could be an important new source of counterexamples to epistemological theories. In this section, I argue this is exactly right. I illustrate my argument by considering two potential counterexamples from the philosophical literature.

First, though, let me offer a principled explanation as to why REC couldn't be a new source of counterexamples in epistemology. As I pointed out in the immediately preceding section, one of the guiding ideas of the REC literature is that cognitive roles occupied by internal biological entities could, in principle, be occupied by radically extended entities instead.Footnote ¹⁷ However, if this guiding idea is correct, one might also expect that cognitive roles occupied by radically extended entities could, in principle, be occupied by internal biological entities. Consider, then, a hypothetical counterexample to an epistemological theory that invokes REC. It would appear that there should be a structurally analogous case in which any radically extended entities playing cognitive roles in the original counterexample are replaced by internal biological entities instead. Assuming the epistemological theory, while invoking various cognitive distinctions, is nonetheless insensitive to REC per se, the structurally analogous case should be just as effective as a counterexample. In other words, epistemological theories are only prone to counterexamples with REC, if they are prone to counterexamples without it.

It will help to illustrate how one might construct a counterexample without REC from an example with. Consider first a potential counterexample from Pritchard (Reference Pritchard2010). He considers a variation on the famous Otto example of Clark and Chalmers (Reference Clark and Chalmers1998). In this variation, Otto's (allegedly) radically extended memory system – in the form of a notebook – is set up for him by his wife. This (allegedly) radically extended memory system delivers true beliefs, but Pritchard suggests these true beliefs are primarily creditable to Otto's wife. This, he judges, may create a problem for the robust virtue theories of knowledge described earlier. Otto knows, but these true beliefs aren't attributable as such to the exercise of his (radically extended) rational abilities (or so Pritchard alleges) because they are attributable to the wife's designs and execution instead.

For our purposes, it doesn't matter whether Pritchard's case is in fact a (successful) counterexample (even though I think it isn't).Footnote ¹⁸ Rather, the issue is whether there is a possible case without REC that would be equally effective as a counterexample.Footnote ¹⁹ I would argue any problem for robust virtue theories of knowledge is equally well created by a case in which the memory system clearly isn't radically extended. For instance, imagine Otto's wife designs a miraculous pharmaceutical drug. In combination with a very strict diet and exercise regimen, which the wife dutifully enforces, this drug bio-enhances (in a distinctively non-natural way) Otto's memory so that it functions as well as the radically extended memory system in the other case. If true beliefs delivered by the radically extended memory system are primarily creditable to Otto's wife, then true beliefs delivered by this bio-enhanced memory system are primarily creditable to Otto's wife too. Notice that in generating a new example, I have simply replaced radically extended entities playing cognitive roles with internal biological entities playing the same roles. As a result, the two cases are structurally analogous in the relevant respects so that they should function equally well as counterexamples. Because in the latter case the memory character traits are not radically extended, it becomes clear that REC is not an essential ingredient for constructing a counterexample.

The same kind of procedure also works for a second potential counterexample from Carter (Reference Carter2013). Carter (Reference Carter2013) is interested in the thesis that knowledge is incompatible with environmental luck.Footnote ²⁰ By way of background, environmental luck might be illustrated by a case in which you form (as it happens) a true belief that you have a five dollar bill in your wallet from having seen it.Footnote ²¹ Ordinarily, the visual experience of the bill might well suffice for knowledge. However, in this particular case, there are a large number of counterfeit five dollar bills in the local area due to a recent visit by North Korean spies. These counterfeit five dollar bills could only be detected by special machines at the Federal Reserve. In this case, it is lucky that you have a true belief that you have a five dollar bill in your wallet even though there is nothing defective about the process by which you form this belief. Lack of defect in the belief-forming process is, in fact, a distinguishing feature of environmental luck. A true belief is environmentally lucky if and only if it is formed in a way that (1) is generally effective at yielding true beliefs, but (2) given local environmental conditions, is not locally effective at yielding true beliefs, but (3) it is rational to believe that the local environmental conditions are not special in this respect (even though, in fact, they are).

It is commonly held that knowledge is incompatible with environmental luck.Footnote ²² Carter (Reference Carter2013) worries that this case might be problematic for the commonly held thesis:

Ignoring the possibility of REC, the jokester case looks like a relatively clear case of environmental luck. There is nothing intrinsically defective with the process by which Otto comes to have an occurrent belief that his appointment with his doctor is at noon. Ordinarily, relying on his notebook – presumably, a kind of self-testimony – does not leave him with false occurrent beliefs; generally, this procedure is effective. However, given the unusual presence of the jokester, this procedure is locally ineffective. Consequently, it looks as if the conditions for environmental luck are met. If knowledge is incompatible with environmental luck, then it seems that Otto does not know in the jokester case.

This last conclusion isn't so clearly right, however, if we suppose that Otto's memory system is radically extended into the notebook in the way that Clark and Chalmers (Reference Clark and Chalmers1998) suggest. Relying on his notebook is not merely a kind of self-testimony for Otto (if it is self-testimony at all); it is ordinarily a kind of remembering. Moreover, the jokester is literally tampering with Otto's memory by fudging the appointment times listed in the notebook. There continues to be a sense in which the belief about the doctor's appointment is lucky; it is lucky that Otto's memory vis-à-vis his doctor's appointment hasn't suffered interference from the jokester. Very easily the jokester might not have overlooked the doctor's appointment listed in the notebook. However, it is not obvious that this kind of luck is incompatible with knowledge. Perhaps, Otto knows in the jokester case after all.

Carter (Reference Carter2013) worries that, if jokester is understood to be a case of REC, it threatens to undermine the thesis that knowledge is incompatible with environmental luck. jokester, he suggests, continues to be similar to paradigmatic cases of environmental luck – for example, the counterfeit money case that I introduced earlier. Thus, Otto should not know in jokester, if subjects don't know in paradigmatic cases of environmental luck. As it happens, I find this reasoning problematic: however similar it might be to paradigmatic cases of environmental luck, jokester only has to be different in one relevant respect to merit a different kind of verdict.Footnote ²³ Moreover, as I highlighted earlier, cognitive distinctions are generally relevant to proper epistemic assessment. If Otto's memory system is radically extended into his notebook, then the problem (if any) in jokester isn't that local environmental conditions aren't favorable for Otto's acquiring a true belief about the doctor's appointment by ordinary means. Rather, the problem appears to be located inside Otto's mind with his memory storage. In other words, jokester is not a case of environmental luck; it is (at least superficially) similar to cases of environmental luck, but cognitively different. I don't see why we should expect that this cognitive difference won't matter to determinations of knowledge.

Nevertheless, we might suppose Carter (Reference Carter2013) is right and jokester does threaten the thesis that knowledge is incompatible with environmental luck. If so, then my view is that there should be some case that doesn't invoke REC, but threatens the thesis equally well. In fact, Carter (Reference Carter2013) suggests a possibility,Footnote ²⁴ but a more closely analogous case might be this one:

Short of skepticism about whether there really is REC in cases like jokester, I don't see any reason to think that there should be a difference in verdict between jokester and bio-jokester. If one grants that Otto's memory is radically extended in jokester, then I don't see any basis for thinking that Otto fails to know the time of the doctor's appointment while Otto* knows or vice versa. Moreover, if the similarity between the jokester case and cases of environmental luck does, in fact, indicate that verdicts about knowledge in those cases should match, then the same should be said of the bio-jokester case and cases of environmental luck. The bio-jokester case is similar to cases of environmental luck in all relevant respects – or, at least, it is if we assume that Otto and Otto* are cognitively alike. (Note that the conclusion from the previous section implies that the obvious difference between Otto and Otto* on this assumption – i.e. how far cognition extends – is not relevant to epistemic determinations.) Thus, if jokester provides reason to worry that knowledge is compatible with environmental luck, so does bio-jokester. The upshot is that Carter (Reference Carter2013) is wrong to suggest that Otto-style REC presents a special problem for the thesis that knowledge is incompatible with environmental luck. Any such problem for this thesis exists independently.

I began this section with a principled explanation of why we ought to expect that theories prone to counterexamples involving REC should also be prone to counterexamples without REC. The cases of Pritchard (Reference Pritchard2010) and Carter (Reference Carter2013) have provided me the opportunity to illustrate how to manufacture potential counterexamples without REC from potential counterexamples with it. To do so, one simply replaces instances of REC with equivalent instances of non-radically extended cognition, mutatis mutandis. I suggested in the previous section that theories in epistemology should not have special carve-out clauses for REC. If they do not, potential counterexamples manufactured by the illustrated method will serve equally well as the original counterexamples with REC.

It follows that any epistemological theory that is materially adequate when restricted to cases without REC will be materially adequate tout court. (If it weren't materially adequate tout court, then there would be a counterexample. There are two possibilities – (a) that the counterexample invokes REC and (b) that it doesn't. If it does, then another counterexample can be manufactured that doesn't invoke REC. Either way, there is a counterexample that doesn't invoke REC. This proves the contrapositive.) Thus, for the purposes of theorizing in epistemology, one can simply ignore the possibility of REC and focus on clear cases of non-radically extended cognition. Or, put another way, the relationship between REC and epistemology is not that the former has consequences for the latter. Unlike, for example, the possibility of epistemic luck, the possibility of REC makes no difference to epistemology.

In closing this section, it's worth briefly addressing a particularly dramatic way in which cases of REC might alter theorizing in epistemology – that of inspiring a paradigm shift. For instance, Vaesen (Reference Vaesen2014) suggests that reflecting on REC might result in a paradigm shift from “analytic” epistemology to epistemology of a more pragmatist bend. In particular, by thinking about how Otto's notebook is interacting with biological-situated states and mechanisms to get him where he's going, we might be better positioned to appreciate the pragmatic “steering” role of doxastic elements (including doxastic activities) as well as the importance of considering the doxastic system and its dynamic activities as a whole rather than individual beliefs in isolation.Footnote ²⁵

For the record, I don't think that “analytic” epistemology has any particular difficulty in adequately addressing the “steering” role of the doxastic,Footnote ²⁶ nor does it incorporate an unduly narrow focus on the evaluating individual beliefs,Footnote ²⁷ but, for present purposes, we can put those issues aside. Nothing put forward in this paper commits me to thinking that REC cannot inspire better epistemological theorizing. Rather, the relevant point is that cases of REC would not be needed to justify theoretical revisions in epistemological – including revisions as dramatic as a paradigm shift. If Otto's notebook is especially different from biological memory in its steering role, connectivity, or general dynamic activity, then motivation for judging it to be a case of REC akin to biological memory is lost. I assume, then, for the sake of argument, that biological memory exhibits similar properties. It follows that if Otto's notebook is indicative of the pragmatic “steering” role of doxastic elements or the importance of considering the doxastic system and its dynamic activities as a whole, then the correlative biological memory is as well. So, even if the target is as large as ‘analytic’ epistemology itself, using a case of REC seems inessential.Footnote ²⁸

4. RADICALLY EXTENDED COGNITION AND THE PRACTICE OF EPISTEMIC EVALUATION

Nevertheless, as indicated in the introduction, REC does have some consequences for the distribution of epistemic properties in the world. In other words, although epistemological theorizing should be unaffected by REC, the practice of epistemic evaluation should be affected. The reason is that proper epistemic evaluation requires sensitivity to whether something is a proper unit of epistemic evaluation. One should not evaluate an accidental, unintended tripping as irrational merely because it fails to satisfy the preferences of the person who tripped; tripping is not usually an action and so is not properly evaluated as such. Unintended tripping isn't a proper unit of action assessment. Similarly, one should not evaluate some scribbling in a notebook as an irrational state of mind, if it isn't a state of mind at all. If the scribbling in the notebook isn't a belief, then it shouldn't be epistemically evaluated as a belief would be. On the other hand, if the scribbling in the notebook is a belief, then it should be epistemically evaluated as a belief would be. Judgments that token processes, states, capacities, and mechanisms are instances of some doxastic-related variety of the cognitive commit one to epistemically evaluating them as such. Moreover, epistemic evaluation of token processes, states, capacities, and mechanisms commits one to judgments that these tokens are cognitive. In addition, the way in which one epistemically evaluates something typically commits one to judging it to be a particular kind of cognitive. For purposes of illustration, consider the contrast between intuition and belief. An intuition and a belief that bodiless pain is possible each clash in different ways with a belief that bodiless pain isn't possible. Having beliefs with contradictory contents is markedly worse, from an epistemic standpoint, than having an intuition and a belief with contradictory contents.Footnote ²⁹ Consequently, epistemically evaluating a mental state token as very strongly clashing with another mental state token with contradictory content gives rise to certain commitment. One is committed to judging that both tokens are beliefs rather than intuitions.

The fact that REC has some consequences for the distribution of epistemic properties in the world is obviously related to the very tight relationship between judging something to be cognitive in a particular way and epistemically evaluating it. However, this same very tight relationship is just as related to the fact that the distribution of epistemic properties in the world has consequences for REC. (Consequence, in the relevant sense, is a metaphysical relation, but not an explanatory one. Consequence is not asymmetrical; the distribution of A-properties may have consequences for the distribution of B-properties even if the reverse also holds true.) Moreover, as I pointed out, there are, in fact, two way rational commitments between judging-as-cognitive and evaluating epistemically, so one might expect two-way consequences between the cognitive and epistemic domains. Notably, it is not clear that our grasp on what is cognitive should be in any way prior to our grasp on how epistemic properties in the world will be distributed.Footnote ³⁰ Indeed, even if the cognitive is metaphysically prior to the epistemic,Footnote ³¹our sensitivity to the cognitive may be mediated in a variety of ways by our sensitivity to the epistemic. As a result, it may be that the best way to weigh the merits of any particular claim of REC is by turning to the epistemic domain. In effect, one should be able to defend a judgment that some radically extended token is of a particular kind of cognitive by defending the epistemic evaluation of it as such.

For example, in other work, I have defended the view that certain radically extended belief-forming processes – pen and paper computations of large arithmetic sums – are cognitive processes by defending the view that these processes are subject to the corresponding kind of epistemic evaluation.Footnote ³² In effect, aspects of these radically extended processes determine whether resulting belief tokens are held because they are evidentially supported, a property these processes would have only if they were part of a distinctly cognitive performance. In fact, one should also be able to defend a judgment that some radically extended token isn't of a particular kind of cognitive by defending the view that it shouldn't be epistemically evaluated in the requisite way. I will turn to this strategy shortly when I argue that certain radically extended processes are not instances of remembering because they are subject to the same kind of epistemic evaluation as genuine remembering is.

Before I do so, however, it is important to emphasize that the effectiveness of these strategies for diagnosing REC turns on the results from §§2–3. Suppose that a radically extended token is alleged to be of a certain cognitive kind, yet this token doesn't exhibit the same epistemic features that an otherwise analogous non-radically extended token of this cognitive kind would. Because REC could not be a genuinely new source of counterexamples in epistemology, this could only mean that the allegation is false. However, if, pace my earlier discussion, REC is a source of counterexamples, then the allegation might be true anyway. The epistemic difference might be explained by the difference between radically extended and non-radically extended cognitive kinds. In other words, noting this epistemic difference might provide, for example, the rational impetus to revise an epistemological theory so as to include a carve-out clause for REC. If epistemic difference might be attributable to radical versus non-radical extension, epistemic difference need not undermine cognitive sameness. Thus, it is precisely because epistemological theorizing is immune to the possibility or actuality of REC that the presence of epistemic features can be used, in principle, to straightforwardly diagnose the presence of the cognitive. In other words, the methodology of using the epistemic as a “mark of the cognitive” crucially depends on the neutrality – claimed in §§2–3 – of epistemology with respect to radically versus non-radically extended cognition.

5. AN ALLEGED CASE OF RADICALLY EXTENDED COGNITION

Whether, in practice, we can use the presence of epistemic features to diagnose the cognitive depends on whether, in fact, we might have a prior awareness of the epistemic features. It is a theoretical possibility that any awareness of epistemic features comes via our overt knowledge of the cognitive. If so, we're not going to learn about the distribution of cognitive features in the world by learning about the distribution of epistemic features since we learn about the latter by already having learned about the former. I've proposed the methodology of looking for the cognitive in the epistemic. But, for this methodology to pan out, we need to be more competent at making explicit epistemic evaluations than we are at judging whether something falls under a particular cognitive kind.

As it happens, we are, at least sometimes, better attuned to normative properties than non-normative one. Famously, subjects do better at identifying potential violations of “If a person is drinking, the person is over the age of twenty-five” than potential counterexamples to “If a card shows an even number on one face, then its opposite face is red.”Footnote ³³ A possible explanation is that detecting cheaters is of far more importance than uncovering counterexamples to abstract generalizations. For similar reasons, it seems plausible enough that we might be more competent at making epistemic evaluations than we are at judging whether something falls under a particular cognitive kind. The epistemic credentials of a person's beliefs or character traits may be, in many cases, more relevant to how to treat the person than the particulars about what exactly in the vicinity of the person is cognitive and in what way. Perhaps, then, due to its practical import, human beings have developed a more intuitive grasp of the epistemic than they have of the cognitive.

Indeed, it is enough for my purposes that it is theoretically possible that we are better at making epistemic evaluations than judgments about the cognitive, and, also, that there are possible explanations for why this might be so. When it comes to justifying the methodology of looking for the cognitive in the epistemic, it wouldn't be wise to rest much on a just-so story. The best way to support this proposed methodology is rather to show how it works in a particular case. I propose to consider a case of remembering – recorded by Sutton et al. (Reference Sutton, Harris, Keil and Barnier2010) – that is advanced as a potential instance of REC.Footnote ³⁴ The case involves a husband and wife remembering in tandem the beginning of their relationship:

The question I will consider is whether the husband's process of remembering, say, that it was Monday night when the pestering began is radically extended into the wife (and vice versa). (There is at least one other theoretical possibility for REC here: the husband and wife could collectively constitute another distinct epistemic agent who is remembering things. Since my purpose is only to illustrate a particular methodology for diagnosing REC, it will suffice to focus only on the possibility that each partner's remembering is radically extended into the other.) I will argue that epistemic evaluation points towards a verdict of “No.” In so doing, I will illustrate the effectiveness of, at least sometimes, using the epistemic as a “mark” of the cognitive.

Before sketching a case for the “No” verdict in the next section, however, it is helpful to consider some of the various alternative interpretations of the case in question. To begin with, it might well be conceded that the remembering processes of the husband and wife are interestingly “coupled” even though the remembering processes of one do not radically extend into the other.Footnote ³⁵ Consider that “coupling” might simply entail capacities for remembering are impoverished without the partner's aid. Presumably, remembering is a performance that each is capable of only in suitable circumstances. When memories are not particularly robust, these suitable circumstances for remembering might involve being in a state of mind typically brought about by priming or prompting. For instance, it might be that although the husband may have a stored memory that the pestering began on a Monday night, retrieval is typically only possible when the wife helps to “jog” his memory. If this is how we should understand the quoted case, then we need not think about it as a case of REC at all even though the husband's memory capacities are impoverished without the help of his wife (thus showing them to be “coupled”). The remembering might be happening entirely within the husband's head even though the background conditions required for this remembering process are typically brought about through conversation with the wife.

Another interpretation of the case worth considering is that at least some of the appearance of remembering in the case is, in fact, confabulation-like fabrication. For instance, it may be that insofar as the husband takes himself to be remembering, he is mistaken. Really, the husband believes on the basis of the wife's testimony, but perhaps this belief comes so naturally that the husband takes himself to be remembering. In fact, the husband has forgotten which day the pestering began, but gains a new belief about it upon hearing the wife.

The husband's remembering would be a radically extended cognitive process if the process is partly located in the wife. For instance, it could be that the husband's memory has been partly offloaded onto and subsequently stored by the wife. However, the practical necessity of the wife for the husband's remembering is not enough for REC. Both the priming and fabrication possibilities intimately involve the wife, and they have to be ruled out if the remembering process is to be established as radically extended. (I'm ignoring, of course, the further possibility that the case is one where the husband's memory is autonomous, i.e. he could simply remember everything on his own without the wife – and vice versa. Such a case does not appear to support the contention that the memories of the wife and husband are “coupled,” much less that remembering is radically extended. So, for the sake of argument, I'm simply granting the case is not of this sort.)

6. EPISTEMIC EVALUATION AS A GUIDE TO COGNITION: A CASE STUDY

The important thing to notice is that it is epistemically significant whether the priming, fabrication, or radically extended cognition scenario obtains. For instance, if the husband is fabricating, there is a deficit in his self-knowledge, but if he is remembering (either through priming or radically extended cognition) there's no such (epistemic) deficit. More importantly for purposes here, where the remembering is happening has an impact on how we evaluate the resulting the belief. Suppose, for instance, that there is unusual activity located within the wife's head. If the husband's cognition radically extends into the wife, then this is potentially a problem with the husband's own cognitive processing that could be relevant to the quality of his cognitive performance, thus impacting the epistemic credentials of any resulting belief. If, however, the husband's cognition does not radically extend into the wife, then although he might be misinformed by the wife, his cognitive processing is itself unaffected.

To make these suggestions more vivid, imagine that the wife takes herself to be remembering the day of the cooking classes that she was taking when she met her now husband, but she confuses these cooking classes with wine tasting classes that she took in the same location a few months earlier. Indeed, she confuses them with the wine tasting classes in the sense that her memory of the day of the week of the cooking classes counterfactually depends on the day of the week of the wine tasting classes rather than on the day of the week of the cooking classes (as it should). Moreover, we might suppose she might even recognize this confusion by realizing that the classes she is remembering took place at a different time in the year than the classes she was taking when she met her husband. However, as it stands, she fails to notice the incoherence of her position. She simply reports that the day of the week of the cooking classes was Monday (because that was the day of the week of the wine tasting classes). In this scenario, the wife's belief token would be rationally defective in the sense that it is the product of a poor cognitive performance. She doesn't have an occurrent belief that she met her husband after her classes on Monday because this type of belief is evidentially supported. Indeed, this is true even if, in fact, she has evidential support for a belief that those cooking classes took place on Monday. For instance, it might be that if she thought about it more, she would realize that these cooking classes took place on the same day of the week as her wine tasting classes had. Still, her occurrent belief that the cooking classes took place on Mondays is not held (even partly) for this reason.

Arguably, if the husband's cognitive processes extended out into the wife's head, the problem we are imagining for the wife would straightforwardly be his as well; the husband's token belief would also be rationally defective in roughly the same sense: he would have this occurrent belief that the day was Monday, but not because he has evidential support for a belief of this type. (Shortly, I will defend this inference, but it is useful to consider the implications of it first.) On the other hand, if the husband's cognitive processes do not extend out into the wife, then this occurrent belief need not be rationally defective in the same way. If the husband has accepted the word of the wife as something he himself remembers – if he is fabricating its origin – then there is certainly something wrong with his second-order belief concerning the origin of his belief that the day was Monday. But, it isn't obvious that there is anything wrong with the first-order belief that it was Monday itself. This belief is (presumably) reasonably held because he is reasonably taking his wife's words on their face (even if he thinks he's remembering). Alternatively, if the wife's utterance simply “jogged” the husband's memory – so that he was primed to genuinely remember the day of the week – then the wife's confusion presents no problem for the husband at all. Her confusions are not his.

The upshot is that it appears that we can adjudicate the claim that the husband's remembering is radically extended by considering the epistemic status of his occurrent belief that the wife's classes were on Monday. If his occurrent belief inherits the epistemic status of the wife's, then there's radically extended remembering. If it doesn't, then there isn't. I contend that, in the kind of scenario envisioned where one partner is confused, this very confusion doesn't transfer to the other. For instance, if the wife's confusion comes to light, it is unnatural to attribute this confusion to the husband. This is especially so if the husband insists that he (now) remembers that it really was Monday and can even report supporting memories. However, it is still unnatural even if the husband defers to the wife and withdraws his statement that the classes took place on Monday once the initial confusion is revealed. Indeed, it seems much more natural to think that the husband has fallen prey to his gullibility (and perhaps lack of self-knowledge) than it is to think that he has fallen prey to confused remembering in the very same way as his wife has. The husband may well have exhibited an epistemic flaw, but it is a distinct one. (Certainly, the husband hasn't confused the day of the cooking classes with the day of the wine tasting classes even if the husband is, in fact, confused!) This epistemic assessment undermines the claim that the husband's remembering is radically extended into the wife – at least in a bad sort of case where there's a problem located in the wife's head. However, if there's no REC in the bad case, then there's no motivation for postulating REC in the good case where there's no problem at all. So, there's no apparent reason to think that there's REC in the original Sutton et al. (Reference Sutton, Harris, Keil and Barnier2010) case quoted in the last section.

This line of reasoning against REC can be summarized succinctly as an instance of modus tollens. From the premise (1) that if the husband's process of remembering radically extends into the wife's head, he would be confusedly remembering if she were and the premise (2) that he would not be confusedly remembering if she were, one infers that the husband's process of remembering is not radically extended into the wife's head. The obvious way to resist this line of reasoning is to deny the premise (2). When the wife confusedly remembers, it merely seems as if the husband isn't also confusedly remembering in roughly the same way. It might seem otherwise because the husband's belief isn't, in fact, rationally defective even though he is confusedly remembering because his remembering process is intertwined with hers. It might be argued, for instance, that the husband has two sources of evidential support: his remembering (which draws on resources located in the wife) and the wife's testimony (which doesn't). Moreover, he believes on the basis of evidential support from both sources at once. So, even if his believing by remembering is defective, his simultaneous believing by accepting testimony is not. In this way, the husband's belief is in better shape than the wife's (even though his remembering process is deeply parasitic on hers!).

In my view, this resistance should itself be resisted. If the wife were feeling especially suspicious, she could take an alienated perspective on her own memory faculties. She could view them as she would an external witness.Footnote ³⁶ After all, these memory faculties largely operate in ways that aren't directly accessible from a first-person perspective just as external testifiers do. Thought of in this alienated way, the wife's memory faculties are “testifying” to her. Still, the wife has reason to think that they are generally reliable, and, on this basis, might accept the “testimony.” In other words, it might be argued that the wife too has two sources of evidential support: her remembering and the “testimony.” These sources are analogous to those of the husband. Moreover, she might believe on the basis of both sources at once – just as the husband does – so there should be no difference between these two beliefs. While this second source of evidential support might be a mitigating factor, it still seems that the wife's belief is defective. Ultimately, it is the product of confusion. If the husband's remembering process is radically extended into the wife, the same is arguably true of him.

A possible rejoinder is that testimony from the wife to the husband is genuine whereas there is only “testimony” – as the view from an alienated perspective – of the wife's memory faculties to her. It might be argued that this difference matters for the evaluation of their respective beliefs concerning the day of the cooking classes. However, this rejoinder is vexed for the advocate of REC. The difference between genuine testimony and “testimony” should only matter for epistemic evaluation if there's an important cognitive difference in the way these two inputs are received. However, if there is an important cognitive difference between the way that the husband treats the input from his wife and the way that the wife treats the input from her memory faculties, then there's a significant failure of “parity” – as discussed in Clark and Chalmers (Reference Clark and Chalmers1998).Footnote ³⁷ The husband is no longer relying on the wife in the way that a subject relies on her memory faculties, so there's no apparent motivation for supposing that the wife is functioning as – and therefore is – a part of his memory system.

7. CONCLUDING THOUGHTS

The discussion in §6 is meant to be suggestive rather than conclusive. Perhaps, further examination will reveal the Sutton et al. (Reference Sutton, Harris, Keil and Barnier2010) case, in fact, to be a case of REC. My interest in this paper is not in arguing for or against REC, but rather in demonstrating a kind of methodology for doing so, a methodology that is grounded in the relationship between epistemology and REC discussed in the earlier sections of the paper. Because epistemological theory is framed in terms of the cognitive (as indicated in §2), but unaffected by the prospects for REC (as discussed in §§2–3), epistemological theories can be used as an effective standard for diagnosing REC. §§5–6 are simply intended to illustrate how this diagnosis might go. Our pre-theoretical tendencies towards epistemically assessing subjects in particular ways – e.g. tendencies not to think that confusions transfer between partners – may reveal pre-theoretical commitments about where the cognition of these subjects is located. Further systematic epistemic assessment may confirm these pre-theoretical commitments. Or, it may not. Either way epistemology provides a starting place for examining whether REC is present. This is the intended moral of this paper.Footnote ³⁸

Nevertheless, if there's a more specific lesson to be learned from thinking about the Sutton et al. (Reference Sutton, Harris, Keil and Barnier2010) case, it is this: while the prospect of recruiting external resources for cognition might provide wonderful opportunities for more extensive cognitive success, it simultaneously provides further locations for cognitive failures. Anywhere a subject's cognition is located is a place where the subject's cognition could be defective. Consequently, before judging that one's cognition extends to some particular place, it is worth considering whether one could be confused, muddleheaded, irrational, etc. in virtue of what happens there. My own view is that it is not generally plausible that any of us could be confused, muddleheaded, irrational, etc. in virtue of what happens where other human beings are located.Footnote ³⁹