1.

In his well-known article ‘Philosophy as a Humanistic Discipline’, Bernard Williams is concerned with this question: ‘[W]hat models or ideals or analogies should we look to in thinking about the ways in which philosophy should be done?’ (Reference Williams2000: 477).

Williams's answer is that the models, ideals, and analogies we should look to are those in history, that is, the academic discipline of history, a discipline Williams assumes to be paradigmatically humanistic. This yields what I call the Williams thesis: that philosophy is like history, at least in important respects relevant to its nature.

This article is a novel exploration and defense of the Williams thesis. The key idea is to apply a central moral from what is sometimes called the analytic philosophy of history of the 1960s (Danto Reference Danto, Ankersmit and Kellner1995; Dray Reference Dray1964; Roth Reference Roth2016; Little Reference Little2017) to the philosophy of philosophy of today. In the article that largely initiated that style of philosophy of history, Carl Hempel argued (roughly) that there is causal explanation in history only if there are laws in history (see Hempel Reference Hempel1942). However, in light of the apparent fact that there are no laws in history, or at any rate that historians are not providing such laws, it seems to follow that there are no explanations either, a result in conflict with the practice of actual historians. The response—and this is the moral I have in mind—is to separate the question of whether there is causal explanation from the question of whether there are laws. Historical explanation is indeed a species of causal explanation, but it need not take the form, as Hempel suggested it must, of an argument one of whose premises is a statement of a universal law.

Causal explanations are not what philosophers are typically aiming to provide. Nevertheless, at least in many cases, they are concerned with something analogous, namely constitutive explanations, explanations that concern how items of philosophical interest are—to put it deliberately vaguely—grounded or realized or necessitated (see, for example, Rosen Reference Rosen, Hale and Hoffman2010: 111). And when they do attempt to provide such explanations, it is often assumed that what they are after is, if not a Hempelian argument that includes a law, then again something analogous, something that, following David Lewis (Reference Lewis1986), I call a unit of explanation. As I show below, to assume that answers to philosophical questions involve a unit of explanation is in this context to assume they must have quite distinctive properties: they must be or include a bridge law or a definition or a set of a priori principles, say, or perhaps be or include an argument of a particularly unchallengeable kind. For Paul Horwich, for example, answers to philosophical questions include, at least as the discipline is traditionally understood, a ‘nonobvious body of a priori principles—one that offers a complete, systematic, precise, and basic account’ (Reference Horwich2012: 21) of some item of philosophical interest. And Peter van Inwagen, in explaining what he thinks would settle disputes over the truth of a philosophical thesis, says that this will occur only when ‘there is a knock-down argument either for that thesis or for its denial’ (2004: 337)

Assumptions of this sort lead to a problem in the philosophy of philosophy very similar to the old one in the philosophy of history, a problem often expressed in terms of whether philosophy makes progress of the sort one sees in the sciences (see, for example, Stoljar Reference Stoljar2017; Lycan Reference Lycan2019; Williamson Reference Williamson2007, Reference Williamson2018). For it seems a descriptive fact that philosophy rarely if ever produces items such as Horwich's a priori principles or van Inwagen's knockdown arguments. If such things are required for philosophy to answer the questions it sets for itself, however, there are very few explanations here either. Indeed, this is a conclusion that both Horwich and van Inwagen either embrace or come close to embracing: both take extremely seriously a pessimistic view about the prospects of progress in philosophy.

I suggest that the Williams thesis solves this problem, allowing us to confront it in a way that the analogous issue in the philosophy of history was dealt with earlier. In particular, we should view philosophy is a sort of synchronic history. On the one hand, philosophy, like history, often consists in the provision of information about what I call dependency structures—not diachronic and causal dependency structures as in history, but synchronic and constitutive dependency structures. On the other hand, explanations in philosophy, like explanations in history, do not usually consist in the provision of laws or any analogous unit of explanation. I also suggest, more generally, that seeing philosophy as in this way analogous to history is a fruitful one for thinking about several aspects of its nature.

In what follows, I begin (sections 2–3) by summarizing the approach to explanation that forms the background to my discussion, which is a modified version of one proposed by Lewis in ‘Causal Explanation.’ I then turn to my main topic, the Williams thesis: the case for it in sections 4–7, objections against it in sections 8–11, and how my defense of it is different from Williams's own (section 12). In section 13, I widen the scope by briefly indicating some further consequences for the proposed analogy between history and philosophy: for providing some context for contemporary metaphilosophy; for understanding the interaction of philosophy and other sciences; and for explaining, or explaining away, the belief that philosophy is a field that requires specific ability or brilliance.

2.

Lewis's view in ‘Causal Explanation,’ as I understand it, is composed of four theses, a point I have developed elsewhere (Stoljar Reference Stoljar2019; Reference Stoljar2017: chs. 5–6).

The first thesis is that the world consists in or contains causal histories, vast systems of events standing in various causal relations. This thesis is intended to be an abstract metaphysical thesis, not tied to any particular theory of causation or of causal relata. Lewis himself of course had views on both topics, but his account of causal explanation is intended to be, and is, separable from them.

The second thesis is that to explain a particular or token event is to provide some information about its position in a causal history; correlatively, to explain a type of event, and so provide a general sort of explanation, is to provide information about a type of causal history, a type that includes events of the relevant sort. As I understand him, Lewis intends this thesis to provide necessary and sufficient conditions on what explaining an event is. However, since the necessary part of this is controversial but not required for my purposes, I ignore it.

The third thesis is that to provide information about causal histories is a special case of providing information about anything. When one provides information about a train system (this is Lewis's example), one generally aims to maximize various virtues, such as truth, relevance, clarity, novelty, and reasonableness, and to minimize various vices, such as falsity and abstraction (Lewis Reference Lewis1986: 226–27). The same is true, he thinks, when providing information about causal histories.

The fourth thesis is that, beyond the fact that causal explanations provide information about causal histories and conform to the general canons of information provision, there is no special form or content they must have. Lewis puts this by saying that there is no ‘unit of explanation’. His main example of an approach that requires such a unit is Hempel's view, noted above, that explanations must take the form of an argument among whose premises is a statement of a universal law. Lewis accepts that causal explanations may on occasion be Hempelian; his point is that they need not be, and in particular need not be in order to be good explanations. But his rejection of the idea of a unit of explanation is not confined to this view in particular: any view that demands something beyond that required by the usual standards of information provision is for him objectionable in exactly the same way.Footnote ¹

3.

This approach to explanation does not apply to philosophy. Philosophers are certainly interested in the nature of causal explanation and whether it is possible to provide such explanations in certain cases, but, as I noted above, they are not typically interested in providing causal explanations themselves. Still, it is possible to generalize the Lewis approach so that it does apply.

One way to do this by taking advantage of a point that has been emphasized by a number of contemporary philosophers:Footnote ² that causal histories are one example of a more general type of structure, which, following a suggestion of Sebastian Lutz, I will call here a dependency structure. Causal histories are one sort of dependency structure because here we have a set of events at a certain time that depends on another set of events at a different time. Another kind of dependency structure is (what I call) a constitutive hierarchy, a vast system of facts (or true propositions) standing in various synchronic (or at least non-diachronic) relations of grounding or realization or necessitation. As an illustration, consider the thesis of physicalism. This thesis entails, whether or not it is true, that the world not only contains but consists in a constitutive hierarchy: according to it, every fact is grounded in, necessitated by or realized by some physical fact.

A version of Lewis's approach generalized to dependency structures may be formulated this way. The first thesis is that the world consists in or contains dependency structures, systems of causal relations among events or constitutive relations among facts. Once again, this as an abstract metaphysical thesis; as before, it is not tied to any theory of causation or causal relata, but nor is it tied to any theory of grounding, necessitation or realization, or of the terms of those relations.

The second thesis is that to explain a particular item (that is, a particular event or fact) in a dependency structure is to provide information about the position of that item in the structure; correlatively, to explain a type of item is to provide information about a type of dependency structure, a type that includes items of the relevant sort.

The third thesis is that providing information about dependency structures is a special case of providing information about anything. One might point out that, since the relations involved in constitutive hierarchies are sometimes necessary, while the relations involved in causal histories are contingent, it is hard to see what providing information about them will consist in, at least if providing information is thought of as a matter of distinguishing among possibilities. I do not go into this matter here, noting only that there must be some way of understanding the sense in which we can provide information about necessary structures. For some discussion of related issues, see Jonathan Schaffer (Reference Schaffer2016b: 71).

The fourth thesis is that there is no unit of explanation. We saw what this amounts to in the causal case: partly a rejection that explanation must be Hempelian, but more generally a rejection of any theory requiring explanations to have a form or content beyond that mandated by the usual standards of information provision. This latter point straightforwardly applies to the constitutive case. To provide a constitutive explanation it is necessary to provide information about a constitutive hierarchy, but it is not necessary to provide a definition or a bridge law or a knockdown argument. Once again, constitutive explanations may on occasion have these forms; but they need not, and need not in order to be good explanations.Footnote ³

4.

Turning now to the Williams thesis, the moral of the analytic philosophy of history I mentioned at the outset—that we should separate the notion of causal explanation from the notion of a law—may easily be expressed in the framework just examined.Footnote ⁴ From Lewis's point of view, it is perfectly possible that historians provide causal explanations even if they do not provide explanations that are Hempelian. Hence, the fact that their explanations are not Hempelian presents no philosophical puzzle, contrary to the assumption of the philosophers in the 1960s.

If we assume this in the case of history, it becomes possible to assume it also in the case of philosophy—so long as we generalize Lewis's approach in the way suggested. From this point of view, philosophers often provide constitutive explanations, even if they do not provide explanations that include bridge laws, definitions, knockdown arguments, and so on.

Above I formulated the Williams thesis as the claim that philosophy is like history in important respects relevant to its nature. Of course, there are many ways in which disciplines may be alike and unalike. Which ‘important respects’ are the relevant ones? The version of the thesis I am interested in says that two respects are important: (1) both historians and philosophers aim to provide information about dependency structures; (2) in neither case must the information in question involve a unit of explanation.

5.

To say that it is possible that philosophy is like history in these respects goes only so far. What reason is there for saying that it is like this, which is what the Williams thesis says?

A central consideration is that the Williams thesis allows us to solve the problem about progress in philosophy I outlined above. One way to develop this point is to see that arguments leading to pessimism about philosophical progress make precisely the mistake Lewis warned against, the unit of explanation mistake.

Look again at Horwich's suggestion that, as traditionally conceived, answers to philosophical questions will take the form of a body of a priori principles. Horwich, of course, does not think that philosophers have produced such principles. This is one of his main reasons for supposing that traditional philosophy ought to be rejected in favor of a Wittgensteinian alternative, on which philosophical problems are ‘pseudo-problems’ quite unlike problems in other subjects. In the light of the Williams thesis, however, we should be suspicious of this line of reasoning. For what are Horwich's a priori principles but a unit of explanation in Lewis's sense? Horwich is assuming that, in traditional philosophy, answers to philosophical questions must take a particular form; when he finds nothing of that form, he declares the enterprise a failure. What the Williams thesis suggests is that we should be skeptical of that assumption.

One might argue on Horwich's behalf that traditional philosophy is surely concerned with a priori generalizations, and that only another a priori generalization can explain an a priori generalization. If so, there is no escape from the idea that traditional philosophy must produce a priori principles, Lewis on explanation notwithstanding.

But it is unclear that only a priori generalizations can explain a priori generalizations. To borrow an example from Gideon Rosen (2009: 119) an explanation for the a priori generalization that all triangles have three angles might be that it is of the essence of a triangle to have three angles; this seems not a generalization so much as a claim about a property, namely, the property of being a triangle.

Moreover, even if a priori generalizations are explained by other such generalizations, it does not follow that, when you provide the explanation, you must provide information that is ‘complete, systematic, precise, and basic’, which is what Horwich requires (Reference Horwich2012: 21). As I noted above, Lewis himself points out that his view extends to the general case. To explain events of a general type, he says, rather than one particular event, is to provide information about a general type of causal history. Transposing this to the synchronic case, to explain facts of a general type, rather than one particular fact, is to provide information about a type of constitutive hierarchy. But again the information in question need not take Horwich's (or Hempel's) form.

6.

Once we guard against the unit of explanation idea, therefore, any force in Horwich's argument that philosophy makes no progress dissipates. The same thing is true when we turn to van Inwagen's well-known claim: ‘disagreement in philosophy is pervasive and irresoluble. There is almost no thesis in philosophy about which philosophers agree’ (2004: 334). This claim does not strictly entail a pessimistic position in philosophy; in principle, I might know the answer to a philosophical question even if I cannot persuade anybody else of its truth. Nevertheless, it is strongly suggestive of pessimism, and entails in context that the epistemological situation as regards philosophy is completely different from that which obtains in history, science, or mathematics.

But what is it that van Inwagen thinks would constitute agreement in philosophy and so would bring the unfortunate situation he describes to an end? On this matter, he is admirably straightforward; indeed, this is the context in which he makes the remark about knockdown arguments quoted above. Van Inwagen's full statement is as follows: ‘Philosophical agreement will come to pass when and only when, for each important philosophical thesis, there is a knock-down argument either for that thesis or for its denial’ (2004: 337). It is hard to see a more explicit commitment to the unit of explanation idea than this! Just as Hempel thought we need a special kind of argument to answer questions in history, so van Inwagen thinks we need a special kind of argument, a knockdown argument no less, to answer questions in philosophy, or at any rate to resolve disagreement over such questions. Since we are hardly ever in possession of such arguments, it would not be surprising, if this standard is accepted, to conclude that disagreements are irresolvable and that progress is impossible. On the other hand, if Lewis's type of view is correct, we should reject this standard, not just in philosophy but generally.Footnote ⁵

It might be replied that any talk of explanation in connection with van Inwagen is misplaced, since he himself says that philosophy is not about explanation. In a comparison of his own views with those of David Armstrong, for example, he says that, while Armstrong postulates universals to explain certain pieces of data, for him, the existence of such entities follows immediately from ordinary descriptions (see van Inwagen Reference Van Inwagen and Calemi2016).

However, while van Inwagen does not apply the word explanation to the products of his philosophical activities, this does not affect the issues we have been discussing. For one thing, the basic message of what Lewis is saying extends beyond explanation strictly speaking, which is a point I return to below (in section 11). For another, the difference with Armstrong that van Inwagen intends to bring out concerns inferences to the best explanation or abductive inferences: he says that Armstrong thinks they have a central role in philosophy, van Inwagen thinks they do not. But this is unsurprising given his commitment to knockdown arguments, which are not abductive. Finally, when van Inwagen elaborates (2014: 53) his own approach to universals, he says it has two parts: the first part concerns the theoretical or functional role that universals play, the second part concerns the nature of the things that play that role. An approach of this general kind might easily be described as concerning what realizes the property role—and that is one sort of explanatory approach in the sense that we (though not van Inwagen) are using the term.

One might also think it is too quick to see in van Inwagen's insistence on knockdown arguments a commitment to units of explanation. When introducing the latter idea above, I said that to be committed to it is to be committed to the idea that explanations have some special form or content beyond that required by the standard canons of information provision. It may be natural to interpret Hempel and Horwich this way: both think that explanations of the relevant sort involve laws or principles. But for van Inwagen what is important are not laws or principles but knockdown arguments. And these, one might think, involve no special form or content at all. If so, he does not make the mistake I am saying he does.

In part, this depends on what knockdown arguments are, which is itself a somewhat complicated matter (see, for example, Ballantyne Reference Ballantyne2014; Stoljar Reference Stoljar2017). As I understand them, however, knockdown arguments are those that have known premises that a priori entail their conclusions in a particularly obvious way. As such, it is reasonable to say that knockdown arguments are distinguished by their form or content. But even if that is not right, it is certainly true that the canons of information provision do not require you to produce arguments with these properties in particular, even if they sometimes require you to provide arguments, and that is enough to render an insistence on knockdown arguments objectionable from the Lewis-inspired perspective.

7.

So one consideration in favor of the Williams thesis is this: it allows us to identify the mistake in various prominent arguments for pessimism or something near it about philosophy. A closely related consideration is that it allows us to think more positively about the nature and plausibility of philosophical progress.

For suppose we think of history as an attempt to provide information about dependency structures. Then the issue of progress in history is a matter of whether we currently have more and better information (using Lewis's desiderata) about various diachronic dependency structures (or about relevant parts of the total diachronic dependency structure) than we had in the past. This is an empirical question, but once the question is posed this way it is difficult not to answer in the affirmative, and so endorse a reasonable optimism about history. Do we currently have more and better information about what happened on the Australian frontier in the nineteenth and early twentieth centuries than we had before? Of course the answer is yes (see, for example, Reynolds Reference Reynolds1999).

If that is true for history, however, the same is true for philosophy—if the Williams thesis is granted. For if philosophy too is construed as aiming to provide information about dependency structures, the issue of progress in philosophy is also a matter of whether we currently have more and better information (using Lewis's desiderata) about various dependency structures than we had in the past. As before, this is an empirical question, but again if we have the issue in this form, it is hard to give it anything but a positive answer.

For example, consider the varieties of proposal that go under the label of functionalism in philosophy of mind. If you look at those proposals in search of a reductive definition of mental states such as belief, desire, perception, and imagination, you are likely to be disappointed; you may even give voice to this disappointment by becoming a pessimist about philosophy (see, for example, Lycan Reference Lycan2019). But suppose, instead, you look at this literature and ask whether it provides better information about these states than we had before. Do we know more about the cognitive, intentional, rational, and phenomenal roles of (for example) perception than we did before? If you consider the question that way, it is again difficult to come away with anything except a yes answer. It is in this sense that the Williams thesis has a transformative effect on the issue of progress in philosophy.

One might reply that there are several potential aspects of historical explanations that have no counterparts in the philosophy case (a recent discussion is Currie Reference Currie2019). For example, it is sometimes suggested that historians, as opposed to philosophers, are concerned with particular events rather than types of events. Similarly, it is sometimes suggested that historians describe the past using representational resources that only become available in the present. However, whether or not these claims constitute genuine differences between history and philosophy—I will take no stand on these issues here—they do not undermine the analogy I am interested in. For it remains the case that progress in both disciplines is a matter of having better information now than we had in the past.

One might also suspect that what I have said makes progress in philosophy too easily attainable. Philosophers who are attracted to pessimism about philosophical progress often make an exception in the case of negative information, that is, about what is not true in philosophy (see, for example, Lycan Reference Lycan2019: 85; van Inwagen Reference Van Inwagen2004: 334–35). Such philosophers might agree that we are currently in possession of more and better information—so long as it is negative. Hence, even pessimists might agree that there is progress in the sense I have described.

One thing to say here is that, while the information we currently have about philosophical topics may include negative information, it is not limited to that. Claims about the functional role of perception, for example, are not negative in any obvious way—on the contrary they typically characterize perception positively as a state that prompts and makes rational various beliefs about a restricted class of properties and relations apparently instantiated in the local environment of the perceiver. But the deeper thing to say is that it is unclear that pessimists are in a position to allow this exception in the first place (see Stoljar Reference Stoljar2017). When one makes a negative point in philosophy, it is almost always the case that a positive point comes along in its wake. Indeed, we need not look very far to find an example of this phenomenon. Lewis's negative point that causal explanation need not be Hempelian brings with it a positive point that causal explanations may take any number of forms so long as they conform to the general canons of information provision. If that is typical, the ‘concession’ that there is progress on negative information is no real concession at all.

Finally, one might worry that the points I am making are too abstract. The examples mentioned above are very brief. Should I not look in detail at various case studies in both philosophy and history in order to show that there is progress in either discipline? Of course I acknowledge the need for detail in general, but I am not aiming to provide it here. The main point instead is theoretical, viz., that once we remove any Hempel-like element from contemporary philosophy of philosophy just as similar elements were removed from philosophy of history in the 1960s, it is rational to approach the whole issue of progress in an optimistic spirit.

8.

I have so far been agreeing with Williams: philosophy is like history in important respects relevant to its nature. And I have pointed out that this idea diffuses the problem about progress in philosophy noted at the outset. But at this point, several more general objections emerge, objections it is natural to divide into two categories. The first category concerns metaphilosophical views that offhand are in conflict with our assumptions so far; the second category asks whether the Lewis-inspired approach to explanation distorts the underlying issues.

One objection in the first category is suggested by Wilfrid Sellars's infamous claim that ‘the aim of philosophy . . . is to understand how things in the broadest possible sense of the term hang together in the broadest possible sense of the term’ (Reference Sellars and Colodny1962: 35. While interpreting a passage like this is always difficult, one may well read into it the suggestion that, whatever is true as regards history, there is no escaping the unit of explanation idea in philosophy, since answers to philosophical questions will consist in the provision of total information, and total information is a unit if anything is. As Sellars himself puts it later, ‘it is “the eye on the whole” which distinguishes the philosophical enterprise’ (Reference Sellars and Colodny1962: 37).

But understood this way Sellars presents no challenge to us. For one thing, if he literally means the (that is, the unique) aim of philosophy, he is mistaken. It is true that sometimes philosophers have their eye on the whole; sometimes they are interested in setting out an entire world-view or Weltanschauung. Those who are interested in the truth of physicalism for example are focused precisely on this sort of question. But in many other cases, the questions at issue in philosophical discussion do not have this totalizing quality. One might perhaps protect Sellars against this criticism by interpreting his claim so that it concerns only things relevant to some particular question, but now it is hard to see how what he says bears on philosophy in particular as opposed to any other sort of inquiry.

Moreover, even if we focus on cases in which the aim of the enterprise is to articulate a world-view, it is crucial to distinguish total information from information about a totality. When philosophers articulate a world-view, they are interested in providing information about a totality, namely, the entire system of existence. But it is implausible that they are interested in providing the total information about that system; indeed, in view of the obvious limitations of human beings, such an aim might well seem fanciful. On the other hand, given that it is information about a totality we are interested in, rather than total information, we lose the sense in which attempts to provide a world-view require anything like a unit of explanation. Providing information about the entire system of existence is in principle no different from providing information about a train system or indeed anything at all.

9.

A second metaphilosophical idea that might seem in tension with my position derives from Frank Jackson's (Reference Jackson1998) defense of conceptual analysis. Jackson argues, to put it roughly but adequately for present purposes, that, if physicalism is true, then, for any fact p, there is a fact q, formulated in the language of physics, such that the conditional ‘if q then p’ is a priori. A generalization of this proposal says that, in any constitutive hierarchy of the sort considered above, and for any constituent fact p of that hierarchy, there is a fact q, formulated in the language that describes the basic facts of the hierarchy, such that the conditional ‘if q then p’ is a priori.

Now, one may well see in this proposal a sort of sophisticated Hempelianism applied to philosophy. For suppose that physicalism is true, and we are interested in why (that is, in virtue of what) some fact F₁ is the case. Jackson seems to be saying that we must answer that question by providing a fact F₂ formulated in the language of physics, and an a priori entailment claim of the form ‘if F₂ then F₁’. And is that not tantamount to saying that explanations are arguments just as Hempel suggested?

One might respond by challenging Jackson's position; it is well known that he relies on premises in epistemology and philosophy of language that at least some philosophers reject. But a better response is that the objection rests on a misunderstanding. For Jackson's claim is existential: he is saying that if F₁ is the case, there is a conditional of the sort he is interested in; likewise, he is saying there is an argument that contains that conditional as a premise. He does not say we are required to mention either the conditional or the argument in the course of answering the question about F₁, though of course it is permissible to do so.

This is clear, for example, in Jackson's response to what he calls ‘Stich's challenge of actual cases’ (1998: 62). Stephen Stich had challenged Jackson to provide a conceptual analysis of grooming behavior, which he (Stich) assumes entails that Jackson produce a fact formulated in physical language that a priori entails that such and such is an example of grooming behavior. Jackson responds that the demand misconstrues what he intended; he intended the claim that there is an a priori entailment, not that anyone produce one. More generally, it is quite consistent with what Jackson says that the Lewis-inspired approach to explanation is correct.

The point here may be strengthened by noting that an analogous issue comes up in the context of Lewis's original theory of causal explanation. At one stage, Lewis compares his view with that suggested by Peter Railton (Reference Railton1978, Reference Railton1981). On Railton's view, there is (what he calls) an ideal explanatory text, consisting—to put it roughly—of a vast system of interlocking Hempelian arguments. To that extent, Railton's view is like Hempel's; indeed, it is presented as an updated and elaborated version of Hempel's position. But in terms of the idea of explanation as the provision of information, it is like Lewis's as indeed he (Lewis) points out. Jackson's metaphilosophical position may be fruitfully seen as akin to Railton's position on explanation. If physicalism is true, there is an ideal explanatory text—that is, a system that includes a huge raft of a priori conditionals—but explanation may nevertheless be understood in Lewis's terms.

I note also a connection between Jackson's position and that of Horwich discussed earlier. As we saw, Jackson should be interpreted as claiming that there is a raft of a priori conditionals; likewise, Horwich might be interpreted as saying that, according to traditional philosophy, there is a body of a priori principles. If so, we arrive again at a reason to resist his suggestion that there is no progress in traditional philosophy. For even if it is true that we have not managed to articulate these a priori principles in the precise way he requires, it remains the case that we have better information about them than we did in the past. And this is what progress requires according to the Williams thesis.Footnote ⁶

10.

I said before that my position provokes two categories of objection. The first, which I just discussed, concerns potential alternatives to the Lewis-inspired account of explanation we have presupposed. The second category is concerned with the account itself and, in particular, with whether it distorts matters when it is applied to philosophy.

One objection in this category is that the Lewis-inspired account assumes that there are both facts and relations of constitution that obtain among those facts. But is it not an open question in various parts of philosophy whether such things exist? The illusionist in philosophy of mind, for example, denies outright that there are phenomenal facts, i.e., facts about phenomenal consciousness (see, for example, the contributions in Frankish Reference Frankish2017). If so, there are no dependency relations among phenomenal facts, or between such facts and other sorts of facts. But then it is impossible that philosophy of mind or consciousness could be understood as attempting to provide information about such relations.

One reply to this objection is that it forgets that information about dependency structures, like information about anything, can often take a negative form. If illusionism is true, the whole truth about dependency structures involving phenomenal facts may be provided quite easily, viz., there are none. A second, compatible, reply is that, while there may be no dependency structures involving phenomenal facts, there are nevertheless related structures—structures involving facts about its seeming to us in introspection that such facts obtain, for example. It is consistent with what illusionists say that philosophy of mind is concerned with structures of this related type; indeed, this is precisely their suggestion.

One might try to put the line of thought here more generally by imagining a tough-minded philosopher who denies the existence of facts outright. If there are no facts, philosophy cannot be a matter of providing information about them. But the problem with this is that any such tough-minded position must eventually accommodate the ‘fact’ (as we might tendentiously put it) that many (many!) parts of ordinary life and science apparently contradict it. I am doubtful there is a successful story to tell about how to avoid this contradiction, but if there is, there is no reason why it will not apply in the case of philosophy as well.

11.

A second objection to the account of explanation I have presupposed points out that it is surely an exaggeration to say that philosophy is concerned with dependency structures. No doubt we are sometimes concerned in philosophy with structures of that sort, but just as often we are interested in what something is—for example, what consciousness or morality or knowledge is—and this seems to have little to do with dependency.

However, while this point is correct, it does not alter the main thing I have been trying to say. Lewis presents his proposal as a view about ‘why’ questions. Such questions, he thinks, are requests for information about causal histories, or, on our generalization, about dependency structures. But, as he goes on to say, ‘all questions are requests for information of some or other sort’ (1986: 221). So, in particular, if we ask what something is—what consciousness or morality or knowledge is, for example—we are requesting certain sorts of information, not about dependency structures, but about consciousness, morality and knowledge. On the other hand, if that is so, versions of the third and fourth theses mentioned above when setting out the Lewis-inspired view apply: providing information as an answer to a ‘what’ question is likewise a special case of providing information in general, and here too we should reject the idea of a unit of explanation, or to put it now more generally, the idea of a unit of information. In short, if we shift from ‘why’ questions to ‘what’ questions, or indeed to other questions, all the principles of information provision apply as before, and hence as before we should view the idea of a unit with suspicion.

It might be thought that, if the account I have in mind is generalized in this way, it no longer applies to philosophy and history in particular: all I seem to have said is that both philosophy and history are in the business of providing information, and that does not distinguish them from anything else. One response emphasizes that it remains the case that some fields have a Hempel-like character, or at least are plausibly interpreted in that way; certainly there is nothing in the Lewis-inspired picture to reject this. But the more important thing to say is that it does not matter to my overall argument if philosophy is like other fields as well as history. Part of what is lying behind the pessimistic ideas about progress we have been considering is that philosophy is a peculiar discipline, to use a phrase Williams (Reference Williams2000: 478) himself adopts from Wittgenstein; that is, a discipline to which normal standards of progress do not apply. The Williams thesis as I have defended it undermines this. For underneath the moral of the analytic philosophy of history that we have been focusing on—decoupling laws and explanation in history—is a more general injunction: do not take the quite exceptional cases of physics or mathematics and treat them as the norm. We should not take that attitude to history, but if the Williams thesis is right, we should not take it to philosophy either—and this gives us a very profound sense in which philosophy is not peculiar in the relevant way at all.

12.

The time has come to look back at Williams's essay and ask how my position differs from his. For while my discussion started with Williams and is inspired by it, the divergences are considerable.

One difference is that Williams does not argue for the thesis in the way I have done. Williams suggests that philosophy is like history, and is unlike science, partly on the ground that it lacks what he calls a ‘vindicatory’ (2000: 487) history, which is, in effect, a history of progress. For us, the analogy to history is used in quite the opposite way, as part of a defense of an optimistic view about philosophical progress. Williams also emphasizes with regard to his thesis that he ‘shall not try to deduce it from the nature of philosophy as compared with other disciplines, or indeed deduce it from anything else’ (2000: 479). Related to this, he warns against what he calls essentialism, the idea that there is a single template that fits all philosophical problems. While I, too, have not tried to deduce anything, I have argued on the basis of various views about the nature of philosophy and history. I do not think I am committed to essentialism, since it is sufficient for my purposes that quite a bit of philosophy conforms to the Williams thesis, not that all does. Still, this is unlike Williams's own procedure.

Not only is the argumentation different, the conclusion is too. Williams's view is that philosophy is like history because it is history, at least in part. He thinks that ‘the content of our concepts is a contingent historical phenomenon’ (2000: 489); in consequence, philosophy must explain why we have the particular concepts we do, and that history is required to explain this. Hence, true answers to philosophical questions will themselves be partly historical.

But the proposal here is that philosophy is synchronic history not that it is diachronic history (that is, history as such), not even partly. I do not deny of course that what concepts we have, what questions we pursue and so forth, is a contingent historical phenomenon, and is well worth study. But that is true in any field, not simply in philosophy. So, unless we are out to portray every field as a humanistic discipline, it is hard to see that there is any particular consequence for philosophy here.

It is worth noting lastly that Williams's specific claims about philosophy and history are not especially plausible. After all, it is a familiar point that philosophy is not typically about why we have the particular concepts we do, but is rather about whether we are free, or whether anything is moral, what the relation is between the mind and the body, and so forth. And even if philosophers raise questions of the kind Williams is interested in, it is not clear that history, as opposed to (for example) psychology, will provide the answers.

13.

My suggestion has been that the Williams thesis is true, though not in the way he himself intended. What implications does the thesis have if it is true? One consequence, as we have seen, is for debates about progress. But there are several further consequences I would like briefly to enumerate, even if it is too late in the paper to give them the attention they deserve.

The first has to do with the larger context of contemporary metaphilosophy. At the beginning of a recent summary of work in the philosophy of history, Paul A. Roth writes, ‘Since the late nineteenth century, philosophical expositions of historical explanation follow a predictable format, one that invariably begins with an observation that such explanations fail to conform to philosophical models of (natural) scientific explanation. These standard accounts typically diverge either in claiming for historical explanations a status of a “science-in-the-waiting,” or as already in some form possessing those virtues typically associated with scientific ones, or by emphasizing the futile and fallen state of debate’ (Reference Roth2016: 1).

It is remarkable how much of this applies mutatis mutandis to philosophy itself, which suggests that the analogy between philosophy and history is deeper than the issue of progress that has been our focus. In particular, questions about realism, reductionism, institutionalization, the relation of public and professional approaches, and the sense in which both disciplines are ‘not quite science,’ could fruitfully be discussed in tandem. I have my own views about how these analogies will play out, but I will not go into them here. The point instead is to stress the potential profit of thinking about philosophy of philosophy in the light of philosophy of history.

The second consequence, which is a natural development of the first, concerns the relation between philosophy and other sciences. One of the more interesting developments in recent philosophy of history is an emphasis on the relation between history as a humanistic discipline, on the one hand, and the historical sciences, such as evolutionary biology, cognitive archaeology, and so on, on the other (see, for example, Currie Reference Currie2019; Currie and Sterelny Reference Currie and Sterelny2017; Currie and Walsh Reference Currie and Walsh2019). No doubt there are many differences here. Contemporary historians working on questions about what happened on the Australian frontier in the nineteenth and early twentieth centuries are in a completely different political situation from evolutionary biologists working on the origin of birds. And yet, if the Lewis-inspired picture is correct, there is a strong sense in which they are engaged in the same sort of inquiry: both are providing information about causal histories by the usual standards of information provision. In that sense, the difference between ‘humanistic’ and ‘scientific’ historical inquiry is much less marked than you might have thought.

A similar point can be made in the synchronic case, if the Williams thesis is true. Philosophy is in part in the business of providing constitutive explanations, but it is clearly not the only effort to provide such explanations—chemistry, psychology, linguistics and other sciences are profitably seen in this light. Once again, therefore, it is natural to think that there is no major difference between, and much scope for cooperation between, philosophy and other sciences that are attempting in different ways to fill in the details about constitutive hierarchies.

The final consequence has to do with something that may at first sight seem unrelated: the well-documented outlier status philosophy has with respect to other disciplines when it comes to the under-representation of women and other minorities. There are several contemporary attempts to explain this, but one hypothesis that is widely thought to provide at least part of the answer is that philosophy is a field that self-conceives as involving a certain sort of elusive unteachable brilliance and that in general ‘women tend to be under-represented in fields believed to require innate intellectual talent for success’ (Meyer, Cimpian, and Leslie Reference Meyer, Cimpian and Leslie2015: 11; see also Antony Reference Antony2012; Dougherty, Baron, and Miller Reference Dougherty, Baron and Miller2015).

While this hypothesis is plausible, it raises various further metaphilosophical concerns: Why does philosophy self-conceive in this way? What exactly is involved in that self-conception? Again, there are several possible avenues to pursue, but one answer is suggested by what we have said. For suppose that we are in the grip of the unit of explanation idea in the case of philosophy. Then it would be natural to think of solving philosophical problems as requiring an unteachable, quasi-sensory capacity, namely, a capacity to detect that special piece of information that answers philosophical questions—a sensus philosophicus as we might call it, echoing the homo philosophicus discussed in Cassam (Reference Cassam2014: 1) and the sensus divinitatis discussed in Plantinga (Reference Plantinga2011: 181).

On the other hand, suppose we are wrong to be in the grip of the unit of explanation idea, just as the analytic philosophers of history were wrong to be in the grip of the counterpart idea in the 1960s; it is worth noting that history is not an outlier as regards under-representation to anything like the extent that philosophy is. Then it would be natural to think of solving philosophical problems as requiring, not some special quasi-sensory capacity, but only the usual techniques of rational inquiry applied to the topics of philosophy. Since such things can be taught and encouraged (and indeed discouraged), we have the intellectual foundation for an intervention in our discipline of the sort that Meredith Meyer, Andrei Cimpian, and Sarah-Jane Leslie (Reference Meyer, Cimpian and Leslie2015) recommend.

14.

Williams's question was what models, ideals, or analogies should we look to in thinking about philosophy. His answer was intended, I think, to be deliberately nonstandard and iconoclastic, at least in analytic philosophy. Analytic philosophers like to conceive of their problems on the model of logic or natural science, but rarely emphasize the analogy to history. While my defense of the Williams thesis has been different from Williams's own, I like to think a nonstandard and iconoclastic element has remained. Even if it has not, I hope at least to have established the plausibility of the Williams thesis, to have demonstrated how it affects the problem of progress in philosophy, and to have indicated a number of its further ramifications—for the relation of philosophy of history and philosophy of philosophy, for the interaction of philosophy to other disciplines, and, lastly, for the ongoing effort to improve its culture.