Eliminative materialists assert that there are no beliefs. This claim has long baffled philosophers. How could it be true? For if there indeed no beliefs, then nothing, the worry goes, could be true, as there would be nothing that could represent reality, correctly or wrongly. In this paper I wish to dissolve this worry by showing that it is possible to read eliminative materialism in a different manner than has been traditionally conceived: not as a thesis about lack of fit between the folk psychological discourse and the mind-independent reality, but rather as a pragmatic, metalinguistic proposal that we should discontinue using belief talk. In this way, as I hope, at least some of the arguments which were to show that eliminativism is incoherent might be refuted.

1 What Is Eliminative Materialism?

As classically defined by Churchland (1981), eliminativism concerning propositional attitudes claims that folk psychology is, strictly speaking, false, and that the entities (beliefs, desires and so on) posited by this theory do not exist. Churchland focuses on propositional attitudes, rather than qualia; for him, it is the idiom of beliefs, desires and the like that constitute the core of folk psychology, and this convention will be followed in this paper. What is important for Churchland is that folk psychology should be interpreted as a theory: this way, it is possible to assess folk psychology according to the standards that all respectable theories should meet. Folk psychology, according to him, fails this assessment miserably: it does not progress; it is unable to explain many vital phenomena and is hard to integrate with other well-established scientific theories.

These considerations led Churchland to the conclusion that folk psychology is false. This, according to Churchland, had dire ontological consequences, as beliefs and desires which are to be understood as theoretical posits of this false theory should and will be eliminated from our ontology. In this respect, Churchland’s reasoning is characteristic of his broadly Quinean approach to ontological issues: only those entities which are necessary postulates of our best theories are worthy of being classified as existing.

Churchland’s version of eliminative materialism is not the only one, and his way of justifying the conception in question is not obligatory. Recently, there has been much interest in so-called scientific eliminativism (see e.g., Machery 2009; Corns 2016). A scientific eliminativist questions the reasoning of the Churchland style “traditional eliminativist”, noting that such an argument relies on a descriptivist account of reference for psychological terms. Instead, a self-described scientific eliminativist claims that we should adopt a causal account of reference for psychological terms. However, even with this new conception, the scientific eliminativist arrives at negative conclusions concerning certain psychological terms (Machery, the concept of “concept”; Corns, the concept of “pain”). According to the new eliminativist, these concepts are defective, as their putative referents do not form a natural kind.

“Scientific eliminativism” is quite problematic to interpret from an ontological point of view, as it is not entirely clear whether the proponents of this view agree that, for example, concepts or pains do not exist. Corns especially (2016) has serious reservations about drawing the conclusion that pain does not exist from the premise that pain does not constitute a natural kind. The ontological aspect of scientific eliminativism would probably require a separate study and, in the present paper, it will just be noted that it is quite possible to justify eliminative materialism on different epistemological and semantic grounds than those assumed by Churchland. Consequently, the arguments presented in what follows will be neutral when it comes to questions concerning the proper reference theory for psychological terms and the proper way of arguing for eliminative materialism. Instead, the focus will be on the metametaphysical status of the core ontological claim of eliminativism about propositional attitudes.

For the purposes of the present paper, the defining feature of eliminative materialism is taken to be the negative ontological thesis that there are no beliefs (this definition is adopted for example, by Daly 2013). The eliminative materialist is to be defined as a theorist who endorses the claim that will be subsequently referred to as the eliminativist thesis that “there are no beliefs”, regardless of the way this thesis is justified. Such definition includes not only the Churchlandian version of eliminativism, but also those proponents of scientific eliminativism who are ready to endorse the negative ontological claim about beliefs. For this reason, the definition of eliminativism provided is somewhat different from the classical Churchlandian one invoked at the beginning of this section: the present paper focuses mainly on the ontological claim of eliminativism, whilst remaining agnostic about the questions about the status of folk psychology and the like.

2 Eliminative Materialism and the Problem of Truth

It would be an understatement to say that eliminativism has not been greeted with enthusiasm among philosophers. For some, the problem stemmed from the steps of Churchland’s argumentation: some have denied that folk psychology should be thought of as a theory (see Ravenscroft 2016 for an overview). Others, like Stich and his collaborators (Stich 1996; Mallon et al. 2009), claimed that Churchland’s argument for eliminativism rests on unjustified acceptance of a descriptivist theory of reference. These are important problems which have generated significant controversies. However, these charges only pertain to the arguments which aim to justify eliminative materialism.

Some philosophers (see e.g., Boghossian 1990) claimed that the problem lay not only with justification but also with the very cogency of eliminativism for such philosophers, eliminativism is not simply wrong, it is incoherent. This charge seems to be far more serious: if arguments to the effect that eliminativism is incoherent are successful, then the position is an obvious non-starter, no matter how well we could justify it.

This kind of argument was labelled “cognitive suicide” by Baker (2004). The general idea of the argument is rather simple: it is impossible to believe that there are no beliefs as the very act of believing in eliminativism proves that there are indeed beliefs. In this way, the eliminative materialist position becomes self-undermining, as in order to accept it, one must accept its negation.

The “cognitive suicide” challenge has two closely intertwined aspects that nonetheless could be separated for the sake of the argument. The first point, strongly pressed by Baker (2004) herself, is what might be called “the problem of attitudes”. The general idea is simple: to accept any given theoretical position is to have a certain psychological attitude towards it (namely, the attitude of accepting or believing). Presumably, an eliminativist accepts her own position. Thus, she must possess an attitude of believing or accepting the eliminativist thesis. On the other hand, it seems impossible for her to do so, as she claims that there are no attitudes.

The other problem is the “problem of truth”. It seems intuitively plausible that if an eliminativist wants to be consistent, she should maintain that her position, namely eliminativism, is true. But, as might be claimed, there is a problem for an eliminativist to retain the notion of truth as, according to eliminativism, there are no beliefs which could serve as a vehicle for representational content and, consequently, as truth bearers. Therefore, it seems that on the eliminativist account, no statement whatsoever can really be true or false. Consequently, the central proposition of eliminative materialism, namely the proposition “there are no beliefs”, also cannot be true. However, a theory that denies that its central claim is true seems to be self-undermining, as it seems intuitively plausible that any theory should allow its central claim to be true.

The problem of attitudes has received some attention in the recent literature. Daly (2013), for example, claims that in order to avoid this problem, the eliminativist should advertise her position as a version of fictionalism. This would allow an eliminativist to combine the error theoretical approach to folk psychology (i.e., the claim that really there are no beliefs) on a metaphysical level, with the embrace of folk psychological discourse in the practical domain. In this way, an eliminativist would be still able to use belief talk and consequently say that she does believe in eliminativism, whilst still denying that that there are any beliefs. This solution is controversial (see Wallace 2016 for the overview of the debate on the consistency of mental fictionalism), but it shows that the problem of attitudes is not necessarily fatal for eliminativism.

An alternative solution has been put forward by Streumer (2017) in the general context of the debate about error theory. Streumer argues that in many areas of philosophy error, theories are impossible to believe in (for various reasons), but this fact does not prove in itself that they are false. Any theory might be true even if it is impossible for us to believe it, as these two issues are logically independent. Although Streumer does not endorse error theory with respect to folk psychology, his point is perfectly general: eliminativism about folk psychology might be true, even if it is impossible to believe in this conception.

Obviously, each of these strategies is controversial and would require much more careful consideration that can be provided here. In what follows, however, I am going to focus on the problem of truth, as without solving this issue, any further considerations about the position in question are futile. If, as critics say, it is impossible for eliminativism to be true, then, even if there were a feasible solution to the problem of attitudes, this would not change the overall negative assessment of eliminative materialism. A theory that claims of itself that it cannot be true seems to be an obvious non-starter for most philosophers. If we could somehow prove that eliminativism is not a candidate for being a true theory, then any further arguments about it being not a believable position or not sufficiently justified are secondary. Thus, the problem of truth seems to be a primary source of worry in the debate about the cogency of eliminativism.

In his unpublished paper Eliminativism without Tears, the importance of the problem of truth was recently highlighted by Alex Rosenberg, who claims that it is the most serious problem for eliminativism. According to Rosenberg, the problem of truth boils down to the fact that eliminativism undercuts the ability to accept the existence of intentional states, which could serve as vehicles of propositional content. If there are no entities with propositional content, the argument goes, then nothing, including eliminative materialism, could ever be true.

This conclusion might be resisted on the grounds that one can see abstract, timeless propositions as the primary bearers of truth and other semantic propertiesFootnote 1. In such a case, the problem of truth seems to be easily avoidable: the fact of eliminative materialism being true might—according to this proposal—consist solely of an abstract timeless proposition “there are no beliefs” being timelessly true. This fact could, in principle, be obtained independently of there being anyone capable of grasping or expressing the proposition in question.

Such a way of avoiding the problem of truth might, however, be treated as an attempt by an opponent of eliminativism to change the subject. What seems to be attacked by an anti-eliminativist is the possibility of coherently formulating the eliminativists’ position as their assert that the claim “there are no beliefs” cannot be either thought or said to be true. This is because, according to an anti-eliminativist, embracing eliminativism undercuts the ability to take any thought or linguistic expression to be true. The problem seems to be that, according to eliminativists, anything formulated in thought or language, including the eliminativist view, cannot be true. In the view of this conjecture, the apparent possibility that the timeless, abstract proposition “there are no beliefs” is true, seems to be too little to placate the worries of an anti-eliminativist. Thus, the debate between an opponent and a defender of eliminativism is about the possibility of expressing truth in thought and language, and the issue of whether timeless propositions are primary truth bearers is left aside (even if there are independent reasons to treat propositions as primary truth bearers).

3 Deflationism and the Problem of Truth

The most natural way to get around the problem of truth is to embrace some sort of deflationary account of truth as such a conception (in at least some of its variants) does not presuppose the philosophically problematic version of the notion of content. Such solution was proposed by Kenneth A. Taylor, who claims (Taylor 1994) that the apparent contradiction of eliminative materialism rests on the assumption that when an eliminative materialist claims that certain sentences (like those that ascribe propositional attitudes to people or substantial truth conditions to statements) are false, she must mean “substantially false”. However, this is obviously not the case.

Let us try to elaborate on this proposal. Deflationism about truth and reference is a theory that comes in many varieties, although, for the purposes of the present paper, it will be defined by the claim that truth is not a “substantial property” and reference is not a “substantial relation” (see e.g., Edwards 2013). According to this theory, we can say that certain sentences are true and certain expressions refer. We can even say that there is a property of truth, but these putative semantic properties and facts do not enter into the genuine explanation of phenomena. Such properties are “mere shadows” of predicates; this leads to the conclusion that there is no need to provide any substantial analysis of them.

Deflationism, so understood, explicitly rejects the assumption that, according to Rosenberg, is central to eliminativism. For him, “Eliminativism holds that there is no original intentionality. Without it, there is no derived intentionality and so our speech and writing have no meaning, they are merely noises and chicken tracks. Without original intentionality, no one can think about anything and no noise or mark they make can have derived intentionality; no noise or mark can be about anything or a symbol of anything” (Rosenberg Eliminativism without Tears, unpublished). Thus, according to Rosenberg, the necessary consequence of eliminativism is semantic nihilism: words do not have any semantic properties and sentences cannot be true or false. This is because elements of language can have semantic properties only derivatively, and there are no mental states that are able to provide a basis for original intentionality.

This is precisely the thesis that deflationists reject. Deflationary accounts of truth and reference allow us to reject semantic nihilism without the need to invoke any claims about original or derived intentionality. The fact that words and sentences can have truth values is rooted only in the way our semantic discourse operates. Truth and other semantic predicates operate in our language and play several important functions; therefore, we are perfectly justified in ascribing truth to sentences and reference to words. As truth and reference are not “substantial”, they do not have to be analysed in terms of “derived” or “original” intentionality. For a deflationist, we do not need to make any rich theoretical assumptions about the nature and sources of intentionality in order to make claims such as “‘Snow is white’ is true”. In doing so, we simply give our assent to the sentence in quotation marks.

On a deflationary reading, if one wants to claim that the sentence “there are no beliefs” is true, one might well do so, if and only one is prepared to claim that there are no beliefs. However, there is no need for such philosopher to assent to any presuppositions concerning the nature of meaning or intentionality. Therefore, pace Rosenberg, for a proponent of eliminativism there seems to be quite an easy way to deal with the problem of truth.

4 The Problem of Representation

As we have seen, adopting a deflationary approach allows eliminativists to retain the notion of truth. Still, this does not mean that the question of cogency of eliminativism is solved. In this section, a new challenge for eliminativism that is based on the notion of representation will be presented.

The general idea of the problem of representation is that once we dispose of the notion of original intentionality and adopt a deflationary account of truth and other semantic notions, then we are unable to make sense of the eliminativist thesis as it has traditionally been conceived. Eliminativism has traditionally been seen as a claim to the effect that belief talk is a victim of a massive reference failure; for example, Stich (1996) takes this to be the defining feature of eliminativism. It seems that here Stich presents what many philosophers have taken to be an unquestionable reading of the basic negative ontological claim of eliminativism, namely the claim that there are no beliefs. Many philosophers have apparently assumed that this thesis must be understood as a claim that there is some sort of mismatch between the folk psychological discourse and reality. If this is so, it means that an eliminativist should believe that there is a relation of representation between the elements of our language and the items in the extra-linguistic world, and that this relation fails to obtain in the case of propositional attitudes talk. Considerations of this sort have been popularised by Boghossian (1990), who aimed to show than an anti-realist about any domain must, at least tacitly, embrace a substantialist approach to the semantic.

If this is true—and an eliminativist must assume substantial account of representation to make a claim about the non-existence of beliefs—then it seems that there is a serious problem with the coherence of eliminative materialism. It seems that eliminativism must, at the same time, assume a substantial and a deflationary theory of representation and other semantic notions. On the one hand, it seems that an eliminativist claims that there is a relation of representation as this seems to be a necessary condition of making error theoretical claims (again, this point is stressed by Boghossian 1990). On the other hand, it looks as if such theorist must deny that there is such relation as the only theory of truth available for someone who denies the existence of original intentionality is the deflationary one.

It might be observed that deflationism about truth and deflationism about reference are in principle separable: one might be a deflationist about truth on the one hand and embrace a substantial theory of reference on the other. This might be of help to an eliminativist who could theoretically resort to claiming that they accept the deflationary account of truth but reject the deflationary account of reference, thereby avoiding the problem of representationFootnote 2. This solution, although intriguing, is not the one I find the most plausible. If we buy into the claim that denying the existence of original intentionality implies that there is no substantial notion of truth, we should also be ready to accept a similar line of reasoning with regard to the notions of reference and representation. The philosophical worries about eliminativism stem from the idea that only by accepting something like original intentionality of propositional attitudes might we reasonably claim that our linguistic expressions possess substantial semantic properties.

A defender of eliminativism might react in two ways. First, they might try to defend the claim that it is possible to maintain realism about at least some notion of representation or reference whilst rejecting realism about propositional attitudes. The other way is to try to accept the deflationary approach for all semantic notions and try to coherently express eliminativism in this framework. Although I do not think the first strategy is essentially misguided, the second seems more plausible. The deflationary account of all semantic representational notions seems to be a natural corollary of eliminative materialism, because once we abandon the notion of mental content, then abandoning the substantial notion of representation seems to be a natural next step forward. For this reason, it is worth inquiring into the possibility of framing eliminativism in the deflationary schema, especially given the fact that there are serious reservations about the cogency of such a project.

Such an approach might also be considered more efficient from the point of view of the dialectics of the debate: we grant that an opponent of eliminativism is right with regard to the consequences of eliminativism, yet we try to show that even then eliminativism might be considered a cogent theory. In this way, we prove our point, whilst granting more ground to an opponent and this seems to be a much more efficient argumentative strategy.

Let us then see the source of the tension between deflationism and eliminativism. Deflationary accounts of truth, reference and other semantic notions usually give primacy to the practice. According to the deflationists, we are justified in making claims about truth and reference because of the way our language operates. But, if this is so, it is hard to see how one might claim that the entirety of a certain discourse fails to refer to anything. Yet, the eliminative materialism goes on to claim that a certain region of discourse fails to represent reality properly, and, because of that, this region of discourse should be eliminated. These two claims seem to be intuitively at odds with each other; if there is no such thing as representation, then nothing can be said to fail to represent.

This theoretical tension is the crux of the problem of representation. It might be argued that intuitions of these sorts have been motivating for the widespread belief that eliminativism is a self-defeating position, and they provide explanation why the argument from truth has gained so much traction. Many philosophers seem to be convinced that eliminativists have made the eliminativist position unstable because eliminativists implicitly deny that their claims are true or represent the world in the substantial sense of truth and representation.

In what follows, I will try to show that this near consensus is mistaken and that it is indeed possible to express the main thesis of eliminativism in a theoretical framework which does not make any assumptions about the existence of the relation of representation, understood in substantial manner. The problems which have beset eliminative materialism stem from commitment to the representationalist framework and might be avoided once this framework is changed.

It is worth noting that the eliminativism–deflationism combination resembles contemporary philosophical anti-representationalism of the kind developed by Price (2013). The main motivation behind anti-representationalism is the conviction that it is virtually impossible to combine consistent naturalism with the idea that there is something like a relation of representation that connects thought or language on one hand and mind-independent reality on the other if this relation is to be taken to exist in a substantial, metaphysical sense. As Rydenfelt succinctly put it, according to anti-representationalism “there is no interesting philosophical theory to be given about truth or reference, or ‘aboutness’ in a semantic sense, for any domain of language” (Rydenfelt 2015, 69).

Although the anti-representationalistic picture of language seems to be easily acceptable for someone who, like the eliminativist, rejects the existence of substantial content bearers, it might be unclear how one might be able to express a negative ontological claim in an anti-representationalist theoretical setup. However, there seems to be a fairly straightforward way to do so that will be developed in the next section.

5 The Carnapian Solution

In the previous section, it was suggested that in order to avoid the problem of representation, an eliminativist should try to express the central ontological claim of eliminative materialism in an anti-representationalistic framework. The way to do so, which I will develop presently, is to adopt a version of the broadly conceived Carnapian approach to metametaphysics and to interpret the central thesis of eliminativism, namely the claim that there are no beliefs, through the Carnapian lens.

There are many competing ways of reading the Carnapian insights into the nature of ontological claims and the debate on how to best interpret Carnapian views on the nature of metaphysics is still ongoing (see e.g., essays in Blatti and Lapointe 2016 for an overview on the recent state of the debate).

For the purposes of the present paper, I shall assume a particular reading of Carnapian meta-ontology. This interpretation has no claim to historical accuracy; its merits should be assessed only on the grounds of its usefulness in shedding light on the issue at hand. This reading has no claim to originality either as it is based on proposals put forward by Price (2009), Thomasson (2015, 2017) and Kraut (2016).

There are two basic tenets of the Carnapian approach to metaphysics as it is to be understood for the purposes of the present paper. The first is a claim that existence questions might be posed and answered in two ways. The first way is to pose an internal question: we adopt a “linguistic framework” and then, by way of empirical (or mathematical) inquiry, decide whether a certain specific object exists. In such a way, we might settle whether there are prime numbers greater than 700 (within the framework of arithmetic) or whether there are dodo birds (within the framework of biology). Such questions, however, are not the ones that metaphysicians are prone to ask: they are rather keen on external questions, which are concerned with the validity of whole frameworks. In such a way, we might try to ask, for example, if numbers exist at all, by questioning the validity of the whole number theoretic framework.

However, for Carnap external questions are not “theoretical” in the sense that there is no way an empirical or mathematical inquiry might allow us to answer them directly. These are pragmatic questions: we decide on pragmatic grounds whether to adopt a certain framework, or whether to discontinue using it. What is important for Carnap is that the decision is “not of a cognitive nature”. Still, it is usually “influenced by theoretical knowledge, just like any other deliberate decision concerning the acceptance of linguistic or other rules” (Carnap 1951).

Carnap sums up his position in the following manner:

“A question like: <<Are there (really) space-time points?>> is ambiguous. It may be meant as an internal question, in which case the affirmative answer is, of course, analytic and trivial. Or it may be meant in the external sense: <<Shall we introduce such and such forms into our language?>>. In this case it is not a theoretical but a practical question, a matter of decision rather than assertion, and hence the proposed formulation would be misleading. Or finally, it may be meant in the following sense: <<Are our experiences such that the use of the linguistic forms in question will be expedient and fruitful?>>” (Carnap 1951).

The second and potentially more exegetically controversial claim of the presently accepted version of neo-Carnapianism is that external questions should be read as being metalinguistic theses. The guiding idea is that when we make an existential claim—understood as an answer to an external ontological question—we are in fact making a normative metalinguistic proposal (this reading of Carnap is pursued by Price 2009, Kraut 2016 and Thomasson 2015, 2017, who bases her ideas also on Plunekett’s 2015 conception of metalinguistic negotiations). According to the picture in question, when we make an existential claim we mention certain terms and decide whether or not to employ them and the associated linguistic rules. The proponent of the negative answer to such a question is proposing not to use a given term, whilst the theorist who opts for the positive answer postulates that we make use of it.

Let us consider the question, asked in the external mode, of whether certain Xs exist. The theorist who answers in the negative makes a claim to the effect that we should eliminate the word “X” from our language and stop making reasoning according to inferential rules associated with the term “X”. A proponent of a positive answer would make an opposite claim that we should retain the term “X” in our language and in the relevant rules of usage.

This general picture might be easily adapted to the case of the debate between the realist and eliminativist about the reality of propositional mental states. This observation was made by Kraut:

“Consider a dispute about the reality of mental events and psychological properties. A Philosopher of Mind, impressed by considerations of explanatory/predictive power and systematic elegance, might wish to deploy a discursive framework that mobilizes the vocabulary and inferential resources of propositional attitude ascriptions. This commitment is expressed in her ontological claim that mental events and psychological properties exist. In contrast, her eliminativist opponent is committed to the adequacy of purely neurochemical explanatory resources, thereby leading him to deny any essential role to belief/desire attributions in adequate explanations of human behaviour. This ontological dispute about irreducibly mental events/properties is configured by Carnap* [i.e. Carnap as seen by Kraut] as a manifestation of conflicting commitments to the adoption of a specific discursive framework: viz., one that gives pride of place to causally efficacious and semantically evaluable internal states (Kraut 2016, 41)”.

Although the reading of Carnap which is pursued by Kraut is not the one I wish to fully endorse here, the general idea that the eliminativist thesis should be given a metalinguistic reading is worth pursuing. According to this proposal, the central eliminativist claim that there are no beliefs should be read as a normative metalinguistic claim that we should eliminate the term “belief” from our vocabulary along with assorted linguistic rules.

Kraut’s reading of ontological debates is, in my opinion, problematic in the context of the issue at hand because he explicitly claims that we should see normative statements as expressions of non-cognitive mental states (this is motivated by his commitment to classical expressivism). However, in my opinion, such commitment might be unwelcome if one wishes to pursue the Carnapian reading of the ontological debate about propositional attitudes. This is because, if one treats ontological claims as normative ones and one considers normative statements to be expressions of psychological stances, commitments or attitudes, then it seems that one must presuppose realism about psychological states in order to be able to make any ontological claim (see Kraut 2016, 54). However, this puts a putative Carnapian anti-realist about psychology in a tight spot, as they must on the one hand assume psychological realism and deny it on the other.

This problem might, however, be easily solved. The talk about “stances, commitments and attitudes” might be reinterpreted in non-psychological jargon because the psychological commitments of Kraut’s version of Carnapianism do not seem to be central to the Carnapian understanding of metaphysical disagreements. The notions of stances, commitments and attitudes might be translated into an inferentialist framework, which, interestingly, is already present in the writings of Price (see e.g., Price 2013, 31–32). In what follows, it will be assumed that external ontological claims, in Carnap’s sense, do not presuppose any substantial psychological commitments (this might go against original Carnap’s intentions, but, as has been said, the aim of the paper is not to provide a historically adequate reading of Carnap, but rather a model of broadly understood Carnapianism to help to solve the issue at hand).

Adapting such version of Carnapianism to the issue of the status of the eliminativist proposal allows the problem of representation to be solved as it frees us of any need to invoke the notion of representation in an explanation of eliminativism. An eliminativist, read in a Carnapian fashion, is someone who makes a normative proposal about the way we talk, and this is consistent with her anti-representationalist assumptions.

A theorist, who argues that there are no vehicles of “original intentionality” and, consequently, no substantial relation of representation between the mind and mind-independent reality, is still perfectly entitled to propose changes in the way we speak, even if these changes involve rejection of certain vocabularies. There is no problem for the eliminativist to make claims about the usage of language as denying that there is anything like intentionality does not amount to the obviously false claim that there are no patterns of linguistic usage. An eliminativist might reflect on the patterns of linguistic use and make reports about such facts. She might also put forward normative revisionary claims about how language should be used. Obviously, the proposal that we should eliminate belief-talk might not be justified by claims about the mismatch between folk psychology and “mind-independent reality”. However, such proposal might be justified pragmatically: an eliminativist is free to argue that changing from the folk psychological to the neuro-scientific idiom in the way we describe ourselves and other people might benefit us in one way or another.

Thus, there seems to be no need for an eliminativist to adopt a substantial conception of semantic notions such as truth or representation in order to be able to put forward the claim that there are no beliefs as this claim simply amounts to the thesis that we should discontinue using the term belief. If the eliminativist thesis is understood in the Carnapian fashion, then the eliminativist is not forced to make any representationalistic assumptions, as her claim is about the proposed changes in linguistic use and not about any word–world relations. In this way, it might be claimed that the problem of representation, which has seemingly motivated many sceptical assessments of eliminativism, is solved. This also shows that there is no deep problem of truth for eliminativism as adopting the deflationary approach to truth is everything that a theorist of eliminativist inclinations needs in order to coherently express the central claim of her theory.

Obviously, adopting a Carnapian approach to metametaphysics does not force anyone to become an anti-realist about propositional attitudes—it only shows that it is possible to do so. Whether it is an advisable option is another issue. Although it is not my aim in this paper to provide a definite answer to the question of whether we should embrace eliminativism (but only to defend the idea that this position might be expressed), some preliminary remarks are in order.

The question “should we eliminate the term belief and the associated inferential rules?” might be understood in two ways. The first is the question of whether we should discontinue to use them in the context of cognitive science. In other words, should we use notions such as belief when we try to construct a respectable scientific theory about the workings of the human cognitive system? The second question, should we discontinue using belief talk in ordinary life?

The distinction between these questions is present in John Collins’s discussion of “meta-scientific eliminativism” (Collins 2007). For Collins, there are two kinds of eliminativism about beliefs and content: he dubs one “futurological” and the other “meta-scientific”. Futurological eliminativism is the view that at some point in history, ordinary people will stop using the term “belief”. The scientific one is the view that mature sciences of cognition will not help themselves to such folk notions.

For Collins, only the second version of eliminativism is plausible. The futuristic one is based on the flawed assumption that showing that a certain concept is defective from the point of view of standards of scientific inquiry would (and should) lead to its elimination from our common parlance. According to Collins, the scientific credentials of a given concept do not determine its use in ordinary life, and it is hard to argue with this conclusion: many, concepts which have been eliminated from sciences like biology and chemistry, are still used in folk contexts.

The issue of whether mature cognitive sciences will use have use for the notion of belief is still an open question as there are important voices arguing for the claim that that folk psychological vocabulary is indispensable in serious scientific research (see e.g., Burge 2010). The issue seems hard to settle given the fact that contemporary cognitive science is still an actively developing field. Still, what might be safely claimed is that, despite the substantial opposition, eliminativism is a viable yet non-obligatory option in this area. Given the state of the evidence, we might grant as much that planning to dispense with folk psychological vocabulary in the scientific context is an admissible option, whereas in the context of folk talk, it remains a rather implausible one.

This conclusion leaves us with two questions: first, is Carnapian eliminativism really justified when applied to the discourse of cognitive sciences? Second, would such possible elimination be enough to make good of the claim that beliefs do not exist, pure and simple? There might be a temptation to say that even if it turned out that belief talk should be eliminated from the scientific discourse, it would be not enough to justify the claim that “beliefs do not exist” if belief talk were retained in the folk domain.

These are complicated issues, and fully addressing them would require a separate study. The present remarks are meant only to show what questions would need to be addressed should one try to actually argue for eliminativism in the Carnapian framework. The aims of this paper are more modest: I have tried to show how one could coherently express eliminativism in this framework, but arguments for actually adopting such an option still to need to be developed and properly assessed.

6 Problems with the Carnapian Solution

As has been shown, the adoption of the Carnapian approach to metaphysics allows the problem of representation to be solved. However, this interpretation leads to certain issues that need to be resolved.

First, it might be observed that the proposed solution is based on a revisionary reading of eliminative materialism. Churchland probably did intend his theory to be based on representationalist assumptions as his main complaint about folk psychology were that it does not meet the criteria that all respectable scientific theories should meet.

In response to this worry, it might be said that revisionary reading of all first-order ontological debates is a characteristic feature of the Carnapian approach to meta-ontology (see Kraut 2016). A Carnapian is someone for whom the literal reading of traditional philosophical debates in ontology is deeply problematic: what does it mean, a Carnapian might ask, to deny (or accept) the existence of numbers altogether? In order to make sense of this apparently incomprehensible disagreement, she is proposing an alternative reading of competing ontological positions.

In the case of eliminative materialism, such revisionary reading seems to be even more warranted if we take into account the fact that the eliminativist position is frequently accused of being inconsistent. In such case, presenting a coherent version of eliminativism might be taken to be an important task, even if consistency is achieved by departing from original formulations and motivations. So, even if we agree that the original version of eliminative materialism was formulated in a representationalistic framework, transposing eliminativism to the Carnapian framework might be a theoretically fruitful move.

Nonetheless, it is worth stressing that such reconstruction resembles the original theory as it embraces the central claim of Churchland’s theory, namely, the negative ontological statement that there are no beliefs. This way, eliminativism read through a Carnapian prism is still a form of eliminativism. The reading of eliminativism presented here differs from the original formulations and motivations as it takes away the representationalistic presuppositions which were present in the Churchlandian version. Still, the central ontological claim of eliminativism is retained in the anti-representationalist version.

It is also worth noting that such reading of eliminative materialism is not entirely new. Indeed, one might find similar ideas in Rorty’s version of eliminativism about sensations that was presented in his early papers. For Rorty, the central claim of what he calls the “disappearance form of identity theory” is that at some point in the future, we might be able to drop terms such as “sensation”, “pain” or “mental image” (Rorty 1965). Rorty explicitly denies that we actually will cease to use such terms as for him that would be most impractical. However, what is important for him is that ceasing to use them is possible in principle.

Rorty’s theory contains the metalinguistic element which is central to the interpretation of eliminativism presented in this paper. Obviously, it would be anachronistic to claim Rorty was a proponent of the Carnapian approach to metametaphysics in the form in which it is developed by contemporary philosophers. Still, it is important to note that a pragmatic metalinguistic streak has always been present in the eliminativist tradition.

As far as Churchland’s philosophy is concerned, the exegetical issues are not as simple as is usually assumed. Although there is a seemingly strong commitment to representationalism in the way eliminative materialism is presented, there is, as Rockwell (2011) has observed, a certain tension in Churchland’s writings between his eliminativist thesis on the one hand, and a generally pragmatic approach to philosophy of science on the other. This is not the place to dwell on the specifics of Churchland’s exegesis, but it is important to note that there are some interpretative issues that might make at least some room for a more pragmatic-friendly reading. Thus, even if the proposed Carnapian reading is revisionary, it might be claimed that it is not unreasonably so.

Another problem with the interpretation of eliminativism proposed here is that, in order to avoid the problem of truth (which has been discussed in section 3), we must allow statements which express external ontological claims to be truth-apt. This might be rejected by some Carnapians (see e.g., Kraut 2016) who see Carnap as being committed to non-factualism about ontological claims. This is a big issue in the interpretation of Carnap, and it is not the aim of this paper to settle such disputes. It is nonetheless important to note that the feasibility of the project pursued in this paper requires a version of Carnapianism that allows ontological statements to be true, at least in deflationary sense.

7 Conclusions

Taken as a claim that there are no beliefs, eliminative materialism about folk psychological ascriptions of belief and similar states should be reinterpreted as an answer to an external question in the fashion of post-Carnapian metametaphysics. On this reading, the main point of eliminative materialism is a normative thesis that we should stop using the term “belief” (and similar terms) and cease to reason according to the inferential patterns associated with this term. In this way, eliminativism can avoid the problem of representation, which might be seen as motivating much of the philosophical scepticism about eliminativism. Cast in Carnapian terms, to put forward its central claim, eliminative materialism does not need to make any substantial assumptions about truth, reference and other semantic notions. Thus, an eliminative materialist is able to express the central claim of his theory with the help of the deflationary approach to truth.

Of course, such manoeuvre only makes eliminativism free from a certain version of self-inconsistency charges and there are many others that might be presented. Moreover, as was noted in section 5, nothing that has been said here purports to prove that eliminative materialism is true (even in the deflationary sense of “true”). If the reasoning presented above is acceptable, this would require showing that disposing of the term “belief” and the associated linguistic rules is indeed a pragmatically good option; however, such an argument seems to be hard to produce. Still, the hope is that at least some theoretical worries have been elevated that have plagued the eliminative materialist position, which it seems has been dismissed too quickly by many philosophers. The position in question should be assessed on the basis of the merits (or lack) of the arguments that support it, and not because the conclusion seems superficially incoherent.

The result is important also because of the fact there is some tendency to associate Carnapian meta-ontology with realistic views on first-order ontological debates (for example, Thomasson (2015) uses the Carnapian framework to prove the existence of ordinary objects). Obviously, a Carnapian might propose a positive response to first-order questions about existence, but the reasoning presented in this paper shows that this is not an obligatory option. There is theoretical room for a Carnapian to adopt a negative approach to first-order ontological questions. In fact, in at least some cases, a philosopher endorsing a negative revisionary ontological claim might be better off adopting a Carnapian approach to meta-ontology than the realist meta-ontological view. This is certainly true in the case of eliminative materialism.