The crucial point here is that Ramsey sentences don’t define concepts like belief, but eliminate them. They give us a way of saying what our everyday theories say without using the relevant concept (there is a kind of state that is caused by perceptions …) If we want definitions, then we need “Carnap sentences”, not Ramsey sentences (Lewis 1970). The Carnap sentence corresponding to “∃!Φ(T(Φ))∃!Φ(T(Φ)T(F)_If_ there is a kind of state that is caused by perceptions …, _then_ it’s belief.) Carnap sentences can plausibly be viewed as akin to stipulations that fix the reference of the relevant concepts, and to that extent as analytic claims that can be known _a priori_. But this certainly does not mean that Ramsey sentences, which make substantial claims about the actual world, are also knowable via _a priori_ analysis. (Note how you can accept a conditional Carnap sentence even if you reject the corresponding unconditional Ramsey sentence. You can grasp the folk concept of belief even if you reject the substantial folk theory of belief.)Can’t defenders of the Canberra programme argue that it is the analytic Carnap sentences that are crucial in setting philosophical agendas, not the synthetic Ramsey sentences? But this seems wrong. We will want to know about the fundamental nature of belief if we suppose that there is a kind of state that is characteristically caused by perceptions, and so on. That is certainly a good motivation for figuring out whether and how the fundamental components of reality might constitute this state. But the mere fact that everyday thought contains a concept of such a state in itself provides no motivation for further investigation. (In effect, the function of a Carnap sentence is to provide a shorthand for talking about the putative state posited by the corresponding Ramsey sentence. It is hard to see how any important philosophical issues could hang on the availability of such a shorthand.)To emphasize the point, consider the everyday concept of a soul, understood as something that is present in conscious beings and survives death. This concept of a soul can be captured by the analytic Carnap sentence: “If certain entities inhabit conscious beings and survive death, then they are souls”. Accordingly, this Carnap sentence will be agreed by everybody who has the concept of soul, whether or not they believe in souls. Yet this Carnap sentence will not _per se_ raise any interesting metaphysical questions for those who deny the existence of souls. These deniers won’t start wondering how the fundamental constituents of reality realize souls—after all, they don’t believe in souls. It is only those who accept the corresponding Ramsey sentence (“There are parts of conscious beings that survive death”) who will see a metaphysical issue here. Moreover, the Ramsey sentence will pose this metaphysical issue whether or not it is accompanied by some analytic Carnap sentence to provide some shorthand alternative terminology. In short, the methodological naturalist can insist that anybody interested in “serious metaphysics” should start by articulating the substantial existential commitments of our folk theories, as articulated in their synthetic Ramsey sentences. Any further analytic conceptual commitments add nothing of philosophical significance.The point generalizes beyond the contrast between Ramsey and Carnap sentences. On reflection, it is hard to see why any purely definitional analytic truths should matter to philosophy. Synthetic everyday truisms can certainly be philosophically significant, and so their articulation and evaluation can play an important philosophical role. But there is no obvious motive for philosophy to concern itself with definitions that carry no implications about the contents of reality. ### 2.3 Synthetic _A Priori_ Intuitions?It is worth noting that not all philosophers who advertise themselves as engaging in “conceptual analysis” are committed to the idea that this involves analytic a priori knowledge. In many cases philosophers who describe themselves in this way go on to explain that in their view “conceptual analysis” is a matter of articulating synthetic claims and assessing them against a posteriori evidence. A particularly clear version of this picture of conceptual analysis is offered by Robert Brandom (2001). Similarly, recent advocates of “conceptual engineering” are explicit that in their view “concepts” embody substantial commitments which are open to criticism on a posteriori grounds (Cappelen 2018, Cappelen, Plunket and Burgess 2019). In the end, it is not clear what differentiates “conceptual analysis” in this sense from the a posteriori assessment of theories. In any case, this species of conceptual analysis seems perfectly consistent with methodological naturalism. (See also Goldman 2007.)Other philosophers are also explicit that “conceptual analysis” issues in synthetic claims, but simultaneously regard it as a source of _a priori_ knowledge (e.g. Jenkins 2008, 2012). This combination of views is less straightforward. In particular, it seems open to the traditional query: how is such synthetic _a priori_ knowledge possible? If some claim is not guaranteed by the structure of our concepts, but answers to the nature of the world, then how is it possible to know it without _a posteriori_ evidence?Yet other philosophers distance themselves from talk of conceptual analysis, but even so feel that philosophical reflection is a source of synthetic _a priori_ intuition (Sosa 1998, 2007). They too would seem to face the traditional query of how synthetic _a priori_ knowledge is possible.In this context, Timothy Williamson has recently argued that the traditional distinction between _a priori_ and _a posteriori_ knowledge is less than clear-cut, and in particular that it breaks down in connection with the intuitions on which philosophers rely. In Williamson’s view, there is a distinctive philosophical method in which intuitive judgements play a central role, but there is no warrant for classifying the relevant intuitions as _a priori_ rather than _a posteriori_ (Williamson 2013).However, it is arguable that this does not so much address as by-pass the underlying question. Perhaps philosophical intuitions are not best classified as clearly “ _a priori_ ”. But, if philosophy’s distinctive methodology relies on synthetic intuitions, this still seems to call for some explanation of their reliability.Doubts about the reliability of philosophical intuitions have been amplified over the past few years by the findings of “experimental philosophy”. Empirical studies have indicated that many central philosophical intuitions are by no means universal, but rather peculiar to certain cultures, social classes and genders (Knobe and Nichols 2008, 2017). This variability of intuitions is in obvious tension with their reliability. If different people have opposed philosophical intuitions, then it cannot be that intuitions of this kind are always true.