Religious fictionalism and non-doxastic religion

Undeniably, fictionalism and non-doxasticism have a lot in common. They both constitute belief-less ways of engaging with religion, challenging the traditional view that belief is fundamental for religious life. Both also tend to stress the moral and spiritual benefits one might receive from a belief-less engagement with religion. However, when it comes to the attitudes towards religion suggested, and to the importance of truth in religious life, there are stark contrasts between the approaches.

As already mentioned, some philosophers treat non-doxasticism and fictionalism as if the terms were interchangeable.Footnote ³ Sometimes this tendency seems born out of confusion, but on most occasions, it is rather the case that the distinction is of little importance to the author in question. Are such authors necessarily wrong in treating the terms as synonymous? Two points are important here.

First, terminology is largely irrelevant for what I have to say. My arguments only rest on the assumption that there are substantial differences between different belief-less approaches to religion, especially between those accounts for which the possibility of truth is important, and those holding that truth is irrelevant. My points would therefore remain even if we were to treat non-doxasticism and fictionalism as synonymous, although they would need to be formulated in other terms.

Second, even though some philosophers use the terms interchangeably, most engaged with these topics do not. There are two main approaches to belief-less religion, one standardly called religious fictionalism and the other labelled in non-doxastic terms. Since the differences I am interested in are differences between these main approaches, I use the terms ‘fictionalism’ and ‘non-doxasticism’ accordingly, well aware that there are other ways one might divide (or refrain from dividing) the field of belief-less religion.

To make the distinction between fictionalism and non-doxasticism as clear as possible, I will focus on the kind of fictionalism advocated by authors such as Andrew Eshleman and Robin Le Poidevin (a view these authors label ‘instrumentalism’), which is also in line with how the term ‘fictionalism’ is commonly understood in a larger philosophical context (Le Poidevin (Reference Le Poidevin1996), Idem (Reference Le Poidevin2003); Eshleman (Reference Eshleman2005), Idem (Reference Eshleman2010) ).Footnote ⁴ The following passage from Le Poidevin presents the position in an illustrative way:

Religious fictionalism is commonly contrasted against non-realist views that take a reductionist stance towards religious language. Borrowing terms from the philosophy of science, Eshleman and Le Poidevin compare their favoured ‘instrumentalism’ with theological ‘positivism’, a view according to which expressions about the supernatural are reinterpreted as being about the mundane. The major difference between fictionalism and positivism is that on fictionalism, religious language is not reinterpreted but retains its original meaning. What changes on the transfer from a realist to a fictionalist account does not have anything to do with meaning, but with truth (Le Poidevin (Reference Le Poidevin1996), 112; Eshleman (Reference Eshleman2005), 187–188).

Truth-value is irrelevant on a fictionalist reading, which is commonly expressed through the claim that fictionalist language is not truth-normed. This must not be confused with the stronger claim that a language is not truth-apt. A language that is not truth-apt is considered non-descriptive and lacks truth-value. A language not being truth-normed only means that the language's truth-value is irrelevant for its use. This view makes it possible to claim that a fiction is literally true or false, while it also has the resources to explain why this does not matter.Footnote ⁵ As we shall see, the irrelevance of truth marks the sharpest differences between fictionalism and non-doxasticism.

The basic idea of non-doxasticism is that some attitude weaker than belief is acceptable as grounds for a religious commitment. Non-doxasticism comes in many different versions, with different accounts suggesting different attitudes to take the place traditionally held by belief. Most of these accounts come in terms of propositional faith, and the main idea is that instead of belief, such faith might be based on some weaker cognitive pro-attitude like acceptance, assumption, or hope.Footnote ⁶ Other accounts challenge the basic assumption that propositional faith can ever be based on belief, offering an analysis where propositional faith and belief are incompatible.Footnote ⁷ Others still suggest that faith is altogether unnecessary, claiming that a religious commitment might be based on hope alone.Footnote ⁸ However, these differences aside, there are at least two defining features that seem to be common to all accounts of non-doxasticism.Footnote ⁹

First, non-doxasticism requires something like desire and/or positive evaluation. This provides motivation for taking on non-doxastic religion and limits its availability to those who actually desire its object or judge it a good thing.

Second, non-doxasticism requires the subject to regard the truth of the proposition towards which its attitude is directed as an epistemic possibility.Footnote ¹⁰ Exactly how this notion of epistemic possibility is spelled out varies with the account, from the very wide sense of it not being known or justifiedly believed to be false to the narrower notion of it having a non-negligible chance of being true.Footnote ¹¹

Summing up these two defining necessary conditions, we can say that for S to have a non-doxastic pro-attitude towards some proposition p, the following two conditions must be met:

• ND1: S desires that p or judges that p is an overall good thing.

• ND2: S believes p to be epistemically possible and neither believes nor disbelieves p.Footnote ¹²

Most accounts of non-doxasticism come with some further conditions. This is especially true for accounts working with complex attitudes such as propositional faith.Footnote ¹³ Hope, however, which is arguably the simplest of the non-doxastic attitudes, is often analysed as being constituted by only ND1 and ND2.Footnote ¹⁴ Therefore, a minimal, common denominator-account like the one ND1 and ND2 aims to provide should not go beyond these two conditions.

Before continuing, it is worth pointing out that the correspondence between this smallest common-denominator account of non-doxasticism and the standard analysis of hope strongly implies that hope is the most fundamental non-doxastic attitude, in the sense that if you have any non-doxastic attitude towards p, you also have a hope directed at p. If this is correct, as I presume it to be, it means that if any non-doxastic attitude is rationally available to the subject, hope is already present. I will exploit this feature in ‘An argument for exclusive availability’.

Summarizing the main differences between the approaches, we can start by noting that the contrast between ND2 of non-doxasticism and fictionalism's denial of religious language being truth-normed marks the most important difference. On a fictionalist account, truth is irrelevant, while a non-doxastic engagement is not possible without truth being an epistemic possibility. Unlike the fictionalist, the non-doxasticist agrees with the common religious believer that religious language is truth-normed, and the possibility of truth is vital.Footnote ¹⁵

The basic difference regarding the relevance of truth is reflected in the different attitudes suggested by the approaches. The fictionalist's ‘make-believe’Footnote ¹⁶ is clearly not a truth-normed attitude and suggests affinity with pretence and playing games. The non-doxastic main attitudes of faith and hope, on the other hand, indicate both the importance of truth, and the position of uncertainty its practitioner is in regarding this truth. You cannot have hope or have faith in something you know is false, and if you know it is true, hope and faith are superfluous.

Non-doxasticism is only open to agnostics (since it precludes outright belief and disbelief), while fictionalism is commonly thought to be open to agnostics and atheists alike. This is important since it opens the question which of the two approaches the agnostic should prefer. However, before answering that question, we need to consider the argument that non-doxasticism cannot be properly distinguished from fictionalism.

Can non-doxasticism be distinguished from fictionalism?

In a recent article, Finlay Malcolm argues that non-doxasticism overlaps with fictionalism, since some agnostic fictionalists might satisfy the conditions of non-doxastic faith.Footnote ¹⁷ According to Malcolm, this threatens to collapse non-doxasticism into fictionalism, and he also claims that the theoretical costs for mending this problem are too high. The argument rests on the assumption that both fictionalism and non-doxasticism are open to the agnostic. While I challenge that assumption in the last major section, here I will grant it for the sake of discussion, pointing to other problems with Malcolm's argument.

As a point of departure, Malcolm uses the following definition of non-doxastic religion:

(3) matches ND1 from our definition in the previous section, while (1) and (2) roughly corresponds to ND2.Footnote ¹⁸ In other words, Malcolm's definition of non-doxastic religion seems quite uncontroversial. He then goes on to define fictionalism:

Malcolm argues that fictionalists might have faith since the definition of non-doxastic faith is unable to rule out agnostic fictionalists (as opposed to atheist fictionalists, which are clearly ruled out by condition 2).Footnote ²⁰ According to Malcolm, there must be something wrong with a theory of faith that is not able to distinguish real faith from pretence (ibid., 224). He even claims that if the agnostic fictionalist counts as having faith, the whole non-doxastic enterprise might collapse into fictionalism:

Taken at face value, Malcolm's claim about a threatening collapse seems exaggerated. To argue that a phenomenon A collapses into another phenomenon B is a strong claim, and what needs to be demonstrated is generally that all instances of A are also instances of B and that they are nothing beyond B. Pointing to a possible partial overlap, as Malcolm does, is nowhere near enough to demonstrate collapse. To collapse into B, the entirety of A needs to be overlapped by B.Footnote ²¹

Given Malcolm's definitions of the two approaches, it is hard to see how a collapse could ever occur. According to Malcolm, the non-doxasticist can be motivated by an ‘affective attitude’, while the fictionalist is only motivated by ‘purely instrumental reason’. Since it would stretch the meaning of the words beyond any reasonable limits to claim that an affective attitude is an instrumental reason, it seems obvious that even on Malcolm's account there are non-doxasticists that are definitely not fictionalists. It is therefore hard to make any sense of his claims about a threatening collapse.

However, I am not sure we should take Malcolm's collapse-claims at face value, since his discussion only seems to presuppose the more modest idea that non-doxasticism cannot be properly distinguished from fictionalism.Footnote ²² At least this is the case with his discussion of possible solutions, which is centred on the possibility of elaborating the definition of non-doxastic faith in a way that rules out the agnostic fictionalist.

Malcolm begins by identifying what is presumably the most important difference between fictionalism and non-doxasticism, namely that truth matters deeply to the subject of non-doxastic faith, while it is irrelevant to the fictionalist. A straightforward way to solve the problem would be to include this difference in the definitions. However, Malcolm is reluctant to do so. He argues that someone with non-doxastic faith might stop caring about the truth, without abandoning her religious commitment (Malcolm (Reference Malcolm2018), 226). While this is obviously true, it eludes me why Malcolm takes this to speak against adding a condition about the importance of truth to the non-doxastic definition. If the difference between the approaches lies in their relation to truth, the subject of Malcolm's so-called counterexample would simply move from non-doxasticism to fictionalism the moment she stopped caring about truth. I cannot see why that would be problematic. So presumably, fictionalism could be ruled out by elaborating condition (3) of non-doxasticism along these lines:

• (3*) a positive evaluative/affective attitude towards p and the truth of p.

Instead of this simple solution, Malcolm goes on to ask why truth matters on the non-doxastic approach. The answer, he claims, is that the goods the non-doxastic subject is hoping to achieve by her commitment are dependent on the truth of her chosen religious propositions. In Christian terms, non-doxastic faith will only bring eternal bliss if it is true that God exists. However, since the attainment of certain goods is also central to the fictionalist, Malcolm thinks an explicit mentioning of ‘eternal goods’ is needed. He proposes the following elaboration of condition (1) from the definition of non-doxastic faith:

• (1**) a positive cognitive attitude towards p, adopted as a means to attain both temporal and eternal goods. (Malcolm (Reference Malcolm2018), 228)

According to Malcolm, this definition clearly separates non-doxastic religion from the fictionalist approach. However, he also finds reason to reject it because the theoretical costs are too high: the new definition only covers religious non-doxastic faith, while the old definition was able to include secular, everyday faith as well (ibid., 228–229).

I agree with Malcolm that a proper account of faith should include everyday faith, since it seems implausible to view the attitudes involved in religion as distinct from the corresponding everyday attitudes. I also think we should reject (1**) for other reasons. First, it seems overly Christian and not applicable to a non-doxastic engagement in religions more concerned with the here and now. Second, if an affective attitude is enough to take on non-doxastic religion, as it is according to the definition Malcolm works with, it is simply not true that one needs to be ‘in it for the goods’. A deep love for the idea of God and a deep longing for its truth could clearly work as motivation, regardless of any goods, temporal or eternal. Third, and most important, we already have (3*), which is all we need to successfully separate non-doxasticism from fictionalism.

In this section I have pinpointed some weak spots in Malcolm's reasoning, and shown how non-doxasticism might plausibly be distinguished from fictionalism. I think there is more to say on this subject, since I think that Malcolm's basic assumption that these approaches are open to the same agnostics is false. If so, the problem of distinguishing them does not even arise. I will elaborate that claim in the argument for exclusive availability. In the next section, however, I will continue to grant the assumption that fictionalism and non-doxasticism are available to the same agnostics. By demonstrating non-doxasticism's ability to handle the major problems associated with fictionalism, I aim to show why a rational agnostic should prefer the former kind of belief-less religion.

Non-doxasticism and the problems of fictionalism

There are at least two major problems standardly encountered in the literature on fictionalism: the problem of justifying religious practice and the problem of justifying the use of religious language. Since, as far as I am aware, non-doxasticism lacks any canonical problems of its own,Footnote ²³ these problems constitute strong reason to prefer that approach over fictionalism.

The first major objection to fictionalism is about the justification of religious practice. The worry is simply that fictionalism is too weak a position to be able to ground a religious commitment. Harrison has formulated the objection as follows:

In this regard, it is important to notice that fictionalists typically do not hold that fictionalism has the same force as realism. Eshleman explicitly rejects the view that fictionalist religion needs to be equally powerful or leads to the same goods as realism. All that is needed to justify fictionalism is that ‘the practice of a fictionalist form of religion is a means to promote human flourishing’ (Eshleman (Reference Eshleman2005), 195).

However, one must not conflate the question whether it could be reasonable to engage in religious fictionalism with the question whether fictionalism is able to ground religious action. That fiction alone is not a sufficient ground for action is something proponents of fictionalism generally agree with (Deng (Reference Deng2015), 209; Le Poidevin (Reference Le Poidevin, Buckareff and Nagasawa2016), 187). A fictionalist needs independent justification for her actions. This is only reasonable. Consider the difference between Don Quixote, who acts as if fiction is literally true without independent justification, and a role-player who dresses up as a knight and partakes in a live-action role-playing game, fighting windmills made to look like giants. The motivation for the role-player might be of various kinds. Perhaps the immersion into fantasy gives him a chance to explore sides of himself that lies dormant in ordinary life. Or perhaps he is just in it for social reasons. The only thing we can be sure of is that unlike Don Quixote, his reason for acting is not to save anyone from threatening giants. Presuming the role-player is sane and rational, his basic reasons for acting will lie entirely outside the fiction.

Like the role-player, for her religious actions to make sense, the religious fictionalist needs independent motivation. For example, to follow the moral code of her fictive religion, she needs already to believe in the soundness of this morality. Or to pray, she needs to find some non-religious, presumably psychological, reason to do so.

Non-doxasticism constitutes a stronger foundation for religious practice than fictionalism since there is no need for independent motivation. The agnostic engaged in non-doxasticism prays because she at least hopes that God exists. The justification of religious action on non-doxasticism is of roughly the same kind as such justification in belief-based religion, even though, obviously, it is not as strong. Unlike fiction, non-doxastic attitudes can be independent reasons for action. This even goes for hope, arguably the weakest of these attitudes. Even though one is seldom obliged to act on a hope, and even though there might often be overriding reasons for acting on hope alone (see next section), a hope can, in the right circumstances, function as justification for action. If you hope to catch an earlier train and you lack any overriding reasons not to act on that hope, you might very well run the last hundred metres to the station. And on stronger non-doxastic attitudes, like propositional faith, the grounds for action become much more robust.

As for the second major problem, one might legitimately wonder why fictionalists use religious language. Since the goods the fictionalist is after, on most accounts, are psychological and moral growth, why not talk about these things in a straightforward and truth-normed way? Why suppose that a non-religious moral and spiritual discourse is so impoverished that one is better off playing a religious game of make-believe?

Fictionalists are not without response to this challenge. Eshleman, who thinks that fictionalist religious life should be centred on ideals such as ‘be imitators of God’, stresses the symbolic nature of religious language, and claims that symbols such as ‘God’ cannot be reduced to other terms:

It is unclear what Eshleman means with ‘for the same reason’ here. Traditionally, realist theologians stress the need for symbolic language because in their view, God surpasses human understanding. The idea is that God lies beyond the reach of what human language can speak directly and truthfully about, why symbolic representation is required. Since Eshleman denies that religious language is truth-normed, it could hardly be ‘for the same reason’ fictionalists require symbolic representation.

I suspect that what Eshleman is really after is that the instrumental and expressive functions of religion he wishes to retain are best promoted by religious symbols like ‘God’, which cannot be reduced to naturalist language without loss of force. This seems reasonable enough, but it also seems that Eshleman might be presuming too much by way of continuity between how these symbols work on a realist and fictionalist reading. It seems unlikely that a symbol of ultimate reality retains its force in the transition from realism to fictionalism. Presumably (and perhaps contrary to the intention of C. S. Lewis himself) reading about the lion Aslan (the personified good) in Narnia will not make you a good Christian, at least not if you read the stories as fiction. However, if you believe in (or commit non-doxastically to) the Christian story, and read Narnia as a Christian allegory, the symbol of Aslan (as a symbol for God) might have an altogether different impact. One cannot suppose without argument that the symbol has the same effect on a fictional reading as it has on a truth-normed reading. To defend the need for religious language on a fictional approach the way Eshleman intends, it is not enough to show that religious language concerning moral and spiritual issues is superior to non-religious language on an ordinary reading. One must also show that religious language remains superior on a fictive reading.

However, it seems doubtful that fictionalism has the resources to back up such a claim without becoming inconsistent. A fictionalist needs independent motivation for her religious action. On any plausible reading, this motivation comes in the form of non-religious motivation, which can hardly be problematic to express in non-religious terms. In order to defend any irreducibility claim, the fictionalist needs to show that her religious game of make-believe contains something important that cannot be expressed in non-religious language. But what can that be, if the fictionalist gets the goods she is after as a result of her independently motivated actions, describable in non-religious terms? How can the symbol of ‘God’ offer additional motivation that non-religious language cannot, if the fictionalist needs independent motivation, explicable in non-religious terms, for all of her actions? The view that independent motivation is needed for religious actions makes it hard to maintain that religious language cannot be reduced to non-religious language without important aspects being lost. Therefore, it seems that the fictionalist would need to find other grounds than irreducibility to defend her use of religious language.

Perhaps the fictionalist could meet this challenge by pointing out that engagement with any fiction can be morally significant. After all, fiction is superior to abstract moral discussion in awakening our morally relevant emotions, adding life and flavour to an otherwise dry moral discourse. In engaging morality through fiction, we get engaged emotionally and not just intellectually.Footnote ²⁴

While this line of defence surely has some merit, it also opens the door for other kinds of objections since it reduces religion to one fiction among many. Most importantly, we must now ask why we should turn to fictionalist religion rather than any other kind of morally relevant fiction. Why choose to play a game of religious make-believe instead of reading a novel or watching a play? Since, as I have argued, the idea that religious language has some uniquely motivating power seems untenable on fictionalism, on what other grounds could the fictionalist argue that religion is the preferable form of morally significant fiction?

I am not claiming that these further objections are ultimately unanswerable to the fictionalist. I only wish to highlight the difficulties facing fictionalist use of religious language, which is significant since on non-doxasticism this is a non-issue. On the non-doxastic approach, religious language remains truth-normed, with functions intact. When the non-doxasticist uses religious language, she means every word of it, at least in the same sense the realist does (since religious language is heavy with metaphor and symbolic figures, exactly what she means and how it is truth-normed may not be entirely obvious, but that holds for realist use as well). That the subject is agnostic and in doubt regarding the truth has no bearing on her language use.

An argument for exclusive availability

In previous sections, I have granted the presupposition that the agnostic can choose freely between non-doxasticism and fictionalism. I will now argue that this presupposition is false. Either the agnostic hopes p, in which case she cannot engage in fictionalism, or she does not, in which case she cannot engage in non-doxasticism. That a hope is needed to engage in non-doxasticism is strongly implied by the assumption that hope is its smallest common-denominator and it means that without hope, a non-doxastic engagement is not possible. That such hope also precludes fictionalism is a much more substantial claim, one I will elaborate and defend in what follows.

Let us carefully consider the position of the religious agnostic. It is constituent of her position that she neither believes nor disbelieves the religious propositions over which she is agnostic. To her, they constitute epistemic possibilities. Therefore, the agnostic by default satisfies the ND2-condition of non-doxasticism. It seems reasonable to suppose that if the agnostic is to leave pure agnosticism behind and take on belief-less religion, be it fictional or non-doxastic, she must have some reasons. Such reasons might be purely instrumental, or they could also be evaluative, meaning that the agnostic is pro-religious and views the religious propositions in question as good or desirable. If she has only instrumental reasons, the agnostic can engage in fictionalism but lacks what is needed (ND1) to be a non-doxasticist. However, if she is pro-religious, she also satisfies ND1, which qualifies her as hoping (since by being agnostic, she already satisfies ND2). In other words, an agnostic with a pro-attitude is a non-doxasticist by default, having the non-doxastic attitude of hope.

Since the pro-religious agnostic has hope, fictionalism is unavailable to her. Why? Because on fictionalism, the truth-value of p is supposed to be irrelevant, but it clearly matters in a significant way to anyone with a hope that p. If the agnostic is to remain rational, she cannot hold the truth of p to be irrelevant and at the same time hope for the truth of p.

Let me repeat the argument a bit more formally:

• Premise 1: An agnostic with an ND1-attitude towards p possesses the non-doxastic attitude of hope towards p.

• Premise 2: Having hope that p is rationally incompatible with taking a fictionalist stance towards p.

• Conclusion: An agnostic with an ND1-attitude towards p cannot rationally take a fictionalist stance towards p.

The first premise builds on the assumption that ND1 + ND2 is enough for hope. It might be questioned by challenging the standard analysis of hope. The second premise seems harder to challenge, but perhaps it could be questioned if hope really precludes fictionalism. I will now consider these challenges in order and modify the premises to account for them.

The first premise presupposes the standard analysis of hope. While I am not aware of anyone who would deny that ND1 and ND2 are necessary for hope, many philosophers claim that the analysis is not complete and that some further conditions are needed.Footnote ²⁵ When such further conditions are added to the analysis of hope, it often results in something similar to James Muyskens's analysis, in terms of necessary and sufficient conditions:

Conditions (a) and (c) are uncontroversial and correspond to ND1 and ND2 of the minimal view. Although quite popular in analyses of hope, at least in a non-doxastic religious setting,Footnote ²⁷ (d) is obviously false. As pointed out by Daniel J. McKaughan, hoping to win the million-dollar jackpot does not constitute a good reason for spending as if the money was already on your bank-account (McKaughan (Reference Mckaughan2013), 113). This does not mean that one might never act on a hope. On many occasions, it might be rational to do so. It only means that action is not mandatory, and that it should not be included in an analysis in terms of necessary and sufficient conditions. Which leaves us with (b) to consider.

The reason to include (b) in the analysis of hope is to exclude cases like the recovering alcoholic, who presumably desires to have a drink badly without having the corresponding hope that she will (since it would ruin her life). We might call this the override condition of hope.

Including an override condition suggests that the agnostic can fulfil ND1 and ND2 while still lacking hope, given that she possesses a relevant overriding reason. Imagine for example an agnostic who, although attracted by the powerful symbols of theistic religion, does not hope for theistic religion to be true since on balance she much prefers the predictability of a naturalistic view of the world. Therefore, the argument has to be further nuanced:

• Premise 1*: An agnostic with an ND1-attitude towards p, who lacks overriding reasons not to hope p, possesses the non-doxastic attitude of hope towards p.

However, introducing overriding reasons changes nothing of importance. Even though an agnostic with overriding reasons not to hope can take on fictionalism, fictionalism is still only available to those agnostics for whom non-doxasticism is not available. Presumably, this point generalizes. If one were to develop the definition of hope by adding further conditions, each such condition would only exclude some further group of agnostics from non-doxasticism, and the point would remain that for any agnostic that fails to satisfy the conditions, only fictionalism is available.Footnote ²⁸

Premise 2 expresses the assumption, crucial to the argument, that the agnostic cannot take on both non-doxasticism and fictionalism at the same time. This is because the approaches treat religious truth-claims in fundamentally different ways. According to fictionalism, truth is irrelevant while on non-doxasticism, the possibility of truth is vital. These different attitudes are mutually exclusive, at least rationally if not psychologically as well. A rational subject aware of her epistemic situation cannot hold that the truth-value of religious language is both vital and irrelevant at the same time.

An agnostic cannot take on both approaches simultaneously, but there is nothing stopping her from alternating between the stances. My analysis suggests two major kinds of situations where this might happen: if she gains or loses some overriding reason for engaging in non-doxasticism, or if she either stops or starts caring about the truth of p. Put more precisely, my claim that the two approaches are not available to the same agnostic means that they are not available to her at the same time. This is important, since it explains some purported counterexamples, like Malcolm's non-doxasticist who stops caring about the truth.

At this stage, one might object and claim that my argument seems to lead to an absurd conclusion, namely that one cannot rationally play a game of make-believe about something that one hopes for. It is easy to imagine counterexamples to this. Consider the children of a soldier at war, hoping for their father's return. Clearly, they do nothing wrong when engaging in a game of make-believe about their father's homecoming, pretending he is already back.

The objection rests on the assumption that every game of make-believe is an instance of fictionalism. I very much doubt that this is the intended reading of the fictionalism advanced by Eshleman, Le Poidevin, and others. The fictionalist stance requires that one considers truth irrelevant and that one takes on the attitude in question to achieve some ulterior goods. These conditions seem hard to reconcile with the idea that every game of make-believe is an instance of fictionalism, since obviously, they do not apply to the most common games of make-believe there are, namely children's games. Children do not restrict their games to things whose truth-value they consider irrelevant. To most children it is really important that there are no monsters in reality (especially not in the cupboard or under their beds), but still they love to play monster-games. Neither do children need any instrumental reason to play games.

Since it seems obvious that one can play a game of make-believe without adhering to the further requirements of fictionalism, it seems safe to suppose that this is the case also with the soldier's children. However, if for some reason I fail to anticipate the fictionalist would be unhappy with instances of make-believe outside fictionalism, this objection too could be handled by claiming that the children effortlessly change between the stances of non-doxastic hope and fictionalist make-believe. In either case, premise 3 holds.

If we substitute the premises of the original argument with the modified ones, and adapt the conclusion, we get a new argument as follows:

• Premise 1*: An agnostic with an ND1-attitude towards p, who lacks overriding reasons not to hope p, possesses the non-doxastic attitude of hope towards p.

• Premise 2: Having hope that p is rationally incompatible with taking a fictionalist stance towards p.

• Conclusion*: An agnostic with an ND1-attitude towards p, who lacks overriding reasons not to hope p, cannot rationally take a fictionalist stance towards p.

The difference between this argument and the original is that it allows some pro-religious agnostics to take up fictionalism, but only those for whom a non-doxastic engagement is not available. The pro-religious agnostic without overriding reasons cannot rationally chose fictionalism. This is an important result, since it shows that far from being threatened by a collapse into fictionalism, non-doxasticism makes fictionalism rationally unavailable for all those to whom the approach is open.

Concluding remarks

Of the two belief-less approaches to religion – fictionalism and non-doxasticism – non-doxasticism seems the obvious or even only choice for all agnostics to whom it is available. The advantages of the non-doxastic approach are numerous, and if my argument for exclusive availability stands, it is the only form of belief-less religion that is rationally available to the agnostic with a pro-religious attitude.

These results indicate that rather than being viewed as competitors, the two approaches might fruitfully be seen as speaking to different subjects: fictionalism to atheists and agnostics who are either religiously indifferent or have some overriding reason against forming a pro-religious attitude, and non-doxasticism to the agnostic with such a pro-attitude. Two points are worth noticing.

First, there are many religious views, and a typical epistemic subject will presumably be agnostic towards some and atheist (or some other term indicating disbelief if the religion in question is not god-centred) towards others. This means that some religious views will be open to a non-doxastic engagement, while regarding others the subject will be limited to fictionalism. For example, you might be an agnostic regarding Abrahamic theism and an atheist concerning the existence of the Old Norse gods, in which case you can be a non-doxastic catholic but only a fictionalist neo-pagan.

One can also take different stances towards different parts of the same religion. A Christian who believes in the existence of God might still lack outright belief that Jesus is God incarnate, only seeing it as an epistemic possibility and therefore taking a non-doxastic attitude towards the divinity of Christ. Presumably, she might also view the more mythological parts of her tradition, like the existence of angels or hell, as useful fictions.

Second, exactly how the notion of epistemic possibility required by non-doxasticism's ND2-condition is spelled out will have a significant impact on who can adopt non-doxasticism and who is limited to fictionalism. If epistemic possibility is understood in the widest possible sense, you have to be absolutely certain that something is false to be limited to fictionalism (in which case perhaps even neo-paganism might be open for a non-doxastic engagement). If epistemic possibility is understood in a more restricted sense, requiring perhaps a non-negligible chance of truth, the options available for a non-doxastic commitment will decrease significantly. In either case, the point remains that for all agnostics to whom it is available, non-doxasticism makes fictionalism redundant.