Next Article in Journal
The Impact of Religion and National Origin on Attitudes towards Refugee Rights: An International Comparative Empirical Study
Next Article in Special Issue
A Non-Doxastic Fear of Hell: On the Impact of Negative Factors for an Agnostic Religious Commitment
Previous Article in Journal
Hostility toward Gender in Catholic and Political Right-Wing Movements
Previous Article in Special Issue
Eric Perl’s Theophanism: An Option for Agnostics?
Open AccessArticle

Meaning Agnosticism and Pragmatism

Faculty of Theology, University of Helsinki, FI-00014 Helsinki, Finland
Religions 2020, 11(6), 302;
Received: 11 May 2020 / Revised: 11 June 2020 / Accepted: 16 June 2020 / Published: 22 June 2020
(This article belongs to the Special Issue Agnosticism)


This paper proposes to reconsider agnosticism by taking a step onto a meta-level, investigating agnosticism not as an epistemic stance regarding the choice between theism and atheism, but as a stance toward the question concerning the cognitive meaningfulness and/or truth-aptness of religious discourse. It is argued that this “meta-level” meaning agnosticism may actually be an attractive articulation of a certain kind of religious attitude. While pragmatists like William James have claimed that (epistemic) agnosticism practically collapses into atheism, meaning agnosticism at the meta-level can in fact be a pragmatist position focusing on our human condition and its limits. Additional issues, such as the relations between agnosticism and the theodicism vs. antitheodicism debate regarding the problem of evil and suffering, are also briefly examined.
Keywords: agnosticism; theism; atheism; religious language; truth; meaning; pragmatism agnosticism; theism; atheism; religious language; truth; meaning; pragmatism

1. Introduction

Agnosticism is usually understood as an epistemic position. In the philosophy of religion, agnosticism is typically characterized as a view that avoids taking any firm stand in the metaphysical and theological debate between theism and atheism by maintaining that we do not, or cannot, know—or that we do not, or cannot, justifiably believe—anything regarding God’s existence or non-existence. There is, according to the agnostic, no way for us to rationally decide the matter either way. We are not, and perhaps can never be, justified in believing that we could finally settle the theism vs. atheism dispute in one direction or the other.1 The agnostic may add to this basic formulation some plausible explanation for this lack of knowledge and/or epistemic justification; for instance, the spatiotemporally restricted human cognitive apparatus may be constitutively incapable of reaching any justified beliefs about the transcendent—or, conversely, about there being no transcendence at all.
Thus formulated, agnosticism may be taken to be a form of skepticism about God’s existence, analogous to, say, external world skepticism, which sets into doubt our capabilities of knowing anything about the existence of a reality external to our mind and thought (without explicitly denying the existence of such a reality). The relation between agnosticism and skepticism would obviously have to be spelled out in much more detail, but that is not an aim of the present inquiry. Here, it is sufficient to note that if agnosticism is understood as a form of skepticism, it is a local variant of skepticism, focusing on our knowledge about God (and, possibly, by extension other religiously relevant topics, such as the reality of afterlife), rather than our knowledge in general. Moreover, it must not be confused with what is known as “skeptical theism”, because the skeptical theist does not subscribe to skepticism regarding their theistic commitment but only regarding to, e.g., divine hiddenness, or regarding how the theistic God can be considered to morally justify the reality of evil and suffering (that is, how God might, for all we know, have good moral reasons to allow apparently innocent suffering). The skeptical theist may be locally agnostic about the way in which the problem of evil, for instance, ought to be solved, but this does not entail agnosticism about the reality of God; therefore, skeptical theists are not agnostics about God’s existence.
Both theists and atheists may (and often do) maintain that agnosticism is a cowardly and unstable middle-ground position not worthy of serious consideration (for a critical analysis of this line of argument, see (Yoder 2013)). One should bravely embrace either theism or atheism, they might argue, instead of suspending the judgment. I will try to show, in contrast to such accusations, that agnosticism deserves much more serious consideration than it often gets. However, the kind of agnosticism I am primarily interested in here is not the standard epistemic kind.
In this paper, I will suggest that we should, in contrast to received epistemic accounts of agnosticism as a non-committal position between theism and atheism, apply the general idea of agnosticism to a somewhat different issue in the philosophy of religion. I will at a general level propose that we may seriously consider meta-level variants of agnosticism. Such forms of agnosticism are not placed on the scale between theism and atheism at all, but they are rather applied to the very availability of that scale for our epistemic consideration. More specifically, a form of agnosticism directed to the question concerning the cognitive meaning of religious language will be investigated (and, with some qualifications, defended) below. It would be impossible to provide a comprehensive argument in a brief paper for the meta-level meaning agnosticism I will describe here; my main intention is to put this view forward as a candidate for serious consideration. I will also briefly explore its connections with pragmatism, partly because some of the best-known arguments against agnosticism have been presented by one of the greatest pragmatists, William James, and partly because I believe that at the meta-level the kind of agnosticism I am (only carefully and with qualifications) recommending can, and should, be interpreted as a pragmatist position. In a more detailed discussion, this proposal would of course have to be connected with an analysis of pragmatist philosophy of religion more generally (see, e.g., Pihlström 2013, 2020).
In principle, one could be “locally” agnostic not just about the theism vs. atheism debate but also about various other issues in the philosophy of religion, such as, say, the evidentialism vs. fideism dispute in the epistemology of religious belief, or the variety of approaches that have been proposed to deal with the problem of evil. In these cases, the agnostic would claim that we do not, or cannot, know how to resolve those debates either way. One might even be an agnostic about the opposition between theodicism and antitheodicism with regard to the problem of evil and suffering (to be revisited toward the end of this essay).2 I will in the following first focus on what I propose to call meaning agnosticism and will return to the problem of evil and suffering only later. I will also show how a pragmatist approach in the philosophy of religion should take agnosticism more seriously than pragmatists (including the most important classical pragmatist philosopher of religion, James) have usually done.
Standard formulations of agnosticism as an epistemic position start from the assumption that it is meaningful to claim God to exist or fail to exist. The statement, “God exists” (or “There is [a] God”), is either true or false, depending on the way the world is, that is, depending on whether God actually exists or not. This statement thus is purportedly factual, and we can judge its epistemic credentials by considering what kind of evidence can be provided pro et contra the statement and/or the proposition it expresses (or the belief that that proposition is true). Let me now suggest meta-level formulations of agnosticism that transfer the debate over agnosticism and non-agnostic positions to the context of the issue concerning the cognitive and/or factual meaningfulness of religious language—formulations, that is, that do not just uncritically presuppose that the statement about God’s existence is determinately either true or false independently of us and our practice-embedded language-use.

2. Meta-Level Meaning Agnosticisms

Most of the different rival views in the philosophy of religion today presuppose that it makes sense, cognitively, to speak about God and God’s existence (and, mutatis mutandis, about other religiously relevant matters, such as the reality of afterlife). The strengthening of this assumption, of course, marks a major turn in the philosophy of religion after the mid-1900s.3 While early analytic philosophers found religious language the main problem in the field, more recent philosophers exploring issues in religion and theology hardly doubt the cognitive significance of our statements about God but are happy to move straightforwardly into considerations about the rationality (vs. irrationality) of maintaining, say, that God exists. In the simple case of theism vs. atheism, the theist indeed believes that God exists, while the atheist denies this. Both, in any event, presuppose that the statement, “God exists” (or “God is real”, or some other equivalent), makes sense, i.e., that it is cognitively meaningful to make such a claim. In terms of standard bivalent truth-conditional semantics, this presupposition roughly means that “God exists” is either true or false depending on whether its truth-condition, the state of affairs of God’s existing, obtains in the mind- and language-independent world.
Another way of putting the matter is to say that both theists and atheists are committed to a certain kind of realism regarding religion and theology (Pihlström 2020). They conceive religious discourse about God as truth-apt: the concept of truth can be applied to religious and/or theological statements about the divinity (as well as, again analogously, other religiously relevant elements of reality, such as, perhaps, the soul and its postmortem existence), and therefore such discourse is to be treated as (in principle) cognitively meaningful.
An antirealist and/or non-cognitivist about religious discourse denies realism and/or cognitivism, maintaining that religious expressions such as “God exists” are not properly speaking factual statements at all. They are not in the business of stating facts that would obtain in the world if the statement were true. They are not made true (or false) by anything in the language-independent world. They are, rather, something else: for example, expressions of emotions, or perhaps implicit prescriptions recommending a certain way of life or expressing one’s participation in certain ritual practices.
This line, or something close to it, was adopted both by the logical positivists (who were mostly harshly critical of religion, rejecting religion and theology precisely due to the alleged cognitive meaninglessness of religious discourse) as well as early analytic philosophers of religion such as Richard Braithwaite (Braithwaite [1955] 1994) and (in his distinctive way) John Wisdom (1944). There is also a sense in which a version of this non-cognitivist approach was continued in the Wittgensteinian tradition, even though Ludwig Wittgenstein himself cannot be clearly classified either as a non-cognitivist or as a representative of any other “ism”, for that matter. Incidentally, it can be claimed that few Wittgensteinian philosophers of religion have simply suggested that religious discourse lacks cognitive meaning; they have, rather, maintained that different language-games establish different criteria of meaning, and the kind of meanings that a properly religious use of expressions like “God”—or even “exists”—may have in religious contexts must not be conflated with non-religious, e.g., scientific, usages of such expressions.
Now, what I wish to suggest here is that we may apply the general and intuitive idea of agnosticism to the debate over the cognitive meaningfulness of religious statements. Instead of either affirming (with cognitivists and realists finding religious discourse truth-apt and purportedly factual) or denying (with non-cognitivists and non-realists finding such discourse lacking in cognitive meaningfulness and not being in the business of stating facts at all) this cognitive and factual meaningfulness, or the truth-aptness of religious discourse, the agnostic at the meta-level maintains that we do not, or cannot, know whether religious discourse is cognitively meaningful, purportedly factual, and/or truth-apt. Or at least there is no way we could rationally resolve the matter either way.
We should, in other words, remain “meaning skeptics” about religious discourse. We cannot, when engaging in religious discourse or when examining it philosophically (or theologically), know, or perhaps even justifiably believe, that it makes sense, cognitively speaking; nor, however, can we know that it does not. Admittedly, it then becomes problematic for us to actually view ourselves (with full confidence) as participants in religious discourse—but this is, in the end, what renders agnosticism a potentially pragmatically valuable position: it makes sense of our participating in religious discourse without full cognitive access to the ways we are, or are not, able to “make sense” of our own language-use within such participation. This position may be regarded as a species of meaning skepticism, but on the other hand it is definitely not a skeptical view in the sense that our being able to know that we can, or cannot, make sense by using religious language would be categorically denied. On the contrary, our adopting an agnostic stance enables us to occupy an open space of philosophical (and, possibly, theological) reflection that is better equipped to acknowledge the distinctive character of religious ways of using language than any dogmatic commitment to either realism/cognitivism or its denial ever could. I will get back to this meta-level significance of agnosticism toward the end of the paper.
There are some obvious attractions in an agnostic view concerning the cognitive and/or factual meaningfulness of religious statements. One would, for instance, be able to avoid taking any firm stand in the debates between realists and non-realists.4 I suppose many of us may find it very difficult to arrive at any clear conclusion here, so agnosticism might, just as in the theism vs. atheism case, prima facie be a way of maintaining a critical middle path, keeping the issue open—without just leaving it open in the sense of failing to care about pursuing it further. Moreover, just as the meta-level agnostic need not settle the realism vs. antirealism issue one way or the other, s/he presumably will not have to settle the evidentialism vs. fideism issue, either. Nor will the agnostic have to worry too much about the fact that realism about religious discourse (understood as truth-apt) may seem to entail a form of religious exclusivism, which we may be ethically and politically motivated to criticize quite independently of the realism controversy.5 Agnosticism may thus help us get rid of a wide variety of troubling and unwelcome philosophical complications—in this case at the meta-level rather than at the “first-order” level of the theism vs. atheism dispute.
Even so, the kind of agnosticism I am (tentatively and with qualifications) proposing for serious consideration here is an attitude of active critical inquiry rather than an attitude of mere passive and uncritical relaxation. We can never be sure that we can continue to rest content with our decision to avoid taking a stand. Agnosticism definitely does not mean that we should cease to inquire into the issue about which we are (at the moment) agnostic (cf. again Draper 2001, 2017).

3. Non-Cognitivism about Religious Language: Braithwaite’s Proposal

In order to concretize the suggestion to develop a meta-level meaning agnosticism regarding religious language, let us consider the view propounded by one of the very few logical empiricists who actually sympathized with religious ideas, Richard Braithwaite. His 1955 lecture, “An Empiricist’s View of the Nature of Religious Belief”, deserves a brief exposition, as it explicitly takes up the issue concerning the cognitive content of religious language.
Braithwaite points out that the philosophy of religion at the time of his writing had made decisive progress precisely by turning its attention to the nature of religious and theological language: “until recently the emphasis has been upon the question of the truth or the reasonableness of religious beliefs rather than upon the logically prior question as to the meaning of the statements expressing the beliefs”—“as if we all knew what was meant by the statement that a personal God created the world” (Braithwaite [1955] 1994, p. 3). It seems to me that this is precisely what contemporary analytic philosophy of religion seeks to do, having turned back to questions of truth and reasonableness and leaving behind the “logically prior question” concerning meaning, and therefore a healthy dose of meta-level agnosticism might in my view serve the field rather well.
Just as Wisdom (1944) had suggested in his famous essay, “Gods”, Braithwaite argues that the hypothesis that God exists cannot be empirically refuted. He takes it as obvious that “most educated believers” do not think of God as being “detectable” and “hence do not think of theological propositions as explanations of facts in the world of nature in the way in which established scientific hypotheses are” (Braithwaite [1955] 1994, p. 6). This is in striking contrast with, for example, the current science vs. religion debates, in which various cosmological explanatory attempts to use theism precisely in that way seem to be the rule rather than an exception. However, Braithwaite moves from this critical account of the cognitively nonsensical character of religious statements to a comparison between religious and moral utterances. Referring to Wittgenstein’s Philosophical Investigations—at that time a recently published posthumous work—he appeals to the “use principle” of meaning, according to which “the meaning of any statement is given by the way in which it is used”, and reminds us that moral statements “have a use in guiding conduct” (ibid., p. 10). The problem of religious belief therefore becomes an empirical problem concerning the use of religious statements to express religious convictions (ibid., p. 11).6
Braithwaite does not subscribe to the standard logical positivist account of moral language as “emotive”; rather, he finds it “conative” and views moral assertions as commitments to policies of action (ibid., pp. 12–13). It is through this analogy to moral statements that he arrives at his solution to the problem of the nature of religious belief:
That the way of life led by the believer is highly relevant to the sincerity of his religious conviction has been insisted upon by all the moral religions, above all, perhaps, by Christianity. ‘By their fruits ye shall know them.’ The view which I put forward for your consideration is that the intention of a Christian to follow a Christian way of life is not only the criterion for the sincerity of his belief in the assertions of Christianity; it is the criterion for the meaningfulness of his assertions. Just as the meaning of a moral assertion is given by its use in expressing the asserter’s intention to act, so far as in him lies, in accordance with the moral principle involved, so the meaning of a religious assertion is given by its use in expressing the asserter’s intention to follow a specified policy of behaviour. […] it is the intention to behave which constitutes what is known as religious conviction.
(Ibid., pp. 15–16.)
This can be read as a commitment to a certain kind of pragmatism about religious belief. Jamesian pragmatists, after all, have also repeatedly quoted the Biblical phrase, “by their fruits ye shall know them”. Indeed, it is no wonder that pragmatist philosophy of religion is often neglected by the analytic mainstream today: it can be taken to be basically just something like logical empiricism, only less formal and, hence, less clear. Conversely, however, the pragmatist seriously interested in the nature of religious language (rather than the epistemic status of religious statements) may find valuable insights in Braithwaite’s analysis.
Indeed, Braithwaite seems to subscribe to a pragmatic holism not very different from the one famously espoused by W.V. Quine and less famously (also in ethics and, in principle, religion) by Morton White.7 Just as scientific hypotheses need to be understood not in isolation from each other but in relation to the “whole system of hypotheses”, we also need to consider “a system of religious assertions as a whole” and the way this whole system is “used” (ibid., p. 17). This leads him to perceive a very intimate connection between religious and moral principles—both are ultimately principles of conduct (ibid., p. 18). Indeed, the following can be seen as a formulation of the pragmatist maxim or principle better known from pragmatist classics like Charles S. Peirce and William James: “The way to find out what are the intentions embodied in a set of religious assertions, and hence what is the meaning of the assertions, is by discovering what principles of conduct the asserter takes the assertions to involve.” (Ibid.)8
Toward the end of his essay, Braithwaite arrives at his characterization of what religious assertions are: “A religious assertion, for me, is the assertion of an intention to carry out a certain behaviour policy, subsumable under a sufficiently general principle to be a moral one, together with the implicit or explicit statement, but not the assertion, of certain stories.” (Ibid., p. 32) Like moral beliefs, understood as intentions to behave in a certain way, religious beliefs are such (moral) intentions to behave, “together with the entertainment of certain stories associated with the intention in the mind of the believer” (ibid., pp. 32–33). This also, in his view, makes religious beliefs open to reasonable critical discussion, as the practical consequences of behavioral principles can obviously be rationally discussed and compared (ibid., p. 34).
I have not quoted at length from Braithwaite’s examination of religious belief in order to endorse his views, which I admit remain too closely stuck with the strict empiricist criteria of meaningfulness set by the logical empiricists. I do think, however, that Braithwaite’s assimilation of religious beliefs to moral principles deserves much more serious attention than most contemporary analytic philosophers of religion would be prepared to grant to it. Among other things, Braithwaite’s analysis carries certain extremely important themes from Kant (religion can rationally and legitimately only be based on morality) and classical pragmatism (the meaning of a concept or statement is ultimately based on its potential practical outcome), as well as the later Wittgenstein (“meaning is use”) to the philosophy of religion—via logical empiricism, to be sure, but without succumbing to the scientistic repudiation of religion and theology that movement is famous for.9
The purpose of my discussion of Braithwaite should now be clear. What I am proposing is not that we endorse the view formulated by him but that we consider adopting an agnostic stance to the question concerning the rational justifiability of the kind of position exemplified by his discussion, in contrast to the realist-cognitivist one. We are well advised to suspend judgment regarding the rational decidability of this meta-level linguistic issue either way—though I should add that consistently with this agnostic attitude I do not wish to make any absolute claims about our need to endorse agnosticism itself, either. This meta-level agnosticism, at any rate, provides us with an open space for critical inquiry into the problems and benefits of both realism/cognitivism and antirealism/non-cognitivism concerning religious and theological language. We may, as agnostics, find a view like Braithwaite’s to be available to us in a sense different from the rather dogmatic neglect of such positions in the metaphysical mainstream of analytic philosophy of religion today.

4. The Pragmatic Meaning of Meaning Agnosticism

Agnosticism has been criticized for many reasons—for example, for the reason that it allegedly fails to care about continuing the inquiry into the matter that is left open—but here it is particularly interesting to note that it has been taken to collapse to atheism. This is, at least from a pragmatist perspective, a much more interesting criticism, as agnosticism in my view clearly does not have to entail any uncritical dismissal of inquiry but can on the contrary emphasize the need to keep the road of inquiry open and search for new evidence.10
In William James’s “The Will to Believe” (1897), agnosticism figures as an inherently unstable position: if you postpone your decision concerning the “live option” of whether or not you should commit yourself to the theistic belief, suspending judgment in the hope of acquiring more reliable evidence for either theism or atheism, then you will, James argues, practically speaking live the life of an atheist, failing to take the practical step of believing in God. Thus, agnosticism is pragmatically equivalent to atheism. This could also be regarded as an application of the pragmatic method, as spelled out in James’s Pragmatism (James [1907] 1975), in particular.
This is how James famously formulates his key thesis in “The Will to Believe”: “Our passional nature not only lawfully may, but must, decide an option between propositions, whenever it is a genuine option that cannot by its nature be decided on intellectual grounds; for to say, under such circumstances, ‘Do not decide, but leave the question open,’ is itself a passional decision—just like deciding yes or no—and is attended with the same risk of losing the truth.” (James [1897] 1979, p. 20; original emphasis.) A few pages earlier he has told us what makes an option “genuine” (ibid., pp. 14–15): it has to be (i) “live” in the sense that both of the rival hypotheses are “live” ones for the subject; (ii) “forced”, i.e., the decision cannot be avoided, as we cannot refuse to face the issue and there is no “third way”; as well as (iii) “momentous”, that is, a unique, highly significant, non-trivial chance. Crucially, suspending judgment—that is, remaining agnostic—may be equally risky as believing, or refusing to believe, both epistemically risky in the sense of possibly losing the truth and ethically or existentially risky in the sense of possibly losing a way of life that could be deeply meaningful. These dimensions of risk are, however, entangled, as an “option” is understood by James as a choice between rival hypotheses—a choice in our lives that has both epistemic and ethical dimensions. Even if agnosticism were a reasonable choice “in theory”, James seems to think that in our practical lives it will not take us anywhere, at least any further than atheism. (See further Pihlström 2013.)
However, it may now be argued that this pragmatic reduction of agnosticism to atheism does not similarly work at the meta-level. Even though we may (I think) agree with James that agnosticism—when seen as a possible position within the theism vs. atheism debate—is indeed unstable and at least runs the risk of collapsing to atheism (or perhaps even directly so collapses), and that therefore from a pragmatist point of view there is no clearly discernible difference between atheist and agnostic positions regarding God’s existence, this conclusion may not hold at the meta-level. That is, a meta-level agnosticism about the cognitive and factual meaningfulness of religious belief, for example, an agnosticism regarding the question whether to embrace Braithwaite’s position in contrast to the more recent mainstream analytic philosophers’ of religion firm commitment to realism, may not be similarly reducible to either non-cognitivism (non-realism) or cognitivism (realism). This is at least partly because the “will to believe” argument in the Jamesian style does not similarly apply at the meta-level, as our “live” weltanschaulichen options hardly include abstract meta-disputes such as the one between cognitivism and non-cognitivism.11
At least I cannot find in Jamesian pragmatism any obvious reason to reject the kind of meaning agnosticism sketched above. On the contrary, agnosticism, as suggested in Section 2 and Section 3, may enjoy several interesting benefits. Instead of hindering our pragmatic commitment to critical inquiry into how exactly we in our religious or non-religious language-use live amidst the questions concerning whether our expressions make sense, and (if they do) what kind of sense, agnosticism, when adopted at the meta-level, may actually be a pragmatic attitude of the critical inquirer who needs to keep the matter open precisely in order to continue investigation.
The kind of meaning agnosticism I have tentatively proposed to be considered may, hence, even have the benefit of serving some humanly vitally important pragmatic purposes and being therefore pragmatically highly significant. This is particularly because it may be taken to play a very important role in acknowledging the kind of existential unease we may rather naturally experience regarding our religious (and non-religious) views about the world. We may, simply, be sincerely puzzled about whether such views, or our attempts to express them in religious language, make sense at all or not. Meaning agnosticism can be regarded as an attempt to acknowledge (some) religious believers’ (the possible inappropriateness of the word “believer” here notwithstanding) puzzlement about whether they are so much as able to make sense by using their religious expressions. They may not be willing to simply deny such sense-making, but it would also be a violation of the depth of their religious convictions to maintain that their religious language-use simply unproblematically makes factual and/or cognitive sense, as that would bring it on a par with everyday or scientific discourse about tables, trees, and viruses—which could even be regarded as banalizing or even blasphemous (depending on the character of the religious outlook at the background of the meta-level agnostic position). Therefore, there can even be theological and/or religious reasons for meaning agnosticism. But there may also be humanistic reasons: agnosticism can, precisely due to its refusal to take a determinate stand on whether our religious language “makes sense”, seek to make sense—at a meta-level—of our human condition itself and the inevitable fragility of its (our) sense-making activities.12
It may, then, be pragmatically extremely important for us—it may play an existential role in our lives—not to take a firm stand on whether our statements are cognitively meaningful or not when we utter religious expressions. Therefore, views such as Braithwaite’s need to be kept on the agenda of philosophers of religion even today. And, therefore, agnosticism is a meta-level stance worth maintaining, at least tentatively. It arguably plays a vital pragmatic role in our on-going consideration of how to respond to the challenge of speaking about matters that religiously matter. (Or, to put it more mildly—more agnostically—we cannot be certain that it does not play such a role.)

5. Antitheodicism and Agnosticism as Humanism

I have argued that focusing on religious language enables and motivates a deeper, more interesting and more exciting form of agnosticism—an agnosticism about the meaningfulness of religious language—than the relatively straightforward agnosticism that simply advocates epistemic neutrality or suspension of judgment regarding the standard apologetic opposition between theism and atheism. Agnosticism about religious meaning, as inherently non-apologetic, may take us deep into the problem concerning what religious language-use is and hence into a reflection on what it means, or could mean, to live one’s life as a believer engaging in such language-use (or, conversely, what it means, pragmatically, not to live such a life). It is, in short, a form of agnosticism that enables us to avoid the apologetic alternative between pro- and anti-religious metaphysical worldviews.
One may even suggest that the meta-level form of agnosticism I have recommended is not only the attitude of pragmatic critical inquiry but also a more humanistic form of agnosticism than the typical epistemic types of agnosticism (though not “humanistic” in the sense of being opposed to religion): focusing on religious language, and especially on the thin line separating the critical views propounded by, e.g., the logical positivists from the more religious-friendly views espoused by thinkers like Wisdom and Braithwaite (and, of course, the Wittgensteinians), entails focusing on the role played by this distinctive and problematic discourse in our human practices and forms of life. This is a focus shared by pragmatism, too, though not as narrowly confined in linguistic analysis. In contrast, the theism vs. atheism controversy today often remains stuck in the context of an endless debate between something like natural theology and its atheist rejection, with cosmological and design arguments (in their classical and updated versions) flourishing among those willing to defend religious belief. Such a metaphysical orientation is, as its practitioners very well realize, a return to something pre-Kantian. The foundational insight of early analytic philosophy—without specific references to Kant—is that we have to start examining religion and its place in our lives and worldviews by first critically investigating how we structure our world in terms of religious discourse. This approach is shared by pragmatism as well, and in these terms both Kant’s critical philosophy and pragmatism may be seen as sharing a meta-level agnosticism—even though developing this idea in scholarly detail would, I suppose, require a book-length treatise instead of a single article.
Various further applications of meta-level agnosticism could also be proposed. They might include agnosticism about theodicies and antitheodicies, for instance. I have argued on a number of earlier occasions (e.g., Pihlström 2014, 2020; Kivistö and Pihlström 2016) that we should embrace a thoroughgoing antitheodicist position for ethical reasons, in order to avoid instrumentalizing other human beings’ meaningless suffering—that is, rendering it allegedly meaningful—in the service of some “overall good”. But I have also argued that theodicism typically threatens to come back even in an antitheodicist context, because even when formulating antitheodicies in contrast to theodicies we seek to view the world harmoniously and “rightly”, thus seeing some “higher meaning”—at the meta-level—in our conceptualization of suffering, after all (see Pihlström forthcoming). At a further meta-level, it may therefore be advisable to suspend any final judgment regarding the ultimate philosophical success of antitheodicism. This agnostic attitude can even play a role in the attempt to acknowledge the puzzlement about our meaning-making capabilities suggested above. It is religiously and theologically relevant for our forms of life or religious practices to maintain that we, as both epistemically and ethically limited and fragile human beings, are not in the position to rationally determine whether we can attach cognitive meaningfulness to our religious language-use, and as an element of this suspension of judgment we may maintain that we cannot ultimately settle the philosophical and theological issues concerning meaningless suffering and evil.
This “dialectics of antitheodicism”, as I have proposed to label it (cf. ibid.), should in my view be taken very seriously, and one way of formulating this view is precisely by applying agnosticism at the meta-level to the resolvability of the problem of evil and suffering. I can still be firmly committed to promoting antitheodicism for pragmatic and ethical reasons while at the same time maintaining intellectual humility by acknowledging that there is no way I can argumentatively or theoretically conclusively resolve this issue either way. Its theoretical irresolvability can indeed be an element of its being a pragmatic and essentially ethical, or perhaps rather existential issue: our taking a stand on suffering defines, in practical terms, who we are.
I am, as indicated above, proposing to regard this kind of agnosticism as a form of humanism—but not, of course, humanism in the sense of disbelief. It is a key idea in meta-level meaning agnosticism that we take seriously our fragile and insecure human predicament, our inability to be fully confident about our abilities to make sense by participating in the (religious) discourses we participate in, as well as the more specific discourses within those discourses, such as the theodicism vs. antitheodicism discourse. My references to “humanistic agnosticism” must not therefore be confused with the kind of humanism that merely epistemic forms of agnosticism—in denying our ability to know anything about God’s existence—may claim to be committed to.

6. Transcendental Agnosticism

Agnosticism in its meta-level form should, as already hinted above, be seen as a chapter of Kantian-inspired critical philosophy, even though no proper discussion of Kant is possible in this essay. Meaning agnosticism is, I would now like to suggest, also transcendental in the sense of addressing and exploring the necessary conditions for the possibility—and therefore also the limits—of religious language-use. (Analogously, the humanism that meta-level agnosticism embraces can itself be regarded as transcendental in this sense derived from Kant—but that is another, and much longer, story.) An even more thoroughgoing transcendental agnostic claims that only a meta-level meaning agnosticism enables our engagement in religious thought and language-use, because religious discourse is, qua religious, unavoidably troubled by its own ability to make sense. It would, according to such agnosticism, be pseudo-religious (cf. Pihlström 2013) to simply presuppose that religious language makes sense, or is “truth-apt”, in a cognitivist and realist manner.13 An unclarity about the ability of religious discourse to make cognitive sense would be regarded as constitutive of such discourse as the kind of discourse it is. Thus understood, meta-level transcendental meaning agnosticism would come close not only to certain Wittgensteinian developments in the philosophy of religion but also some “postmodern” views that reject philosophical theism as an adequate account of religious faith.14
Agnosticism, in this meta-level shape focusing on the conditions of meaning, would however also be transcendentally entangled with a kind of humanism, as already suggested above. Continuous puzzlement at the limits of religious meaningfulness—of our capacities of engaging in meaningful discourse within a religious form of life—could be taken to characterize the human condition, at least for believers who are critically self-reflective about their faith. Moreover, even the non-religious can be deeply puzzled about human language and its limits of sense more or less in the same sense; Wittgenstein’s famous metaphor of language as a “cage” (e.g., in his 1929 “Lecture on Ethics”, available in (Wittgenstein 1993)) easily comes to mind here.
However, when speaking about, and even defending, a meta-level form of meaning agnosticism, or even transcendental agnosticism, I am not assuming that the distinction between “first-order” epistemic agnosticism and “second-order” or meta-level agnosticism would always be crystal clear. Some philosophers might suggest that it makes little sense to speak about agnosticism at the meta-level: from a Carnapian point of view, for instance, questions concerning the choice of a linguistic framework, as distinguished from existence questions within such a framework, are not cognitive questions at all. Thus, for someone like Rudolf Carnap it would hardly make sense to be agnostic about religious meaning: questions of meaning must be settled first as, ultimately, matters of pragmatic choice, and only then can cognitive issues concerning our postulation of entities within a chosen framework arise (cf. Carnap 1950).
However, my proposal in this paper is “post-Quinean” (rather than Carnapian) in the sense that I admit that our way of drawing the distinction between the different levels—the “first-order” level of epistemic concerns and the meta-level concerning meaning—is itself contextualized in terms of our previous and on-going ways of drawing such distinctions and operating within the contexts defined by them. Our conception of what we can meaningfully say depends on what we take the world to be like—and this, of course, could also be regarded as one of the insights of the later Wittgenstein.15 The semantic thus partly depends on the epistemic, and vice versa, to the extent that we may (borrowing a phrase sometimes used in Quinean contexts) talk about a “reciprocal containment” of epistemic agnosticism and (meta-level) linguistic meaning agnosticism. Our meta-level agnosticism about religious language may thus also partly depend on our metaphysical puzzlement about the world we take ourselves to be living in, including puzzlement about evil, suffering, and God. This idea also functions well in the context of Morton White’s (1956, 2002) pragmatic holism cited earlier.
Let me note, however, that I am not claiming that the meta-level meaning agnostic has to employ the Kantian transcendental vocabulary in articulating their agnosticism. This vocabulary is optional. Even so, the focus of agnosticism on questions concerning the conditions and limits of meaning is strongly emphasized in the kind of meta-level agnosticism I have formulated.

7. Conclusions: Agnostic Humanism, Hope, and Truth

I have in this essay only discussed selected examples of meta-level agnosticism—primarily agnosticism about religious meaning and in Section 5 also, albeit very briefly, agnosticism about the resolvability of the theodicism vs. antitheodicism controversy concerning the problem of evil and suffering. I have also suggested that this kind of agnosticism is “humanistic” in the sense of seeking a plausible account of the human predicament.16 These are clearly central and relevant examples in the philosophy of religion, but here they have primarily served as, indeed, mere examples. The main point of my investigation is the proposal that the concept of agnosticism deserves to be articulated and further developed considerably beyond its initial and obvious epistemic context of employment.
Even in the epistemic context there would still be much further reflection to do. For example, it might be interesting to examine in what sense exactly the agnostic may still hope that God exists. Hope in this sense may be adequate as a religious attitude weaker than full belief, as has been proposed by, e.g., Simo Knuuttila.17 At a one step higher meta-level, it could be investigated whether it would be sufficient for a religious agnostic to merely hope to be able to believe that God exists. Such an agnostic would not believe that God exists, nor necessarily even hope that He does, but they could hope to be able to so believe (or, of course, to so hope). In principle, an indefinite number of nested doxastic operators could be added to account for ever more complex cases of iterated states of belief and hope.
Furthermore, a Jamesian pragmatist may ask whether one’s hope to be able to believe might actually—at least in some special cases of human life—yield an active will to believe (cf. again James [1897] 1979). In principle, one can remain agnostic about whether one has good—or even morally acceptable—reasons to hope that God exists, or again to hope (or will) to be able to believe that God exists. Such an agnosticism about hope could express a deeply human puzzlement about what one’s proper religious (or non-religious) response to, say, evil ought to be. For example, one could maintain that given the irresolvability of the problem of evil and suffering, it would be immoral not just to believe in God’s existence but even to hope that God exists.
When this intuitively rather straightforward analysis of the relation between belief and hope is brought up to the meta-level investigated in this essay, the relevant kind of hope will focus on our abilities to make sense (in the factual and/or cognitive sense) of our religious expressions. One might thus hope to be able to utter cognitively meaningful and determinately true or false sentences when engaging in religious discourse—or, conversely, one might (in principle) hope to be able to avoid making literal sense by using such expressions.
I would like to conclude by re-emphasizing the idea that agnosticism—especially in my meta-level transcendental sense—is (as already tentatively suggested) a form of humanism. Far from being a manifestation of intellectual cowardice or disregard for truth, as its opponents often claim, it is an active attitude of investigating what it means to be human in our complex contexts of belief, disbelief, hope, and despair. Agnosticism actually is a way of caring for truth, and hence for our lives and souls, because it is an attempt to walk the thin line between making a commitment to a truth-claim and avoiding such a commitment—always for a definite reason. This caring for truth is a key not only to “first-order” epistemic agnosticism about God’s existence but also to the meta-level meaning agnosticism analyzed above. The latter critically avoids taking a firm stand to the question concerning the truth-aptness of religious discourse—for various reasons, arguably also including the tendency of realism and cognitivism to entail ethically problematic views such as religious exclusivism and theodicism.18 We should not too easily assume that the discourse we engage in is truth-apt (and we may even end up maintaining that a serious commitment to religious discourse transcendentally necessitates our not making such an assumption), but we should at the same time continue critically inquiring into the pros and cons of the assumption that it is.
Moreover, if we take seriously the iteration of doxastic (and voluntarist) attitudes, it is clear that one may also be an agnostic about agnosticism itself. The agnostic, when reflexively self-conscious, need not believe agnosticism to be “true”—whatever that exactly means in this case. They might avoid taking a firm stand even here. Agnosticism can then be defended as a tentative standpoint for critical inquiry, more like a hypothesis that will be further tested, rather than a proposition to be believed to be true. At the meta-level, the agnostic need not believe that it is true that we cannot know whether or not religious discourse is truth-apt. They may remain agnostic about our being able to possess, or come up with, such knowledge. There is, clearly, a certain resemblance between this view and the Pyrrhonic form of skepticism familiar from antiquity—but such a historical comparison is of course impossible in this paper.
Agnosticism, then, considerably broadened from its epistemic version, can be designed to acknowledge our intellectually and existentially fragile human condition and its limits, analogously to the way in which Kantian critical philosophy seeks to acknowledge those limits, criticizing various “inhuman” (especially transcendentally realist) attempts to overstep or neglect such conditions and limits. This humanistic sense of agnosticism can, in my view, be most vigorously affirmed and defended insofar as we engage in a pragmatist rearticulation and possibly a transformation of agnosticism. It is through (meta-level) agnosticism that we may then pragmatically create and recreate our relation to whatever philosophy of religion is about—be it God, religious language, or evil and suffering.


This research received no external funding.


Thanks are due to Francis Jonbäck who invited me to contribute this essay to this special issue on agnosticism, as well as to three anonymous referees for highly useful critical comments.

Conflicts of Interest

The author declares no conflict of interest.


  1. Braithwaite, Richard Bevan. 1994. An Empiricist’s View of the Nature of Religious Belief. In Theory of Games as a Tool for the Moral Philosopher and An Empiricist’s View of the Nature of Religious Belief. Bristol: Thoemmes Press. First published 1955. [Google Scholar]
  2. Carnap, Rudolf. 1950. Empiricism, Semantics and Ontology. Revue Internationale de Philosophie 4: 20–40. [Google Scholar]
  3. Draper, Paul. 2001. Seeking but not Believing: Confessions of a Practicing Agnostic. In Divine Hiddenness: New Essays. Edited by Daniel Howard-Snyder and Paul Moser. Cambridge: Cambridge University Press, pp. 197–214. [Google Scholar]
  4. Draper, Paul. 2017. Atheism and Agnosticism. In Stanford Encyclopedia of Philosophy. Edited by Edward Zalta. Available online: (accessed on 10 March 2020).
  5. Hasker, William. 2005. Analytic Philosophy of Religion. In The Oxford Handbook of the Philosophy of Religion. Edited by William J. Wainwright. Oxford: Oxford University Press, pp. 421–46. [Google Scholar]
  6. James, William. 1979. The Will to Believe and Other Essays in Popular Philosophy. Edited by F. H. Burkhardt, F. Bowers and I. K. Skrupskelis. Cambridge and London: Harvard University Press. First published 1897. [Google Scholar]
  7. James, William. 1975. Pragmatism: A New Name for Some Old Ways of Thinking. Edited by F. H. Burkhardt, F. Bowers and I. K. Skrupskelis. Cambridge and London: Harvard University Press. First published 1970. [Google Scholar]
  8. Kearney, Richard. 2010. Anatheism. New York: Columbia University Press. [Google Scholar]
  9. Kivistö, Sari, and Sami Pihlström. 2016. Kantian Antitheodicy: Philosophical and Literary Varieties. Basingstoke: Palgrave Macmillan. [Google Scholar]
  10. Martin, Michael. 1990. Atheism: A Philosophical Justification. Philadelphia: Temple University Press. [Google Scholar]
  11. Pihlström, Sami. 2013. Pragmatic Pluralism and the Problem of God. New York: Fordham University Press. [Google Scholar]
  12. Pihlström, Sami. 2014. Taking Evil Seriously. Basingstoke: Palgrave Macmillan. [Google Scholar]
  13. Pihlström, Sami, ed. 2015. The Bloomsbury Companion to Pragmatism. London and New York: Bloomsbury. [Google Scholar]
  14. Pihlström, Sami. 2020. Pragmatic Realism, Religious Truth, and Antitheodicy: On Viewing the World by Acknowledging the Other. Helsinki: Helsinki University Press. [Google Scholar]
  15. Pihlström, Sami. Forthcoming. Theodicy by Other Means? Rethinking “God after Auschwitz” through the Dialectics of Antitheodicism. Unpublished manuscript.
  16. White, Morton. 1956. Toward Reunion in Philosophy. Cambridge, MA: Harvard University Press. [Google Scholar]
  17. White, Morton. 2002. A Philosophy of Culture: The Scope of Holistic Pragmatism. Princeton: Princeton University Press. [Google Scholar]
  18. Wisdom, John. 1944. Gods. In Proceedings of the Aristotelian Society. Hoboken: Wiley, vol. 45. [Google Scholar]
  19. Wittgenstein, Ludwig. 1953. Philosophical Investigations. Translated by Gertrude Elizabeth Margaret Anscombe. Oxford: Basil Blackwell. [Google Scholar]
  20. Wittgenstein, Ludwig. 1969. On Certainty. Translated by Denis Paul. Oxford: Basil Blackwell. [Google Scholar]
  21. Wittgenstein, Ludwig. 1993. Philosophical Occasions. Edited by James Carl Klagge and Alfred Nordmann. Indianapolis: Hackett. [Google Scholar]
  22. Yoder, Scot D. 2013. Making Space for Agnosticism: A Response to Dawkins and James. American Journal of Theology and Philosophy 34: 135–53. [Google Scholar] [CrossRef]
In a sufficiently broad construal, atheism may also cover at least some forms of agnosticism, if the former simply amounts to the lack of belief in God’s existence (for a comprehensive discussion of the definitions of atheism, see (Martin 1990)). Usually interesting forms of atheism place the doxastic operator differently: the atheist believes that there is no God, rather than just failing to believe that there is. For a lucid discussion of the relations between (various forms of) atheism and agnosticism, see (Draper 2017). Following Draper (among others), it is important to distinguish between agnosticism as a mere psychological state (involving disbelief in either theism or atheism) and agnosticism as a normative commitment to the proposition that neither God’s existence nor non-existence is known. For Draper’s reflections on his own agnosticism, see (Draper 2001).
As I have suggested elsewhere (Kivistö and Pihlström 2016; Pihlström 2020), the opposition between theodicism and antitheodicism should be understood as the choice between those approaches to the problem of evil and suffering that consider it a normative requirement for the discussion to deliver a theodicy and those that do not. Accordingly, both theists maintaining that a theodicy can be delivered and atheists denying this (while expecting theists to deliver one) are theodicists in my sense, while antitheodicists reject this way of setting the dispute in the first place by criticizing (e.g., for ethical reasons) the theodicist expectation or requirement itself.
It might be suggested that the move toward theological realism (and out of antirealist or non-cognitivist views) in the philosophy of religion after mid-1900s is parallel to the emergence of scientific realism in the philosophy of science, as well as the more general re-emergence of realistic metaphysics in analytic philosophy, in particular (cf. Pihlström 2020).
In a more comprehensive examination of realism regarding religion and theology, the distinction between antirealism and nonrealism would also have to be discussed (see again, e.g., Pihlström 2020). Here, I will not dwell on that issue, as realism is not my main topic in this paper.
For the argument that (strong) realism entails exclusivism, according to which at most just one religion is true (either in a straightforwardly propositional sense or in a more theologically pregnant soteriological sense), see again (Pihlström 2020).
Let me note that I am not taking any stand on whether Braithwaite’s account of Wittgenstein’s (1953) “meaning as use” idea is correct (that is, I remain agnostic regarding the accuracy of his reading of Wittgenstein).
On White’s holistic pragmatism, see both his early (White 1956) and late (White 2002) work; cf. also, e.g., (Pihlström 2015).
Braithwaite’s analysis continues by the observation that religious stories play a crucial role in shaping the practical attitudes of conduct that the religious believer is committed to. Such stories do not have to be interpreted as empirically veridical (Braithwaite [1955] 1994, p. 27); rather, the religious person may pragmatically interpret them “in the way which assists him best in carrying out the behaviour policies of his religion” (ibid., p. 29). This may sound too straightforwardly instrumentalist even for a pragmatist’s taste, but the basic idea of entangling religious language-use very closely together with moral conduct is something that Braithwaite shares not only with the pragmatists but also, going further back in history, with Kant, for whom religious concepts like God and the immortal soul were, famously, only available to us through the use of practical reason.
Certainly Braithwaite’s position does not deserve the brief dismissal by Hasker (2005) as a mere surrendering of the cognitive content of religious belief “in the interest of defending its personal and ethical significance in the life of the believer”, because with a suitably pragmatist reinterpretation, the latter kind of practical significance might actually be regarded as the source of any content a belief might have.
For an excellent critical discussion, highly relevant to the concerns of this paper, on how both atheists like Richard Dawkins and philosophers more sympathetic to religious belief like James have argued against agnosticism, see (Yoder 2013).
On the other hand, I do maintain that James’s “will to believe” can in principle be applied in very different contexts, potentially including meta-level contexts of various kinds. In a sense, I here wish to remain agnostic regarding the actual scope of applications of the will to believe principle; the purpose of this paper is not to offer any scholarly interpretation of James anyway. On James’s philosophy of religion and its relevance, cf. further (Pihlström 2013).
On agnosticism as “humanism”, see Section 5 and Section 7 below.
From a pragmatist perspective, it is of course another question what it would even mean to regard religious discourse as “truth-apt”: would this have to be construed in terms of something like James’s (James [1907] 1975) pragmatist account of truth? Here I must leave this issue aside (cf. again Pihlström 2013, 2020).
For example, Kearney’s (2010) “anatheism” might be an interesting relative of meaning agnosticism.
Wittgenstein’s On Certainty (Wittgenstein 1969), in particular, would be the key source here.
Again, I wish to emphasize that this “humanism” is not humanism in the sense of disbelief in God (nor is it, however, committed to any theistic belief, of course). See Section 5 above.
Knuuttila has, as far as I know, defended this view only in some of his publications in Finnish. In brief, the main idea is that for a religious person it is sufficient to hope that God exists; it is not necessary for one’s religious attitude to believe that He does. Thus, an agnostic could clearly be a religious person in Knuuttila’s sense, if they merely hope that God exists. Presumably, even an atheist could be “religious” in this sense, if they hoped, despite disbelieving, that God exists – given that it is possible to hope something to be the case that one believes not to be the case (as long as one does not believe it to be impossible).
As analyzed in (Pihlström 2020).
Back to TopTop