1. Disagreeing and Voicing Disagreement

In 1994, Richard Herrnstein and Charles Murray published The Bell Curve. In it, the authors notoriously argued that the difference in performance on IQ tests across races is due to genetics rather than socialization. They argued that difference between African American average IQ scores and white American scores is caused, at least in part, by genetic differences. They further argued that this difference in genetics also explains differences in income, job success, and criminality, among other things.Footnote ¹ These claims, according to Herrnstein and Murray, were the results of years of work in behavioural genetics.

The Bell Curve was met with wide public protest. There were (and continue to be) sit-ins and calls for Murray to resign his various academic posts (Herrnstein died soon after publication). Some behavioural geneticists, particularly those interested in animal behavioural genetics, dismissed the work as well. Geneticist David Botstein, for example, called the text, ‘so stupid that it is not rebuttable’.Footnote ²

However, the academic reaction was not entirely negative. In the 1995 keynote address to the Behavioral Genetics Association, Glayde Whitney, professor of behavioural genetics, suggested that Murray and Herrnstein's work be pursued – that The Bell Curve opened many lines of important research. Indeed, many geneticists interested in intelligence voiced their support of the book, going so far as to call it an important contribution. An editorial in The Wall Street Journal, signed by fifty-two geneticists defended Herrnstein and Murray against accusations that their data and analysis were problematic.Footnote ³

Despite signing this editorial, many of these same geneticists thought that the science of the book was deeply flawed. Subsequent interviews with signatories to that editorial revealed that they never thought that the relationship between genetic correlates for intelligence and race were informative or even possibly informative. These scientists, then, disagreed with the findings of The Bell Curve. They just didn't say so.

It seems to me that these disagreeing scientists failed to meet their obligations. The interviews were anonymous, so we'll call two of the scientists Lucy and Ethel. Lucy and Ethel failed in a couple of ways. First, Lucy and Ethel participated in the publication of an editorial that said, or perhaps implied, something they did not believe. To some extent, they lied or misled the readers of that editorial.Footnote ⁴ Second, Lucy and Ethel ought to have said something about what they in fact believed. Intuitively, Lucy and Ethel should have publicly testified that they disagreed with the findings in The Bell Curve.

Lucy and Ethel probably had some practical reasons not to voice their disagreement: the political climate was uncertain, Murray, Whitney and other proponents of the book were powerful, and the field of behavioural genetics was new. To fight within the field might have cast doubt on the whole discipline. Further, they may have had principled reasons to protect scientists from intimidation and protest – it is reasonable to think that scientific progress depends on a high degree of academic freedom, and that that freedom is only protected if scientists can publish their results regardless of public outcry.

These reasons, however, don't seem to override the sense that Lucy and Ethel ought to have voiced their disagreements with their findings. We can imagine a graduate student in Lucy's lab being confused or misled by The Bell Curve. This grad student might go on to pursue a faulty plan of study. Lucy could, plausibly, have prevented this by telling her graduate student that she disagreed with the findings. We can imagine Ethel's African-American neighbour coming to believe that she will, as a matter of genetic predisposition, be unable to retain her job or keep her children out of prison. If Ethel had testified that she disagreed with Murray and Herrnstein's findings, she could have helped assuage these fears. These obligations seem to over-ride the practical considerations against testifying.

So, there were good reasons for Lucy and Ethel to testify that they disagreed with the findings of The Bell Curve. They had, I think, an obligation to do so.Footnote ⁵ Further, I think they had an epistemic obligation to do so. That is, Lucy and Ethel, by failing to testify that they disagreed, were failing to behave as they epistemically ought. And this is the case whether or not they believed as they epistemically ought.Footnote ⁶ I will defend this claim in this paper.

In the next section I will briefly discuss epistemic obligations in general. In section three, I'll demonstrate that our obligations to voice disagreement are plausibly epistemic, even if they are sometimes also moral. Then, in section four, I'll offer a hypothesis of what it takes to discharge obligations to voice disagreements. Finally, in section five, I'll consider four accounts of the source of these obligations. These accounts are not competitors – however, by highlighting multiple possible accounts I will attempt to make the best possible case for the claim that we have a prima facie epistemic obligation to voice disagreements.

2. Epistemic Obligations

The discussion of epistemic obligations in the philosophical literature is largely concerned with the formation of beliefs. Usually these obligations are negative – they are obligations to refrain from certain epistemic actions.Footnote ⁷ For example, we ought not believe that p if our evidence rules p out. We ought (perhaps) not to remain as confident in our belief q if our epistemic peer disagrees. Positive epistemic norms are often considered to be permissions rather than obligations. Given a set of evidence E, you may believe that p. So, epistemic obligations to testify to disagreement, if indeed we have them, have gone largely unnoticed in the literature.

It is not plausible that we have an ultima facie obligation to testify when we disagree, even if a case can be made for some kind of obligation to do so. If I voice all of my beliefs whenever they happen to be in disagreement with content already expressed,Footnote ⁸ I'd likely be violating other sorts of norms. Testifying to all my disagreements would be weird, rude, and exhausting. I probably have no all-things-considered obligation to voice my disagreement with a stranger I overhear on the bus, for example. Considerations of politeness and practicality will override what obligation I do have. I can imagine a case (say, of overt harmful racism or verbal abuse) in which my obligation is not overridden in such a context, but these cases are in the minority. I'll return to this in more detail below, but for now let me note the following: the obligation that I'll be defending is a prima facie duty to voice disagreements – one that is more or less easily overridden in various contexts.

How might the features of a context affect the strength of an obligation? Considerations of manners, morals, practicality, and propriety are stronger in some contexts than in others. At a dinner party hosted by my spouse's boss, practical and conversational considerations will probably outweigh many of my obligations to voice my disagreements. If, however, I can do so at little cost to manners or propriety, or if the subject of disagreement is particularly important, my obligation may be mitigated but not outweighed.

Put a slightly different way, sometimes obligations are (or at least appear to be) in conflict. In some contexts, my obligation to protect someone's feelings, to advocate for myself, or to reach some desirable conclusion might conflict with my obligation to voice my disagreement. If I'm not terribly confident in the belief I'd be voicing, my obligation is easier to override. If I have a high degree of credence, and the stakes and conflicts are comparatively low, my obligation is stronger than if I barely count as believing the proposition and expressing it would be rude. In cases like this, the conflicting obligations must be weighed against one another to determine what ultima facie obligations I have. Figuring out my ultima facie obligation may be complicated, and will certainly depend on the details of my situation.

These conflicting obligations, however, press us to address a further question: are the obligations to testify merely moral obligations to do something epistemic? Or are we epistemically obligated to voice (at least some of) our disagreements? We'll turn to that in the next section.

3. Distinguishing Epistemic and Moral Obligations

We might coherently hold that Lucy and Ethel have an obligation to voice their disagreement and yet deny that this obligation is particularly epistemic. Perhaps it seems that Lucy and Ethel have a moral obligation, and it is merely incidental that the action they are obligated to perform has an epistemic flavour. The fact that they're obligated to testify does not, by itself, demonstrate that they have an epistemic obligation.Footnote ⁹ The subject matter, and the consequences of The Bell Curve make the case for a moral obligation easy – the book contributed to and appeared to justify the continuation of the systematic subordination and disenfranchisement of African Americans, and those racialized as black. This is seriously immoral, so Lucy and Ethel quite plausibly have a moral obligation to stop it if they can. But I don't think this accounts for all the obligations they face.

To bolster the case for epistemic obligations to voice disagreement it would help to have an example that prompts the intuition that there is an obligation to testify, but where there are no (or no major) moral stakes.Footnote ¹⁰ So, imagine that you're having lunch with your friend Stella, the mechanic. You mention that you're planning to replace your brake pads to fix the squealing sound your car makes. Stella thinks that this is a mistake. She does not believe that replacing brake pads will fix squealing. Nonetheless, she does not say anything to you about her disagreement. She remains quiet, or makes non-committal noises until the subject changes.

It is plausible that in a case like this Stella has an obligation to tell you she disagrees with you. And yet, this does not seem like a case with major moral stakes. This suggests that Stella's obligation, if indeed she has one, is epistemic.

On the other hand, while this case is not of moral import, perhaps we take it to have too much prudential import to cleanly demonstrate that Stella's obligation is epistemic. I don't find this convincing, myself. Perhaps you prudentially ought to get more information before you make your purchase, and perhaps you prudentially ought to ask Stella for her help, rather than merely assert your plans to her. Whatever you should prudentially do, it doesn't seem that Stella has prudential obligations. However, let's grant, for the sake of argument, that Stella's case is still too fraught to demonstrate that there is an epistemic obligation here.

So, we need an example in which there are neither major moral nor prudential stakes, and yet there is an epistemic obligation. So, imagine that Carla, a historian of ancient art, is giving a presentation to her research group. She is a well-respected, emeritus scholar, speaking to other well-respected emerita faculty. As a small aside of her presentation she makes the claim that Babylonian architecture influenced Assyrian ceramic art. Carla's colleague, Theresa, disagrees. Theresa believes that Babylonian architecture did not influence Assyrian ceramic art. Nonetheless, she remains silent, even through the question and answer portion of the presentation. She does not voice her disagreement.

Here we have a case with no major moral or prudential stakes. Carla and Theresa will, in all likelihood, face no practical consequences as a result of this (lack of) conversation. Carla may independently decide that that part of the presentation lacks evidentiary support and leave it out in future work. Or she may bolster it with footnotes and further claims. Theresa might do some investigating of her own, or she may recall that she never cared much about Assyrian ceramics anyway. At any rate, it is unlikely that correcting this aside has great prudential import. So, if Theresa has an obligation to voice her disagreement, and neither prudential nor moral stakes can account for this obligation, then the obligation is probably epistemic.

Perhaps it seems more plausible that Theresa, like Stella, has an obligation in so far as she is an expert. That is, expertise might confer special obligations, and this explains why we don't feel that Theresa does as she ought when she fails to testify. While I'm not inclined toward this explanation of Theresa's and Stella's obligations because I don't think that the obligations they have are particular to experts, I don't think it is entirely off the mark. Expertise plausibly affects the strength of one's obligation to testify. But notice, even if this explains the strength of their obligation, this does not mean that the obligation to voice disagreement is not epistemic. After all, expertise is an epistemic category. If Theresa's expertise generates her obligation to testify, then this is just an example of the kind of obligation I'm getting at. An obligation due to expertise just is an epistemic obligation.

Of course, we may doubt that stories or examples like these help us cleanly distinguish epistemic obligations from other kinds of obligations. It may be that our intuitions about Stella, Theresa, and the rest are insufficiently fine-grained to help us in such an analysis. I think there is something to this concern. Indeed, I think that there is probably not much to be gained from drawing too sharp of a line between kinds of obligations. And I take myself to be in fairly good company on this score: Miranda Fricker, in her 2007 text, argues for a similar point with respect to her idea of testimonial justice.Footnote ¹¹ Fricker argues that there are so-called hybrid virtues – virtues that are both epistemic and moral. When we are concerned with truth, the virtue's epistemic side is more apparent, and when we're concerned with ethics, the virtue is apparently moral. The hybrid virtue doesn't bother Fricker, so perhaps hybrid obligations need not bother us.

If we agree with Fricker that the line between the moral and the epistemic is not as thick or sharp as it might appear, then there is good reason to think that we need not distinguish our moral from our epistemic duties too cleanly. At least Fricker's conception of a hybrid virtue opens the conceptual space for a similarly hybrid obligation. All I want to claim, here, is that the obligation to testify to disagreement is an epistemic obligation, even if it is also a moral one.

4. Discharging the Obligation

Let's grant, for now, that we have a defeasible epistemic obligation to voice our disagreements. When these obligations aren't overridden, how are we to discharge them? What, in other words, must a speaker like Theresa do to meet her obligation (assuming it isn't overridden)? While the details of this are likely to depend on the situation, in this section I'll hypothesize about the generalities.

In general terms, to discharge one's obligation to voice disagreement with proposition p requires that one add to the conversational scoreboard some content that appears to be incompatible with p. I'll spend the rest of this section explaining and (to some extent) defending this hypothesis.

According to this requirement, an agent like Theresa must put forward some content in order to discharge her obligation. Putting content forward is the correct level of specificity for this requirement. While I'm calling this an obligation to voice disagreement, the agent need not literally use her voice. Signs, signals, notes, and gestures are all possible methods of discharging one's obligation. We might imagine Carla taking a poll to see which of her audience members disagreed with her. If she did this, then Theresa's raised hand would surely discharge her obligation. We can also imagine making a face or hand movement to express disagreement. Demanding verbal expression, in some cases, is asking too much. On the other hand, one must put forward some content or other. Theresa disagreed with Carla's claim about Babylonian architecture, and her speech action should (to some extent) convey that.

Still, it doesn't seem like the only way for Theresa to discharge her obligation is to say ‘Babylonian architecture didn't influence Assyrian ceramic art’, or ‘I disagree about the influence on Assyrian ceramic art’. Though these responses might be sufficient, they are not clearly necessary. Theresa could, instead, say, ‘I think you should check into your sources on Babylonian architecture’, or ‘Some of the inferences about influences seemed off to me’. It is too specific to require speakers to voice not p when they disagree with an expression of p, or that they make explicit that they are disagreeingFootnote ¹².

Surely, though, Theresa can't express just any content. There are many utterances Theresa could make that would not discharge her obligation. It won't discharge her obligation to ask, ‘Do you think that Babylonian architecture influenced Assyrian art?’ in most cases (though asking ‘Do you really think that…’ might). If she says, ‘I read that Babylonian architecture was terrible’ she probably doesn't do the job either. What she expresses must be (or appear to be) incompatible with the objectionable content to some extent.

Also, if Theresa tells her students, or her partner, or her barista some content that is incompatible with Carla's aside, she also does not seem to have discharged her obligation. If she does this, Theresa has added her belief to a conversation, but not to the same conversation that inspired her disagreement. One way to put this is that Theresa's contribution is made to a new or different conversational scoreboard.Footnote ¹³ So, this part of the requirement means that the content must be put on to roughly the same conversational scoreboard as the objectionable content.

There are, however, pragmatic concerns about voicing objectionable content, given facts about social subordination and privilege: It can be pragmatically difficult, especially for members of socially subordinated social groups, to voice disagreement with content, particularly content expressed by those in positions of privilege. This can be because of the ways in which members of these social groups are socialized, or because of the ways in which speech acts by members of these groups are perceived. But it would be a knock against the hypothesis if it placed unfair burdens on agents who already experience systematically unfair social and epistemic situations. If Lucy, for example, is a person of colour, it might be more pragmatically difficult for her to voice her disagreement, especially if she were obligated to voice her disagreement to Charles Murray or some other powerful white scientist.

I think there are two ways to go in light of this concern. First, it may be that Lucy can discharge her obligation by voicing her disagreement to some but not all of the participants in the original conversation. If she can tell some other scientists, some known allies, or even members of her own lab that she disagrees, that plausibly discharges her obligation. Second, it may well be that Lucy's obligation to voice disagreement is outweighed by the practical considerations in this case. There will certainly be contexts in which voicing disagreement is too risky, given unequal social powers and privileges. There are probably contexts in which one or the other of these options will be appropriate for disagreeing agents.

There's a further concern, here. Returning to our art history example, surely Theresa isn't obligated to express her disagreement with every potentially objectionable proposition. Many utterances could be understood to mean, presuppose, suggest or implicate something objectionable. This is because what is said when someone utters some content, is, famously, vexed.Footnote ¹⁴ Theresa's obligations to voice disagreement don't extend to all possible content that might be expressed or implicated.

An example will help us here. Imagine that Theresa and her friend Roger have a running feud over ancient art such that any mention of Assyria in a conversation with Roger would be objectionable to Theresa because of the implications that content would have in that context. If Roger is not at Carla's lecture, then these implications plausibly will not arise. In a case like this, Theresa need not voice her objection just because the content would be objectionable in some other context.

The most natural and precise way to address this concern and to specify which objectionable contents warrant voiced disagreement is to turn, again, to the conversational scoreboard, or to what Stalnaker calls the common ground. The common ground of a conversation is, according to Stalnaker, the background information of that conversation – what the conversers take, or act as if they take for granted.Footnote ¹⁵ For a proposition p to be on the common ground is for the conversers to presuppose that p and to presuppose that their conversational partners also presuppose that p.

When Carla says that Babylonian architecture influenced Assyrian ceramic art, she attempts to add that content to the common ground. Because of the meaning of the parts of this sentence, the presupposition that Babylonian architecture preceded or was concurrent with the development of Assyrian art is also put forward for inclusion in the common ground. Perhaps so too is the presupposition that an artist in Assyria was aware of Babylonian architecture. I'd like to suggest any content that is in the common ground or that a converser has attempted to add to the common ground is a candidate for objectionable content. Hearers who disagree with this content, then, would be at least prima facie obligated to voice that disagreement. This means that the content that would be generated in a similar conversation with Roger is not on the common ground in the conversation with Carla. This is because Carla would not take those propositions to be added to the common ground, and Theresa would not expect her to. So, Theresa need not make her opinions about Babylonian architecture known generally – she need not make known all possible disagreements, but the obligation to voice disagreements ranges over objectionable content that is part of, or is put forward to be part of the common ground.

It should be noted that conversational scoreboards, like conversations, are messy and their boundaries are vague.Footnote ¹⁶ There might be cases in which Theresa could discharge her obligation by voicing her disagreement to an audience whose membership was not identical to Carla's. If, for example, some of the faculty leave Carla's talk before the Q and A, Theresa is not obligated to track them down and tell them about Assyrian art – the conversation and the conversational scoreboard intuitively survives the loss of those members. On the other hand, if there is no overlap in audiences then presumably Theresa's obligation is not discharged. Many cases, however, will be in between these extremes.

It would also be asking too much of Theresa that she convince Carla, or anyone else, to agree with her about Babylonian architecture. If Carla and her cohort are intransigent, surely that doesn't reflect on Theresa's epistemic behaviour. Further support for this point comes from the fact that we sometimes disagree with people who aren't in a position to be convinced. Herrnstein, for example, died soon after the publication of The Bell Curve. So, convincing him by voicing disagreement is too much to ask.

It is more plausible that sometimes intelligibility by one's interlocutors or reasonable exchange with them is required. If Theresa voices her disagreement to her English-speaking colleagues in Danish, she's not discharged her obligation. And, plausibly, if she yells ‘Nuh-uh!’ and then runs out of the room, her obligation goes unfulfilled. As we'll see in the next section, though, the details of this will depend on the source of the obligation.

And, finally, according to my hypothesis, the content that Theresa expresses must appear to be incompatible with the objectionable content. I hedge on requiring incompatibility because incompatibility is not always obvious. Sometimes we can be surprised by what is compatible with what, and it is too much to ask that Theresa be sure of the logical consequences of the content she disagrees with and the content she uses to express that disagreement. It may be enough (in many contexts) that the content she expresses (if true) makes the objectionable content less likely. The degree of incompatibility required will vary from context to context, determined by the details of the conversation and beliefs of the participants. If Theresa and Carla share a false presupposition that p and q are incompatible, and Carla has expressed that p, Theresa can discharge her obligation by expressing that q. This suggests that the appearance of incompatibility is sufficient.

So, for these reasons, I suggest that for an agent to discharge her obligation to voice disagreement with proposition p requires that that agent add to the conversational scoreboard some content that (at least) appears to be incompatible with p.Footnote ¹⁷ This might seem unsatisfyingly vague, but as above, there are going to be considerations that are specific or internal to particular conversations that will bear on the details of when contents count as disagreeing, and so when people expressing those contents have discharged their obligations. Further complicating matters is the fact that here, as in the discussion of convincing interlocutors, the particular details of the obligation may depend on the source thereof. That is, the different sources will determine different kinds of obligations, obligations that arise in different contexts, and obligations with different requirements. In the next section, I'll look at the most plausible candidate sources.

5. The Source of the Obligation

If we grant that there is a prima facie epistemic obligation to testify as to our disagreements, the next natural question is this: what generates this obligation? There are at least four plausible ways we might go about answering this question. One way (following Fricker) is to understand our obligation as self-regarding – to see it as a way to build the epistemic skills necessary for epistemic well-being. Another way (following Catherine Elgin) is to see it as an obligation incurred in virtue of participating in inquiry. A third way (following Margaret Gilbert) is to understand the obligation as part of a commitment to joint action. A fourth source of the obligation is to take a lesson from Mill seriously – namely that beliefs and actions cannot be justified without being discussed and submitted to scrutiny. While these four views are more or less compatible (that is, some combination of the four could be true/generating the obligation), I'll treat them each separately for clarity.

Fricker offers a theory of epistemic justice and injustice.Footnote ¹⁸ In so doing, she describes a variety of ways in which an agent might be better or worse off epistemically. One component of epistemic well-being is that agents are able to testify and to have their testimony taken seriously – they are able to participate in the epistemic economy. This is part of why, when an agent's testimony isn't taken seriously due to identity prejudices, that agent is the victim of epistemic injustice. One of the dangers of epistemic injustice is that the victims thereof lose confidence and internalize the credibility deficits they suffer in the eyes of their prejudiced interlocutors. This makes them worse off, epistemically. Voicing disagreement, in certain contexts, could help agents avoid some of these pitfalls. Recall the example of Stella the auto mechanic. Auto mechanics is a male-dominated field. It seems likely that, in many cases, women's testimony hasn't been properly attended to in conversations about auto mechanics. Their testimony has either been pre-empted, or hasn't been given its due epistemic weight. We can imagine that Stella has experience with this. Say that Stella doesn't voice her disagreement about your brakes because she lacks confidence. Voicing her disagreement could be instrumental in overcoming this lack. And, if this is the case, then agents like Stella have an obligation to themselves to testify that they disagree.

So, perhaps the epistemic obligation to voice disagreement is self-regarding, and grounded in the conditions for epistemic thriving. This yields a fairly far-reaching obligation – all agents whose epistemic well-being could be improved by voicing their disagreements are obligated to do so. And many agents are in this position.

To see how considerations of epistemic flourishing generate an obligation to voice disagreement, let's return to the case with which we began: Lucy and Ethel and The Bell Curve. Recall that one component of epistemic well-being is participation in the epistemic economy. One way for Lucy and Ethel to do this is to publish editorials. However, if they're not sincere in their contributions to the economy, it's not clear that they're flourishing epistemically. And, if they're not adding their sincere beliefs to the conversation, it is not clear that they're contributing to the epistemic (rather than, say conversational or rhetorical) economy. A sincere contribution, especially when objectionable content is being considered, is a better way to contribute.

Another component of epistemic well-being is epistemic justice. Suppose that Lucy is a member of a socially subordinated group (this is plausible if she is a woman in the sciences), and that part of her social identity involves being subject to a credibility deficit. If this is the case, and if voicing her disagreement is one way for her to participate in the epistemic economy, then she has a defeasible obligation to do so. And if Fred is not a member of a socially subordinated group, then for him, this self-regarding obligation does not arise (at least in this case), as he does not need to voice his disagreement to overcome an unfair perceived credibility deficit.

Of course, this depends on a contingent generalization about psychology: the claim that voicing disagreement develops one's ability to participate in the epistemic economy and to overcome perceived credibility deficits. It seems like a plausible claim to me, but psychological work would have to be done to confirm it.

Another worry, here, is that sometimes voicing disagreement could epistemically harm someone. If I am teaching a class and a student expresses some objectionable content, I might epistemically harm them if I voice my disagreement, especially if that student already faces some identity prejudice involving credibility deficits (like stereotype threat, or imposter syndrome). Don't I have an epistemic obligation, then not to voice my disagreement?

I think not. There is a difference between an obligation to stay quiet and an obligation to be kind, polite, or pedagogically responsible. In this case, while I might have a good reason to be careful about how to voice my disagreement, the obligation to voice it remains. And in most cases our obligations to keep quiet won't be epistemic, but rather will stem from the overriding of our epistemic obligation to voice disagreement by moral or prudential considerations.

So, an account like Fricker's generates an epistemic obligation to voice disagreement that agents have insofar as voicing their disagreements helps them develop epistemically, participate in the epistemic economy, and avoid the harmful effects of negative identity prejudices. However, if the obligation indeed comes (solely) from this source, the obligation is contingent on the psychological facts of the agents involved. That is, if Stella's confidence won't be improved by voicing her disagreements, then she is not obligated to do so. This might be a palatable result, but we might plausibly want something that is not dependent on agent psychology in this way.

The second possible explanation for the obligation is inspired by the work of Catherine Elgin. Elgin argues that, ‘Science requires collaboration. [And] that collaboration requires trust’.Footnote ¹⁹ Science requires collaboration and trust, she argues, for the same reason it requires models, inference, and idealizations: the subject matter is too complicated, and our findings are too fallible for any scientist to make much progress on her own. She must be able to trust her fellow scientists to let her know when she's in error, or going astray. In other words, she needs to be able to trust her fellow scientists to voice their disagreements with her findings or her methods. And she needs to be able to rely on their collaborative contributions.

Elgin's argument extends beyond the natural sciences. She argues that pursuits like arts and auto mechanics make similar requirements on those who pursue them.Footnote ²⁰ Perhaps participation in these pursuits, then, generates the epistemic duty to voice disagreements: collaboration and trust are unlikely if agents keep disagreements to themselves. Carla would probably not value presenting to her peers if they did not make a habit of voicing their disagreements. And in his work on The Bell Curve, Panofksy argues that concealed disagreements contributed to the collapse of behavioural genetics as (what he calls) a well-behaved science.Footnote ²¹ Agents who pursue collaborative inquiry, then, are obligated to voice disagreements because of the basic requirements of that pursuit.

One kind of trust involves believing that people are sincere in their utterances. When trustworthy people assert, they believe the content asserted. When they ask questions, they're not sure about and are interested in the answers. This is similar to the kind of trust described by Hinchman and Moran in their interpersonal accounts of assertion.Footnote ²² This minimal trust is not the kind of trust that Elgin has in mind – it would not be sufficient for science to function. Science requires a more substantial kind of trust.

Science, and inquiry more generally, requires substantive trust because the subject matter is too broad and the literature too vast for any one scientist (or even team) to read and assess it all. A scientist can't reproduce all of the experiments on which their work is based and still have time to do innovative and creative work. Thus they have to trust the process that reviews and publishes scientific findings. This trust requires that, when people disagree with findings or methods, those disagreements are voiced. For the scientific process to be reliable and to proceed, scientist are required to voice their disagreements – the process generates an obligation to do so. And, as inquiry is an epistemic pursuit, so too is the obligation it generates.

Similarly, collaboration generates a requirement to voice disagreement. Indeed, it is hard to see how someone is collaborating at all if they don't make their disagreements known. Perhaps collaboration alone won't generate Stella's obligation to voice her disagreement in the brakes case (unless conversation counts as a science-like inquiry), but surely it would be sufficient to generate obligations for Theresa and for Lucy and Ethel.

Unlike in the case of the Fricker-style account above, here the inter-personal efficacy of the testimony matters. Because the voicing of disagreement is instrumental to the functioning of science, it matters that it contributes to that function, at least enough of the time. And for that to be the case the disagreement must alter the conversational scoreboard in some way.

It might be objected that Theresa and Carla, as colleagues, have only a very slight acquaintance, while Lucy and Ethel have none at all with Murray and Herrnstein. What kind of trusting relationship should we expect (and require agents to honour) between strangers and mere acquaintances?

The kind of trust in question, and the kind that Elgin has in mind, is not the kind of personal trust that develops between intimates. Nor is it an unrestricted trust that extends to all parts of the agents’ conducts. Instead it is an impersonal trust, which is entirely independent of the familiarity or the relationship between the person who is trusted and the person who is trusting. It is a kind of trust that is more than the minimal trust described by Hinchman and Moran, but less than the kind of trust required to make people deeply connected and intimate. It is the kind of trust one would have of a tool or method a scientists has used with good results. Any scientist should, according to Elgin, be trustworthy and (at least to some degree) trusting in this way. To be otherwise is to commit scientific misconduct and would be wildly inefficient.

So, Elgin's account generates an epistemic obligation to voice disagreement when participating in science or other inquiry-based practices (broadly construed). This obligation is generated by the basic requirements of the practices themselves. These practices require collaboration and trust, each of which is only possible in communities in which disagreements are voiced. This means that, according to Elgin's account, the epistemic obligations only arise in those conversations that take place in communities that require these features. Perhaps this is reassuringly narrow – perhaps this focus is a benefit – but this source certainly generates an obligation for fewer agents than that which is generated by Fricker's account.

Third, perhaps our epistemic obligation to voice disagreements is generated by a joint commitment – here we're following Margaret Gilbert.Footnote ²³ My reconstruction of Elgin, above, claimed that mere participation in a sphere of inquiry generates obligations for participants. It might seem strong, however, that mere participation in the same pursuit generates obligations for agents to voice their disagreements. We might think that obligations like these are more or less operative given certain commitments agents have made (or perhaps they're only operative when commitments have been made). The intuition might be that I am not obligated to do anything regarding someone else's expressed position unless I have some kind of commitment to them or for which they are relevant. Gilbert's account of joint action offers an understanding of our obligations to voice disagreement that is based on commitments.Footnote ²⁴ According to Gilbert, when agents make joint commitments, the individuals are committed to doing something together – not merely at the same time, but as members the same group. These commitments generate obligations for the agents involved.Footnote ²⁵

To see how this source generates an obligation to voice disagreements, it will be helpful to look at one of Gilbert's central examples. She illustrates what it is to jointly commit to some activity by contrasting taking a walk beside another person and taking a walk with another person. When you and I are taking our lunchtime walks, and our paths overlap for a ways, we're walking besides one another. If I change my mind and go home early, or take a different route, I've done nothing wrong and owe you no explanation, because our overlapping paths don't generate any obligations for me (except, perhaps, staying out of your way or being polite). If, however, we're walking together, the situation and my obligations are different. If I invite you to take a walk with me, I must explain or excuse myself before changing destinations, or quitting and going home. Our joint commitment has generated obligations for me (and for you too). While there are many more details to Gilbert's analysis, this difference is symptomatic of a joint-commitment.

Agents who have jointly committed to some activity have obligations that are specific to that activity. In the case of you and I walking together, I have the obligation (among others) to match my pace to yours, or to ask you to slow down. You have the obligation to wait for me while I stop to tie my shoe, etc. And other activities will generate other activity-specific commitments. If we're jointly committed to organizing a conference, then we have conference-organization-specific obligations. And if we're jointly committed to some science or area of inquiry, that will generate some obligations specific to that activity.

It is easy to see how Teresa and Carla are jointly committed to an area of inquiry. Part of what explains Carla's talk is that she is seeking input on her inquiry into art history. The function of the practice of giving academic talks to one's colleagues is, at least in many departments, to get feedback on one's work and ideas. Teresa, by attending the talk, is committed to working on that with Carla. This is what makes academic talks different from traditional plays or performances. The oboist playing a concert at the conservatory is not (or at least not obviously) committed to any joint activity with her audience. A fellow expert in the oboist's audience might have feedback (and notice, if she doesn't offer this feedback there's the sense that she's failed to do as she ought), but the oboist doesn't give a performance in order to get constructive criticism from her audience members. Teresa and Carla, on the other hand, are jointly committed to scholarship. This generates some commitments and some obligations for them.

Does a joint commitment to inquiry generate an obligation to voice disagreement? I think it does, for at least two reasons. For one, as Elgin points out, inquiry is an activity that benefits from (perhaps even requires) honest voicing of disagreements. Second, voicing disagreement is the sort of behaviour that could help jointly-committed agents come through on their joint commitments to inquiry. Another example of joint commitment will help us to see why.

Say that Jamie and Jean decide to buy a house together. They jointly commit to paying the mortgage. This generates all kinds of mortgage-paying-specific commitments for each of them, and neither can decide to stop paying into their shared account without telling the other. Imagine, however, that Jamie has recently started going to the local casino. Jamie hasn't yet failed to pay into the shared account, but Jean is concerned. Does Jean, in light of the joint commitment to pay the mortgage, have an obligation to express her concern to Jamie? It seems plausible that she does. It seems especially plausible given Michael Bratman's account of shared cooperative activities. According to Bratman, one of the features of a shared cooperative activity (like paying a mortgage or studying some area) is a commitment to mutual support.Footnote ²⁶ A commitment to mutual support requires that jointly committed agents are each committed to supporting the efforts of the other to play her role in the joint activityFootnote ²⁷. So, if Jean takes Jamie to need support to avoid gambling, then Jean should tell Jamie that she's worried or otherwise convince her to save her money.

Similarly, if Theresa and Carla are jointly committed to art history, they are also (if Bratman is right) committed to mutually supporting one another in the aims of that activity. Given their disagreement, Theresa may take Carla to be pursuing a sub-optimal line of inquiry or resting on false premises in her paper. If she does, then their joint commitment to art history and the commitment to mutual support that that joint commitment requires generates an obligation for Teresa: she must point out (what she takes to be) Carla's error as part of supporting Carla in her efforts to discover art-historical truths.

So, perhaps we can understand the obligation to voice disagreement as part of and as generated by the joint commitments undertaken by the obligated agents. This would significantly limit the number of such obligations we have – Theresa and Carla are jointly committed to scholarship, but you and your mechanic friend Stella haven't clearly undertaken such a commitment. So, Theresa is obligated to voice her disagreement, according to an account like this, while Stella is not. Whether this limit is a feature or a flaw I leave to future discussion.

Finally, the obligation to voice disagreement might be due to the nature of doxastic justification itself. John Stewart Mill argued that for an agent to be justified in a belief that belief must be subjected to scrutiny and discussion.Footnote ²⁸ Mill argues for this in the context of advocating against prohibitions on free speech however, his reasons have as much to do with epistemic considerations as with political or moral ones. He says,

So, Mill is explicit that wise agents subject their beliefs to scrutiny and discussion. They must not only subject their own beliefs to scrutiny, but must (at least on occasion) voice their disagreements with what they take to be fallacious. And this is for practical and political and moral reasons, but also because such behaviour is a necessary condition of doxastic justificationFootnote ³⁰.

If Mill is right about our fallibility and the nature of doxastic justification in light of that fallibility, then voicing our disagreement contributes to doxastic justification. And, contributing to doxastic justification is contributing to an epistemic good. We are, then, epistemically obligated to voice our disagreement insofar as we are epistemically obligated to contribute to epistemic goods. And this seems like a plausible understanding of epistemic obligation.