Hybrid expressivism and epistemic justification

Abstract

Epistemic expressivists maintain, to a first approximation, that epistemic assertions express non-cognitive mental states, like endorsements, valuations, or pro-attitudes, rather than cognitive mental states such as beliefs. Proponents of epistemic expressivism include Chrisman (Philos Stud 135: 225–254, 2007), Gibbard (Wise choices, apt feelings, 1990, Thinking how to live, 2003), Field (Proc Aristot Soc 96:359–379, 1996, Philos Stud 92:1–24, 1998, Philos Stud 143:249–290, 2009), Kappel (Acta Anal 25:175–194, 2010), and Ridge (Proc Aristot Soc 81:83–108, 2007a), among others. In this paper, I argue for an alternative view to epistemic expressivism. The view I seek to advocate is inspired by hybrid expressivist theories about moral judgments (see e.g. Barker (Analysis 60:268–279, 2000), Copp (Soc Philos Policy 18:1–43, 2001, in: Shafer-Landau (ed.) Oxford studies in metaethics, 2009), Finlay (J Ethics 8:205–223, 2004, Philos Impr 5:1–20, 2005), Strandberg (Philos Phenomenol Res 84:87–120, 2012)). According to these hybrid views, moral judgments express semantically cognitive or representational states and pragmatically convey the speaker’s non-cognitive mental states via implicatures. I will argue that a particular version of this view can reasonably be extended to epistemic judgments and that it has several advantages over its expressivist and cognitivist competitors. In particular, I will try to show that there exist certain phenomena in the epistemic domain that seem to be best accounted for by expressivist theories of epistemic judgments. However, a version of hybrid expressivism that maintains that epistemic judgments convey the attributor’s non-cognitive mental states via generalized conversational implicatures is able to account for these phenomena just as well without running afoul of the main problems that have been identified for different versions of epistemic expressivism.

1 Introduction

Virtually all first-order epistemologists assume that attributions of justification in the form of “S is justified in believing that p” or “It is epistemically rational for S to belief that p” do not significantly differ from claims that attribute a descriptive property to an object or individual. Epistemic claims thus feature the following characteristics: (i) they are genuine assertions with (ii) representational contents; (iii) the contents of epistemic claims are in principle true or false; and finally, (iv) there exist (presumably objective) epistemic facts that determine whether particular epistemic claims are true or false.¹ A view that accepts the first three of these theses could be referred to as “epistemic cognitivism.” If one further endorses the fourth thesis, then the resulting position might be called “epistemic realism.” It is evident that the core commitments of these views are equivalent to those of their moral counterparts, viz. moral cognitivism and moral realism. However, epistemic expressivists deny, to a first approximation, that epistemic claims should be characterized by theses (i) through (iv). Proponents of classical versions of epistemic expressivism maintain that epistemic claims serve to express the non-cognitive mental states of the attributor such as their endorsements, valuations or pro-attitudes.² Hence, they reject the first three claims characteristic of epistemic cognitivism listed above (and thereby also reject the final claim). Proponents of non-classical versions of epistemic expressivism, in contrast to supporters of the classical versions, concede to the epistemic realist that epistemic claims can be true or false. However, according to these positions, epistemic claims are only true in a “deflationist” or “minimalist” sense. Still others endorse ecumenical versions of epistemic expressivism. Ecumenical expressivists hold, very roughly, that epistemic claims express both beliefs and non-cognitive mental states. However, according to this position, the beliefs expressed by epistemic claims depend for their very content on the presence of the non-cognitive mental states [see (Ridge 2006, 2007a)].

In this paper, I will explore an alternative position to epistemic expressivism. The view I seek to advocate is an extension of hybrid expressivism (sometimes also called realist-expressivism) to epistemic judgments. Hybrid expressivism concerning moral discourse is advocated by Barker (2000), Copp (2001, 2009), Finlay (2004, 2005), and Strandberg (2012). According to these theorists, moral claims express cognitive or descriptive contents semantically, but pragmatically convey the speaker’s non-cognitive mental states via implicatures. I will attempt to show that when applied to epistemic judgments, a version of this hybrid view exhibits several advantages over its expressivist and cognitivist competitors. In particular, a hybrid expressivist view of epistemic judgments preserves what I take to be the core insight of expressivism without running afoul of the main problems that have been identified for different versions of epistemic expressivism.

The paper is organized as follows. In the first section, I will briefly present some considerations that seem to support epistemic expressivism. Although there are several (quite diverse) ways to motivate epistemic expressivism, I will focus only on those that I view as the most pivotal. In the second section, I will briefly consider different expressivist positions and some of the arguments that have been presented against them. I maintain that the problems that have been identified for these different versions of expressivism will vanish if we opt instead for hybrid epistemic expressivism. In the third section of the paper, I will provide a rough sketch of the version of hybrid expressivism that I think is the most promising for epistemic judgments.

2 From moral expressivism to epistemic expressivism

As noted in the introduction to this paper, there are several different versions of epistemic expressivism. These expressivist positions differ with respect to which of the core commitments of epistemic cognitivism or epistemic realism they reject. However, all of them maintain that the chief purpose of epistemic judgments is to express non-cognitive, desire-like mental states. But why should one endorse expressivism regarding epistemic discourse in the first place? In the moral case, expressivism can be motivated by various considerations. Perhaps the consideration that is most often presented in favor of moral expressivism is that moral expressivism provides the best account of the so-called practicality of moral judgment.³ The practicality of moral judgment consists of at least two closely related phenomena. First, there seems to be an “internal” connection between sincere moral judgment and motivation to action. For instance, a sincere utterance of “ϕ is wrong” would seem to imply that the speaker is at least to some extent motivated not to engage in ϕ. We generally assume that a person is motivated to act in accordance with her moral judgment, even if we do not possess any information about the individual’s particular situation. By contrast, if a speaker utters a descriptive sentence such as “It is snowing outside,” we do not assume that she is motivated to pursue a particular course of action. Only if we possess additional information—say, if we know that the speaker likes to take walks when it is snowing—can we assume that the person will be motivated to go outside. If moral judgments display this connection to the motivation to act, then there needs to be given an explanation as to why moral judgments exhibit this feature. According to the Humean theory of motivation, beliefs alone do not intrinsically motivate an agent to perform a particular action, but rather his desires together with his beliefs. If the expressivists are correct and moral claims merely express a speaker’s non-cognitive mental states, then sincere moral judgments imply that the speaker is in an action-guiding mental state and is therefore motivated to some extent to act in accordance with that judgment. Thus, if moral expressivism is true, the relationship between moral judgment and motivation to action can be accounted for in a straightforward fashion.

The practicality of moral judgments entails a second phenomenon based on the specific role that moral judgments are assumed to play. As expressivists such as Gibbard (1990, pp. 115–119) and Blackburn (1998) have pointed out, by making moral judgments, speakers intend to have an impact on the behavior of others. If a person sincerely states that it is wrong to ϕ, for instance, then we typically take this utterance to imply that she disapproves of actions of this type, and, moreover, that she does not want her audience to perform them. By contrast, if a speaker utters a descriptive sentence, we generally do not assume that she intends to have an impact on our behavior. In this way, Gibbard and Blackburn maintain, moral utterances can serve a certain function in society, namely regulation of the behavior of others. If the expressivists are right and moral judgments merely express desire-like mental states, then this phenomenon can also be accounted for in a straightforward manner. Moral cognitivists, however, who claim that moral judgments are descriptive and simply express representational contents (such as beliefs), have difficulty explaining this phenomenon.⁴ Consequently, the two considerations that have been presented in favor of moral expressivism make it seem likely, at least to a certain extent, that moral judgments are accompanied by non-cognitive mental states.

What about the epistemic case? Do these considerations carry over to epistemic judgments?⁵ In order for the two considerations to favor the existence of non-cognitive mental states that are expressed by epistemic judgments, similar phenomena need to exist in the epistemic realm. First, is there a connection between sincere epistemic judgment and epistemic motivation? Second, do epistemic judgments play a role similar to that of moral judgments in regulating epistemic behavior? As I will now try to argue for, I find that both of these questions should be answered in the affirmative.⁶

First off, to establish a close connection between sincere epistemic judgment and non-cognitive mental states, we should consider this phenomenon in light of cases of overt epistemic behavior. Reading reliable newspapers, seeking out experts or reliable informants when in doubt about a particular issue, repositioning oneself to obtain a better visual perspective, attending seminars or lectures, and so on are all common methods that epistemic agents employ in order to achieve better epistemic positions. But if an epistemic agent A sincerely judges that S will be justified or form an epistemically rational belief if S were to engage in behavior of this sort, then we tend to assume that the attributor A himself is motivated to act in accordance with his judgment. If an attributor would not display some motivation to engage in these forms of behavior in case she is disposed to make the corresponding epistemic judgment, then we would judge him to be either an incompetent user of epistemic terms like “epistemically justified” or “epistemically rational” or cognitively impaired. Thus, as with the moral case, the presence of an epistemic judgment seems to be reliably correlated with the corresponding motivation of the epistemic agent.

The cases presented so far have concerned overt epistemic behavior. But, on closer inspection, other cases that few philosophers would be tempted to classify as instances of epistemic actions do not seem to elicit a different verdict in us. If an attributor A sincerely judges that a person S is epistemically justified in believing p, where S is justified in this case via perception, introspection, or memory, for instance, we nonetheless tend to assume that the attributor A possesses some motivation that corresponds to this judgment. The motivational state might consist, for instance, in a desire to conform to epistemic norms or a desire to use (types of) belief-forming processes or methods. There are even more options available. I’ll address this question shortly. Though having perceptual experiences and believing in accordance with those perceptual experiences presumably does not qualify as engaging in a form of action, our verdict as to whether the attributor possesses some motivational state that corresponds to his judgment is not affected by this fact. Note that possessing a non-cognitive mental state that possesses some motivational force on the one hand and being, in principle, able to exercise this motivation such that there results an action on the other hand seem to be entirely distinct issues. An agent can desire to comply with a particular set of epistemic norms or desire to use particular (type of) belief-forming processes, though this does not imply that in every case in which this agent goes on to form a belief, she will ipso facto exercise this motivation.⁷ Thus, I think that the connection between epistemic judgments and being motivated to act in accordance with that judgment is fairly robust and is not limited just to some cases of attribution of justification or epistemic rationality.

If this correlation between sincere epistemic judgments and action-guiding, desire-like mental states indeed holds, then something needs to be said about the non-cognitive mental states in the epistemic domain. Since epistemic judgments are concerned with the evaluation of beliefs instead of actions, it seems difficult, at least on the face of it, to come up with an epistemic analog of a non-cognitive mental state that is expressed by a moral judgment.⁸ I have already hinted as to what the epistemic action-guiding states might consist in. First, one might conceive of the non-cognitive mental states in the epistemic domain as desires to conform to epistemic norms. However, it remains controversial what the precise content of these epistemic norms is. Fortunately, no stance towards this issue is needed in the present context. Second, the non-cognitive mental states could consist in desires to use (types of) belief-forming processes or methods. If a person were to judge that some epistemic agent S is justified in believing p, then, according to this account, this person would desire the utilizing of the belief-forming process or method that underlies her ascription of justification or rationality. In the present context, it doesn’t matter which of these accounts of non-cognitive mental states is ultimately correct.⁹ I’ve presented these options in order to illustrate that the concept of an epistemic action-guiding state is not ultimately confused. Moreover, both options possess a fair amount of generality (i.e., they are not limited only to some attributions of epistemic justification or rationality).

If it is indeed the case that epistemic agents express non-cognitive mental states when they tend to make the corresponding epistemic judgments, then it seems plausible that the second consideration in favor of moral expressivism described above will also carry over to the epistemic case, viz. in making epistemic claims, we let our audience know that we want or do not want them to engage in belief-forming behavior of the type in question. Dogramaci (2012, p. 520) has emphasized that our epistemic evaluations have the tendency to influence the epistemic behavior of others. He says¹⁰:

When we make our evaluations of others’ beliefs, we are intending for them to follow the correct rules. Actually, I’ll claim something slightly stronger; I claim this: our evaluations have an overall tendency to influence our audience to follow the endorsed rules. (…). What I am claiming is just a commonsensical truth about our being built to respond to each others’ evaluations in accommodating ways.¹¹

Thus, if I am correct, then the two considerations usually presented in favor of moral expressivism carry over to the epistemic case. If we assume that there is a close connection between epistemic judgment and the presence of motivational states and that epistemic judgments serve to regulate epistemic behavior, then this casts doubt on the claim that epistemic judgments are purely descriptive and serve only to express descriptive or cognitive contents. In light of these phenomena, epistemic expressivists will maintain that the considerations presented above support the adoption of some form of epistemic expressivism. However, like its counterpart moral expressivism, epistemic expressivism is by no means an uncontroversial position. In the following section, I will examine three versions of epistemic expressivism.

3 Classical, non-classical, and ecumenical versions of epistemic expressivism

In the introduction to this paper, I briefly introduced three different versions of epistemic expressivism—namely, classical, non-classical, and ecumenical versions of expressivism. Proponents of the so-called classical versions of expressivism maintain that the main function of epistemic discourse is to express the non-cognitive mental states of attributors, such as their endorsements, valuations, or pro-attitudes. Field (1996, 1998) and Gibbard (1990) have developed versions of epistemic expressivism of this sort. Although Field’s version of epistemic expressivism in his (1996, 1998) is primarily concerned with a priori justification, his view can nonetheless encompass cases of empirical justification. Field argues that in calling a belief justified, we do not attribute any property to the belief or the agent, but rather adopt a policy or stance towards the belief in question (see e.g. Field 1996, p. 377). When we claim that a belief is justified, we express our positive stance towards certain evidential norms that license S’s belief that p. Although Field maintains that we express non-cognitive mental states in attributing a justified belief to someone, he nonetheless assumes that we additionally express beliefs. However, these beliefs merely concern whether an agent indeed satisfies the epistemic norms in question (see Field 1998, p. 6). As I alluded to in the introduction, Field’s core claim implies that there are no epistemic facts as the epistemic realist conceives of them.

Non-classical versions of expressivism are characterized by their rejection of the thesis that epistemic discourse is not truth-apt—a claim that is characteristic of classical versions of epistemic expressivism. Proponents of non-classical expressivism, such as Blackburn (1998), Timmons (1999), Gibbard (2003), and Field (2009), hold that epistemic claims express non-cognitive mental states and are nonetheless true, but only in a “deflationist” or “minimalist” sense. According to Cuneo’s (2007, pp. 148–162) exposition of the basic commitments of this version of epistemic expressivism, the non-traditional expressivist adopts three distinctive theses.¹² First, there is what Cuneo refers to as the “Modified Alethic Thesis” concerning epistemic claims. This thesis maintains that calling a statement true does not entail any ontological commitments to the effect that epistemic facts as the realist conceives of them exist. In order for a statement “p” to be true, the deflationist contends, one simply has to assume p. Second, the non-classical expressivist adopts a particular thesis concerning epistemic assertions. According to Cuneo’s exposition of this view, when an agent “sincerely utters a predicative epistemic sentence, that agent thereby does not assert an epistemic proposition, but rather ‘asserts’ a quasi-proposition” (2007, p. 147). Finally, the third claim endorsed by the non-traditional expressivist (according to Cuneo) is the “Modified Ontic Thesis.” Whereas the epistemic realist is committed to the idea that there are epistemic facts that account for the truth of epistemic claims, the expressivist must introduce a surrogate for the realist’s notion of a fact. Cuneo calls this concept of facts “quasi-facts.”

Finally, there is Ridge’s (2007a) ecumenical expressivism. Although Ridge (2007a) develops the epistemic version of ecumenical expressivism for attributions of knowledge, here I will consider how this view fares with respect to attributions of justification.¹³ In Ridge’s theory, if a speaker utters “S is justified in believing that p,” then she expresses (i) an “epistemic endorsement of certain procedures for deciding what to believe” and (ii) the belief that S’s belief that p is causally regulated by “either (a) those procedures that she endorses in (i) or (b) procedures that are close enough to those procedures, so far as p goes, or (c) more fully informed successors to those procedures” (Ridge 2007a, p. 103). The complexity of Ridge’s second condition is due to the fact that it is intended to explain cases in which we attribute a justified belief to someone who uses either less sophisticated or more sophisticated methods of belief-formation than the attributor. Moreover, the belief that is mentioned in condition (ii) of Ridge’s theory refers anaphorically to a non-cognitive state and thus depends for its very content on the presence of such a non-cognitive mental state. There exist even more versions of epistemic expressivism, such as the theories proposed by Kappel (2010) and Chrisman (2007), though I will mainly focus on the views just sketched.

What is common to all the views outlined is that they are committed to the assumption that non-cognitive mental states are expressed semantically by epistemic judgments. One very common strategy to call this assumption into question—and thereby to show that expressivism is problematic—is to present cases in which an epistemic utterance is true, though the attributor is not in a non-cognitive mental state that is said to be part of the semantic content of the judgment.¹⁴ Kvanvig (2003), for instance, has mentioned in passing that it seems possible to him that there might be Spock-like individuals who are highly intelligent but lack any affective states (see Kvanvig 2003, pp. 177–179). If a Spock-like individual were to assert that a particular person is justified in believing a proposition p, then this utterance would not express the non-cognitive mental states of the attributor, since this individual lacks the non-cognitive mental state in question. It is evident that Kvanvig’s Spock is just the counterpart of the amoralist, who is said to cause trouble for the moral expressivist (and motivational internalism). Moreover, Kvanvig (2003, p. 177) has also drawn attention to even more mundane phenomena that call into question that non-cognitive mental states are expressed on every occasion in which an epistemic utterance is made. He thinks, for example, that cases of depression cause problems for views like expressivism, since these cases allegedly demonstrate that an epistemic agent might be deprived of mental states that motivate this agent to act, though the agent thereby does not cease to be in a position to possess justified beliefs or make true epistemic judgments. Kvanvig’s case might even be strengthened by reference to further cases that could be referred to as cases of “epistemic indifference.”¹⁵ Cases like these are characterized by the fact that an agent is in some psychological condition that causes the agent not to care about epistemic matters and thereby to possess no motivation to act in accordance with her epistemic judgments. The most trivial examples of this sort are cases of listlessness. Still other cases might involve agents that do not care about epistemic matters to the same degree as they care about other matters. For example, stubborn people tend to care more about not giving up their own beliefs than about having true beliefs. And these people tend to concede at times, when their beliefs have been challenged, that it is rational or justified to believe the opposite of what they believe. Still, they are lacking the motivation to act in accordance with their judgment. Moreover, a person of this sort cannot be described as being akratic or being in possession of some initial motivation that is trumped or overridden by other concerns.

There already exist some (very sketchy) replies to these concerns by proponents of epistemic expressivism. Chrisman and Carter (2012, p. 327, footnote 3) have responded on behalf of Field to Kvanvig’s first objection. They have focused on Kvanvig’s Spock case and have maintained that Kvanvig’s Spock is no threat to epistemic expressivism, since Kvanvig’s original case is allegedly construed such that this individual does not make utterances to the effect that someone knows or is being justified in believing a particular proposition. But, of course, if Kvanvig’s original case were modified in the way indicated above, this reply will not work. Chrisman and Carter offer a second reason for why they doubt that an individual such as Kvanvig’s Spock should be considered a counterexample to epistemic expressivism. Chrisman and Carter seem to think that an individual such as Kvanvig’s Spock is not possible after all, since they find it unclear how Spock might be motivated to engage in epistemic behavior (such as gathering evidence for a particular hypothesis). However, one might reply that denying that Spock can be motivated to engage in epistemic behavior is a very speculative hypothesis that is thus unsuitable to carry the burden of proof that Chrisman and Carter intend it to carry. (It might be contended, on the contrary, that the assumption that Spock can be motivated to engage in epistemic behavior, though he is deprived of all non-cognitive mental states, is an equally plausible hypothesis, given all we know about this particular character.) Finally, even if we were to concede to Chrisman and Carter that an individual such as Kvanvig’s Spock is not possible, there are still the more mundane cases, such as those mentioned above, that suggest that not every attribution of a justified or epistemically rational belief expresses the non-cognitive states of the attributor. This worry is not addressed by Chrisman and Carter’s reply.

A further reply in defense of epistemic expressivism to the objections outlined above can be extrapolated from how Michael Ridge intends to handle cases of akrasia in the moral domain (see Ridge (2007b).¹⁶ This sort of reply is specific to Ridge’s version of expressivism and might thus be of help in saving his account from the problems discussed above. Ridge points out that most expressivists have tried to deal with cases of akrasia by positing conflicting motivational states—an approach that Ridge deems to be unsuccessful. His brand of expressivism (i.e. ecumenical expressivism) has the resources to handle cases of akrasia in a different way. Ridge claims that the akratic agent is still in possession of a non-cognitive state, though this state is not directed at any one particular action but rather directed at actions of a certain general sort, such as the state that is specified in condition (i) of Ridge’s account described above.¹⁷ However, according to Ridge, agents that suffer from akrasia are not in possession of a proximate intention that confers some motivational “oomph” on the agent. Moreover, according to Ridge’s diagnosis, the person suffering from akrasia still has a belief that engaging in some action ϕ is the best way to achieve a particular end. According to Ridge, this is what makes akrasia irrational.

But is this proposal really of help in dealing with the cases I have introduced above? With respect to Kvanvig’s Spock, for instance, who is ex hypothesi deprived of all non-cognitive states, it makes little sense to maintain that Ridge’s strategy will be successful. Since it is assumed that Kvanvig’s Spock does not possess any non-cognitive states, Ridge’s contention that an endorsement of a very general sort is still present seems somewhat implausible. Moreover, I think that even with respect to other cases, the hypothesis that an agent is still in possession of an endorsement of a very general kind is, on closer inspection, not convincing. Think of cases of severe depression. It seems to me that the depressed agent is not in possession of non-cognitive states of this general sort. Thus, Ridge’s account will not be able to handle this case if his proposal consists of just denying that the agent lacks a proximate intention in cases such as these.

Nonetheless, to the defender of epistemic expressivism (or motivational internalism in the epistemic domain), there are even more strategies available for dealing with the problems discussed above.¹⁸ These strategies are fairly prevalent in the literature on moral expressivism and motivational internalism, though up until now they have not been explicitly discussed in print in the literature on epistemic expressivism.¹⁹ Since there is insufficient space here to fully delve into this issue, my discussion of these strategies will be, to an extent, cursory.²⁰

One very popular strategy that has been employed by various defenders of motivational internalism in the moral domain consists of qualifying the claim that sincere moral judgments are accompanied by motivational states. The generic claim of these positions is that if a person sincerely judges that ϕ is wrong, then she is motivated to ϕ only in case she satisfies a certain further condition C. It has been proposed that C might be the condition of (a) being psychologically normal, (b) being practically rational, or (c) being a fully competent user of evaluative terms.²¹ It might thus be suggested that, for epistemic judgments, a qualified view holds as well. However, it is highly doubtful whether any of these conditions is able to deal with all the cases that call into question that the connection between sincere evaluative judgments and the presence of motivational states is a matter of conceptual necessity or semantic content. It is doubtful, for instance, whether someone who cares less about epistemic matters than about other matters or is suffering from temporary listlessness fails to satisfy any of these conditions. Moreover, these proposals are plagued by further problems on their own, such as that they seem to conflict with the core internalist claim or that they leave it unexplained as to why fulfilling any of these conditions gives rise to the presence of motivational states.²²

A further strategy that has just recently been proposed for dealing with the cases outlined above in the moral domain involves questioning the evidential value of the intuitions that cases such as the amoralist elicit in professional philosophers. It has been maintained that externalist-friendly intuitions (i.e., intuitions to the effect that persons such as the amoralist who lack motivational states but nonetheless possess genuine moral beliefs or moral understanding are genuinely possible) are theoretically biased and thus do not represent neutral data.

Though two surveys exist that seem to show that ordinary folk have externalist-friendly intuitions—namely one conducted by Nichols (2004, chapter 3) and another done by Strandberg and Björklund (2013)—a more recent survey seems to suggest that things aren’t as straightforward as Nichols’ and Björnsson and Strandberg’s results allegedly demonstrate. Björnsson et al. (2014) surveyed the intuitions of purportedly theoretically unbiased persons and found that the test-persons were divided concerning the attribution of moral beliefs to individuals such as the psychopath. However, the survey also seems to show that cases in which temporary listlessness causes the character of the case that Björnsson et al. present to their subjects not to be in a motivational state are, by most subjects, still conceived of as being instances of moral belief and moral understanding (see Björnsson et al. 2014, p. 15).

If the results of Björnsson et al. (with respect to the latter case) are to be trusted, denying the intuition that a person who suffers from listlessness is in possession of a moral belief thus does not seem to be a viable option. Moreover, in case the results should transfer to the epistemic domain—what seems likely but is still an open question without the existence of any empirical surveys—the epistemic expressivist is not able to debunk the intuitions that put her view into jeopardy. Of course, whether experimentalist surveys are, in principle, able to confirm a particular philosophical position is still a contested issue, since several philosophers have expressed doubt about the significance of experimentalist surveys.²³ Moreover, there is much room for debate about the design of a particular survey and whether the particular conditions that the subjects are in shape the results of the surveys in an important way.²⁴ Finally, there is the live possibility that the subjects mistake semantic with pragmatic properties of a sentence or utterance and that there are thus pragmatic explanations available as to why the subjects respond in the way they do. But nevertheless, at least with respect to the kind of reply under consideration—viz. denying the intuition that cases of listlessness are not cases of genuine moral belief—does not seem to be a very plausible option.

4 Hybrid expressivism and epistemic discourse

As I indicated in the previous section, all expressivist positions seem to be problematic. In particular, one difficulty that I identified for every expressivist position is that it seems possible that a sincere utterance of “S is justified in believing p” can be true even though the attributor is not in an action-guiding mental state. This phenomenon seems to demand that one should divorce action-guiding attitudes from the semantic content of utterances that attribute a justified belief to someone. Hybrid expressivism with respect to epistemic judgments is a position in which the non-cognitive mental states of an agent are pragmatically conveyed via implicatures. Thus, the basic idea of hybrid expressivism in connection to epistemic judgments is as follows:

Hybrid Expressivism (Epistemic): “S is justified in believing that p” (i) semantically expresses a belief or descriptive content and (ii) pragmatically conveys the speaker’s non-cognitive mental states via implicatures.

If hybrid expressivism is considered in light of the problems that I reviewed for expressivist positions in the previous section, it becomes evident that hybrid expressivism is in principle immune to these charges.

As I briefly mentioned at the outset of this paper, hybrid expressivism with regard to epistemic judgments is simply an extension of the hybrid expressivism that has been proposed in different guises for moral judgments. The guiding idea of these accounts is that the sincere utterance of a sentence such as “It is wrong/good to ϕ” (i) semantically expresses a descriptive content or proposition and (ii) pragmatically conveys the speaker’s non-cognitive mental states via implicatures. However, precisely what kind of an implicature is generated by a moral utterance is a matter of some controversy. Copp (2001, 2009), for instance, maintains that moral utterances convey the speaker’s non-cognitive mental states via conventional implicatures. Finlay (2004, 2005) argues that moral utterances carry particularized conversational implicatures to the effect that the speaker possesses a particular non-cognitive mental state. In addition, Strandberg (2012) has made the case that moral utterances trigger generalized conversational implicatures that the speaker is in a certain non-cognitive mental state. Hence, there are at least three different options with respect to the type of implicatures generated by moral utterances. Which of these three potential options is most promising for the epistemic case? I will seek to argue that the most plausible option is an analogue of Strandberg’s position, i.e., a position that maintains that epistemic utterances convey the speaker’s non-cognitive mental states via generalized conversational implicatures.

4.1 The case for the generalized conversational implicature view

Because I will attempt to make a case for the theory proposing that these non-cognitive states are conveyed via generalized conversational implicatures, I will now examine the two other options in greater detail, in an attempt to show that my preferred option is the most promising.

I start with the particularized conversational implicature view. These types of implicatures possess the following characteristics: (a) they tend to be indeterminate, (b) they are reinforceable, (c) non-detachable, (d) cancellable, and (e) calculable [see Grice (1989) and Sadock (1991)]. If judgments of the sort “S is justified in believing that p” were to trigger only particularized conversational implicatures, then the implicatures generated by these judgments would exhibit features (a) through (e). However, it is evident that some of these features are not compatible with the properties of implicatures generated by epistemic judgments concerning the presence of action-guiding mental states. Consider first the property (a), viz. that particularized conversational implicatures tend to be indeterminate. The indeterminacy of implicatures can be illustrated by the following example. Assume that Professor Smith were to state “Jones has good handwriting” upon being asked what distinguishes Jones from Professor Smith’s other students. Upon closer inspection, it is evident that there are multiple implicatures that this utterance might generate. This utterance could carry the implicature that Jones is a bad student, but it also might convey the implicature that Jones is a mediocre student, or that Jones is a total idiot, or that Professor Smith is only knowledgeable about Jones’ handwriting skills and therefore cannot say anything about how Jones fares as a student. As this example illustrates, there are multiple implicatures that the audience could infer from Professor Smith’s utterance to make it compatible with Grice’s overarching principle of cooperation. However, I think that it is unpromising to classify the implicatures generated by epistemic claims as particularized conversational implicatures. Like their moral counterparts, epistemic claims carry highly determinate implicatures concerning the presence of action-guiding mental states. In this respect, the implicatures generated by epistemic claims differ significantly from the implicatures triggered by claims such as the one concerning Jones’ handwriting. What about feature (b), i.e., that particularized conversational implicatures are reinforceable? An implicature is reinforceable if the content of the implicature can be made explicit without causing any redundancy in the informational content carried by this particular utterance. Consider the case discussed in connection with the first feature of particularized conversational implicatures. The implicature generated by Professor Smith’s utterance of “Jones has excellent handwriting” could be made explicit, for instance, if Professor Smith were to say “Jones is just a mediocre student.” Since particularized conversational implicatures tend to be indeterminate, Professors Smith’s second utterance is not redundant. If epistemic judgments are considered in terms of this feature, it is less apparent that their implicated content can be reinforced by explicitly stating it. If I were to claim “Jones is justified in believing that p” and then continue, “I endorse the norms that license Jones’ belief,” then we would typically take the second utterance to be to some extent superfluous. If this is indeed the case, then this provides further evidence for the claim that epistemic judgments do not generate particularized conversational implicatures.

What about the second option, treating the implicatures triggered by epistemic judgments as conventional implicatures? In contrast to particularized conversational implicatures, conventional implicatures are (a) highly determinate, (b) not reinforceable, (c) detachable, (d) non-cancellable, and (e) non-calculable [see Grice (1989) and Sadock (1991)]. Copp (2001, 2009) has forcefully argued that in the moral case, the meaning-like connection between moral judgments and the possession of a non-cognitive mental state can only be captured by the assumption that moral terms conventionally implicate that the agent is in some non-cognitive mental state. Copp maintains that moral terms behave semantically and pragmatically like pejoratives, in that they possess both descriptive and evaluative content. For instance, if a speaker states, “Paolo is a wop,” this utterance conveys that Paolo is an Italian and that the speaker feels contempt for Italians. The evaluative content of this utterance (i.e., that the speaker feels contempt for Italians) is, according to Copp, implicated conventionally. Copp’s account enjoys several advantages over the particularized conversational implicature view. Concerning the first characteristic that conventional implicatures exhibit, their determinacy, one could argue that this feature makes them more suitable to account for the relationship between moral judgments and the presence of action-guiding mental states. As I laid out in the discussion of particularized conversational implicatures, it is highly implausible to conceive of the implicatures that are triggered by moral and epistemic assertions as being indeterminate. Thus, in this regard, the conventional implicature view seems more promising.

The second property of conventional implicatures mentioned above also makes them more suitable to explain the connection between epistemic judgments and the presence of non-cognitive mental states. Consider the following utterance: “Paolo is a wop and I feel contempt towards Italians.” The second conjunct is somewhat redundant, since the use of the derogatory term “wop” already indicates that the speaker feels contempt for Italians. As I indicated previously in my discussion of particularized conversational implicatures, the implicatures generated by epistemic judgments do not seem to exhibit the property of being reinforceable by an explicit statement that the speaker is in some action-guiding mental state. In this respect, then, the conventional implicature view seems to be better suited to account for the implicatures generated by epistemic claims.

One problem that complicates the conventional implicature view is that conventional implicatures are not cancellable in the way that particularized and generalized conversational implicatures are cancellable. For instance, if someone states, “Paolo is a wop” and does not feel contempt for Italians, the speaker would thereby call into question her use of the term “wop,” and we would be tempted to assume that she is not a competent user of this term. Thus, the conventional implicature view is unable to handle the cases discussed in the second section of this paper that suggest that the implicatures generated by epistemic claims (as well as those generated by moral claims) are indeed cancellable. Copp (2009) has anticipated this objection to his theory of moral utterances. He contends that it is semantically inappropriate to use moral terms when one has no action-guiding mental states that correspond to the moral judgments one is disposed to make. However, I think that this response will not work. As the examples described in the second section of this paper suggest, if someone is devoid of all action-guiding states, as Kvanvig’s Spock is, then this person would commit linguistic mistakes when she judges that someone is justified in believing that p. I find this assumption highly questionable.

Moreover, conventional implicatures are also generated when the terms that trigger these implicatures are embedded in complex sentences, such as conditionals. For example, consider the sentence “If Paolo is a wop, then we should not invite him to our party.” When a speaker utters a sentence like this one, then the audience will typically infer that the speaker feels contempt for Italians. This is explained by the fact that conventional implicatures are tied to the conventional meaning of an expression and are therefore generated by every use of the term in question. However, epistemic and moral terms do not seem to exhibit this kind of behavior. Consider the following examples: “If it is rational to believe p and p entails q, then it is rational to believe q” and “If it is wrong to beat up one’s children, then it is wrong for Jenny to beat up her son John.” Neither sentence seems to carry the implicature that the speaker subscribes to the relevant moral or epistemic norm, respectively. In this regard, conventional implicatures do not seem to be triggered by epistemic and moral claims.

What about the final option on the table, the generalized conversational implicature view? Generalized conversational implicatures are (a) highly determinate, (b) do not tend to be reinforceable, (c) are non-detachable, (d) cancellable, and (e) calculable.²⁵ If the generalized conversational implicature view is correct, then utterances that contain epistemic predicates will behave similar to the utterances “Harry and Sue bought a piano” or “Mary has three children.” In most conversational contexts, the statement “Harry and Sue bought a piano” generates the implicature that they bought it together, not one each, whereas an utterance of “Mary has three children” will trigger the implicature that Mary has no more than three children. If attributions of justification indeed generate generalized conversational implicatures, then an utterance of “S is justified in believing that p” will carry in most conversational contexts the implicature that the speaker possesses some motivation that corresponds to that judgment. As the previous discussion has made apparent, the particularized conversational implicature view struggles to account for the fact that epistemic judgments generate highly determinate implicatures. Generalized conversational implicatures, by contrast, are highly determinate. In this respect, then, it is more promising to conceive of the implicatures generated by epistemic claims as generalized conversational implicatures. Moreover, due to the fact that generalized conversational implicatures are cancellable, this view is not affected by the problems that beset the conventional implicature view. Since generalized conversational implicatures are not part of conventional meaning, they can be cancelled without thereby implying that the speaker has committed a linguistic mistake. Therefore, the generalized conversational implicature view is able to handle the cases that speak against the conventional implicature view, i.e., cases in which this implicature is for instance contextually cancelled. In sum, I find that the generalized conversational implicature is the most promising view concerning the implicatures generated by epistemic claims.

4.2 How are these generalized conversational implicatures generated?

If the main thesis of the present paper is correct and epistemic judgments indeed carry generalized conversational implicatures to the effect that the attributor is in some non-cognitive mental state, there needs to be an explanation given for what the underlying mechanism is that prompts a hearer to infer these implicatures.²⁶ In contrast to conventional implicatures, which are grasped “immediately” according to Grice, conversational implicatures are calculated on the basis of certain principles. Grice has presented a very general pattern that is said to underlie the calculation of particularized as well as generalized conversational implicatures. The basic form of this pattern can be put schematically as follows, where S is a speaker and H is a hearer (cf. Grice 1989, pp. 49–50):

• (i) S has said that q, and for H there is no reason to assume that S is not observing the cooperative principle and the conversational maxims;

• (ii) In order to make S’ utterance consistent with the cooperative principle and the conversational maxims, H needs to assume that S thinks or intends to convey that p;

• (iii) S believes or knows that H will assume that S intends to convey that p in order to make S’ utterance consistent with the cooperative principle;

• (iv) S has done nothing to stop H from thinking that S intends to convey that p;

• (v) S indeed has implicated that p.

With respect to the calculation of generalized conversational implicatures, it has been suggested that Grice’s maxim of relation (“Be relevant”) and the maxim of quantity (“Be as informative as required/not more informative as required”) give rise to these types of implicatures.

Levinson (2000), by contrast, has outlined a more refined proposal concerning the generation of generalized conversational implicatures that is, however, still Gricean in spirit. Levinson maintains that generalized conversational implicatures belong to the level of utterance-type meaning (i.e., a third layer of meaning in between literal-meaning and speaker-meaning) and arise out of certain expectations about how language is in general used. He thinks that generalized conversational implicatures result from default inferences that hearers usually make. Levinson suggests that there are three different inferential heuristics or principles at work when hearers infer generalized conversational implicatures—namely, the quality heuristic (“What isn’t said isn’t”), the informativeness heuristic (“What is expressed is stereotypically exemplified”), and the markedness heuristic (“What is said in an abnormal way isn’t normal”) (cf. Levinson 2000, pp. 35–39). Each of these heuristics can be derived from some of the four conversational maxims that Grice has introduced.

For the case at hand, Levinson’s informativeness heuristic is the particular heuristic able to explain why utterances in the form of “S is justified in believing that p” generally convey that the speaker is in some non-cognitive mental state. Levinson assumes that hearers have the tendency to interpret utterances in line with what they take to be normal or typical in a given situation. According to Levinson’s proposal, the informativeness heuristic is supposed to capture this tendency of hearers since it licenses inferences to the effect that a certain stereotypical property is exemplified when an utterance of a specific sort is made. Levinson provides a very detailed account as to how to understand this heuristic (see Levinson (2000, pp. 112–116). Because the details of Levinson’s proposal are not important for present purposes, I will just focus on the two main principles that comprise the informativeness heuristic or principle (Levinson 2000, p. 114):

(I-Principle)

Speaker’s Maxim: the maxim of Maximization. “Say as little as necessary”; that is, produce the minimal linguistic information sufficient to achieve your communicational ends (…).

Recipient’s corollary: the Enrichment rule. Amplify the informational content of the speaker’s utterance, by finding the most specific interpretation, up to what you judge to be the speaker’s m-intended point, unless the speaker has broken the maxim of Maximization by using a marked or prolix expression.

The following examples illustrate how these two principles give rise to generalized conversational implicatures (cf. Levinson 2000, p. 32).

• (1) The blue pyramid is on the red cube.

  Licensed inferences:

  • (a) The pyramid is a stereotypical one (i.e., one that is on a square rather than a hexagonal base).

  • (b) The pyramid is directly supported by the cube. (There is no intervening slab.)

• (2) Lebron James scored 30 points, and the Cavs won the game.

  Licensed inference:

  • (a) Lebron James was the main cause of why the Cavs won the game.

The implicatures (1a)–(2a) are inferred by a hearer on most occasions in which (1) and (2) are uttered because both speaker and hearer rely on the informativeness principle. In case an utterance of (1) or (2) is made and the speaker has not done anything in order to stop the hearer from thinking that the speaker herself wants to convey the implicated content, then the implicature concerning the exemplification of the stereotypical property in question will be inferred by the hearer. Still, there exist different ways by which the speaker can cause the hearer to not infer the implicated content. For example, in case Lebron James did not cause the Cavs to win the game (and this fact were somehow relevant to the conversation) and the speaker said instead “Lebron James scored 30 points, and the Cavs won the game, despite the fact that Lebron James committed too many turnovers and a dumb foul at the end the of the fourth quarter that almost cost them the victory,” the hearer will not assume that the speaker believes that the stereotypical property in question (Lebron James being the main cause of why the Cavs won the game) is exemplified. In this case, the speaker would have blocked this inference by stating that Lebron James committed many turnovers and a foul that almost caused the Cavs to lose. But if a speaker does not choose to use this particular sentence or does not signal in some other way that (2a) is not the case, the hearer will typically infer that (2a) is the case when (2) is uttered.

Similarly, if a speaker makes an utterance to the effect that someone is epistemically justified in believing a particular proposition, the hearer will infer that another stereotypical property is exemplified in this particular case as well—namely, that the speaker possesses some motivation to act in accordance with that judgment. As laid out in the first section of this paper, sincere epistemic judgments seem to be reliably correlated with the presence of motivational states and the presence of these states should thus be considered as typical or normal on most occasions. It can be assumed also in this case that the hearer bases her inference of the implicated content on the informativeness principle. However, as with the examples discussed in the previous paragraph, there exist cases in which this implicature will not be inferred by the hearer. In case an utterance is made by someone who is devoid of any action-guiding states by being in a certain psychological condition and the hearer is aware of this fact, it is safe to assume that the hearer will not make the inference that the speaker instantiates this stereotypical feature. Thus, I think that Levinson’s informativeness principle allows for an explanation as to why epistemic utterances generate generalized conversational implicatures that the speaker possesses some motivation that corresponds to her judgment.

5 Conclusion

In this paper, I have attempted to show that hybrid expressivism with respect to epistemic judgments preserves what I take to be the core insight of epistemic expressivism but is nonetheless capable of withstanding the charges that have been levelled against different versions of epistemic expressivism. Although there are different options in terms of the kind of implicature that epistemic claims might generate, I have argued that a version of hybrid expressivism in which the non-cognitive states of the attributor are conveyed via generalized conversational implicatures is the most promising.