Stupefying

1.1 Common Ground and Conversational Tone

To explain in more detail how content is snuck into the common ground, we’ll need to say just what common ground is. We’ll work with the view, due primarily to Stalnaker, that the common ground is the set of propositions which all of the conversational participants accept in a suitably public way.⁷

What does it take for acceptance to be suitably public? One way to think about publicity is on the infinitely iterated common belief model of Stalnaker (2002): each of the interlocutors believes that I accept p, believes that each of the interlocutors believes that I accept p, and so on.⁸ I will leave open whether publicity needs to be understood in this demanding way, or whether we should make do with something less.⁹

And what does it take to accept a proposition? Following Stalnaker, we’ll take this to be a rather thin notion, where one accepts a proposition iff one treats it as true for the purposes of the conversation. One can accept by believing, but acceptance is weaker than belief, in that it includes pretense, supposition, and so on. Not all conversations are serious, sober ones where everyone is taken to believe what they say. But by saying something they are still, in some way, for some purpose, acting as if what they have said is true. In such cases contents can be accepted without being believed, and so can be common ground without being believed.

This introduces a puzzle for us, though. We’re interested in how making additions to common ground could be used to control others. But on its own, something’s counting as common ground in a conversation doesn’t seem to mean much. Acceptance doesn’t require belief, so one may accept something, act as if it is true for the purpose of the conversation, then go on one’s merry way believing the opposite.¹⁰ If Bob confronts Anne with the fact that she was committed in some conversation to p but now seems to be acting otherwise, Anne can sometimes easily explain herself by saying “Oh, we were just messing around. I didn’t actually believe any of that” or “We were just talking about what would be the case supposing that q. I don’t know whether q is true, though, and haven’t made up my mind about p”. How, then, does getting something into the common ground make for an effective means of influence?

The answer: not all conversations are ones of pretense or reasonably regarded as such. Anne’s above excuses won’t get her out of what she said or agreed to in any ordinary conversation, and it would be a poor defense of having given perjurous testimony under oath. We can draw here on what Yalcin (2007, p. 1008) calls conversational tone: the attitude which is commonly known by the interlocutors to be the attitude they will all take towards the content in the common ground.¹¹ In many conversations, it is public information that everyone will come out of the conversation believing what has been established there, unless new evidence appears. So if one gets something into the common ground where the conversational tone is belief, one will have succeeded in directing the judgments and conduct of other people.

1.2 At-Issue/Not-at-Issue Updating

Getting something into common ground, then, can have real influence. But how does one do it? The paradigm way content is added to common ground is through assertion by a speaker and assent by the other interlocutors (or in some cases, arguably, absence of dissent).¹² There may be various means of making assertions, but the usual way it happens in spoken English is by a speaker uttering a declarative sentence with falling intonation. And assent from the addressee is indicated by an affirmative answer particle (yes, okay, right, and so on) or gesture (like nodding), or simply by continuing the conversation in a way that doesn’t reject or otherwise resist the assertion.¹³ If ϕ is the sentence being asserted, what is proposed to be added to common ground is ⟦ϕ⟧, the semantic content conventionally associated with ϕ.¹⁴ If the addressee accepts, into the common ground ⟦ϕ⟧ goes.¹⁵

But this picture is too simple. Often there are multiple contents semantically encoded in a single sentence which have importantly different statuses. Compare the following:

• (3) The modern antivaccine movement can be traced to a paper published 16 years ago in The Lancet, a respected medical journal.¹⁶
• The Lancet, which published a paper 16 years ago to which the modern antivaccine movement can be traced, is a respected medical journal.

Both sentences encode the proposition that The Lancet is a respected medical journal and the proposition that the modern antivaccine movement can be traced to a paper published in The Lancet 16 years prior to the utterance. But they differ in how they present these propositions. In (3-a), the information about the origins of the antivaccine movement is presented as the main issue, whereas the status of The Lancet is backgrounded. This can be seen by considering how (3-a) sounds in response to a question like “Where does the antivaccine movement come from?” (good) and how it sounds in response to a question like “What is The Lancet?” (bad). Things are just the reverse with (3-b). Potts (2005) introduces the term at-issue content for the information foregrounded in this way and not-at-issue content for the backgrounded information. In light of this, we need to make a revision to our view of assertion: it is the at-issue content of an assertion which is explicitly proposed to be added to common ground.¹⁷

I’ll take not-at-issue content to include any other content conveyed by an utterance. This will be a broad and disunified class, including the contents expressed by supplements like nominal appositives and non-restrictive relative clauses (as in (3)), conversational implicatures, and presuppositions. Such breadth makes our next claim relatively easy to defend: some not-at-issue content gets added to common ground.

There are various controversies over just what not-at-issue content gets added to common ground and how it does so, but these will not concern us. All we need is that it sometimes does happen. One relatively uncontroversial case of not-at-issue update to common ground is through presupposition accommodation.¹⁸ When one utters (4), one presupposes that Mary smoked at some point.

• (4) Mary stopped smoking.

That is, one treats it as already established—already part of the common ground—that Mary smoked. But one can use such an utterance in cases where it isn’t already established. In such cases, the presupposition typically becomes common ground, perhaps with the pretense that it was there all along.

Interestingly, accommodation seems to happen by default. Presuppositions are not always accommodated; one can challenge them or outright reject them. But without any explicit challenging, they will typically go into the common ground without any effort from the participants of the conversation, regardless of what happens with any at-issue content of the utterance.

Even in cases where the at-issue content is rejected, the presuppositions will often be accommodated anyways.

• (5) Has Mary stopped smoking?
• I don’t know, I didn’t even realize that she did smoke.

Here the issue of whether Mary has stopped smoking is left unresolved, but the issue of whether Mary ever smoked is tacitly settled, as von Fintel puts it, “quietly and without fuss” (2008, p. 137).

The distinction between updating the common ground through at-issue content and doing so through not-at-issue content (as with presuppositions) has been a crucial one to work on non-ideal speech. Stanley, for example, argues “that propaganda typically affects the not-at-issue content of an utterance. It enters into the common ground by routes other than assertion” (Stanley 2015, p. 172).¹⁹ Langton argues that back-door speech acts—“low profile speech acts, enabled by presuppositions and their ilk, that tend to win by default” (Langton 2018a, p. 146)—are an effective tool for “evil” speech, since they can be difficult to block. This is in part because “[b]ack-door speech acts can have an under-the-radar quality absent in assertion. When someone says ‘Even George could win,’ attention is on ‘George could win,’ not on ‘even,’ still less on ‘George is an unpromising candidate’” (Langton 2018a, p. 159).²⁰

The idea, then, is that we can sneak content into the common ground through the not-at-issue back door. Asserted, at-issue content, however, lacks this ability. I will flesh out this idea more explicitly in the remainder of this section.

1.3 The At-Issue/Attention Link

The account turns on what content speakers are aware of. To track this, we can enrich standard models of discourse with a class of agent-relative modal operators A_i, where A_iϕ says that the agent i is aware of ϕ.²¹ But when does an agent count as aware of a sentence? As we’ll see in §2, there are a few related notions of awareness that will be useful to include in our theories, but for now we’ll follow Franke and de Jager (2011) and others in taking the relevant kind of awareness to be attention. More specifically, we’ll think of it as attention to the content of the sentence; attending to the mere phonetic or orthographic object won’t suffice. A_iϕ, then, says that A is attending to ϕ’s content, where what it takes to attend to a content is to be actively thinking or entertaining it. I attend to the content of ‘It’s raining’ by attentively considering whether it is raining or entertaining the thought that it is raining.²² Even more specifically than that, we want A_iϕ to indicate attention to the content of ϕ as the content of ϕ. I’m not attending to the content of a sentence in the relevant sense if it’s written in a language I don’t understand but I just happen to be entertaining the content it expresses by coincidence. Nor am I attending to its content in the relevant way if I am considering the question whether this sentence, whatever it means, is true. For A_iϕ to hold, i must be entertaining the content of the sentence itself as the content of that sentence.²³

This is close to what we need, but it’s still not quite right. As we’ve seen, sentences often have multiple contents, so it can’t be attention to the content of the sentence that we’re after. What we want instead is something which says which content, if any, an agent is attending to. One way to do this is by indexing our attention operators to contents as well as to individuals, like so:

So for each content c of ϕ, there will be an attention operator for each individual i that says i is attending to c as a content of content of ϕ.²⁴ $A_i^{[\varphi ]}\varphi$, for instance, says that i is attending to ϕ’s at-issue content, [ϕ]. And if π is a presupposition of ϕ, then $A_i^{\pi }\varphi$ says that i is attending to that presupposition.

Now that we’ve singled out the relevant notion of attention, we need to link it to what happens in a conversation.²⁵ As various authors have observed, it’s hopeless to look for an account which fully specifies how attention shifts in response to an utterance. What people attend to varies dramatically from person to person and context to context, and not in any reasonably systematic way. However, there’s one claim that seems to be generally agreed upon by those who have worked on the dynamics of attention in conversation: asserted content is publicly attended to.

Bledin and Rawlins (2016), for example, rely on a principle called Drawing Attention, which says that if a speaker utters ϕ, then the attention state of every other discourse participant a is immediately updated to ensure $A_a^{[\varphi ]}\varphi$, prior to any further processing (such as allowing into the common ground or challenging the assertion in some way).²⁶ For Franke and de Jager (2011), it’s important that “unawareness from inattentiveness is lifted spontaneously whenever agents process linguistic information that contains mention of an unaware contingency” (p. 80). Others concur.²⁷ Accepting any such proposal will imply the following principle:

At-Issue/Attention Link: At-issue content of an utterance is always attended to by each interlocutor before being added to common ground.

What about not-at-issue contents? Some, no doubt, will in fact be attended to. And perhaps certain ones, like those associated with appositives (as in (3)), will be attended to as automatically as at-issue content is.²⁸ But often they are not. Presuppositions, for instance, are not always attended to before they are added to the common ground. This was just Langton’s point. There is thus no strong link between not-at-issue contents and attention.

Not-at-Issue/Attention Separation: Not-at-issue content is not always attended to by each interlocutor before being added to common ground.

What results from these two principles is a plausible account of how conversational short-circuiting can happen. One way to ensure that content is not sufficiently deliberated about before it is added to the common ground is through inattentional short-circuiting: getting it into the common ground without your interlocutors attending to that content. If At-Issue/Attention Link is true, then inattentional short-circuiting cannot be done with at-issue content. However, given Not-at-Issue/Attention Separation, it can be done with not-at-issue content.

Like sleight of hand, inattentional short-circuiting relies on misdirection of attention, the oldest trick in the book.²⁹ While you attend to one thing, the speaker has snuck something else into the common ground. Coming out of the conversation, at least if it had the right conversational tone, you may find yourself with new commitments you didn’t knowingly assent to and new beliefs (if only implicit ones) to which you never gave adequate consideration.

If this is right, then there’s an important means of conversational short-circuiting which is only available for contents packaged in a not-at-issue way. We can see why not-at-issue content would be a more promising vector for short-circuiting, just as Stanley and Langton suggest. It is thus unsurprising that not-at-issue content has been the focus of much of the work on manipulative speech.

This is an attractive account, but as we will soon see, it is wrong.

2.1 Two Kinds of Stupefying

Stupefying involves an addressee accepting an assertion without understanding it. When Eben Byers accepted claims like (1) about why radium water would help him, he was stupefied. Not only did he not understand what was said, but he couldn’t have understood what was being said. Having no idea what ionization or alpha rays are, he lacked the relevant concepts to grasp the sentence’s content. This kind of stupefying, which arises from an addressee’s lacking the conceptual resources to grasp the asserted content, is what I’ll call stupefying by content.

Sometimes we are stupefied not because we can’t grasp the relevant content, but because we don’t recognize that the assertion that’s just been made is a way of expressing that content. Suppose on your visit to the Outer Banks in North Carolina, you ask an Ocracoke Island local whether it’s a good idea to stay at a particular cottage. They tell you it is not, then explain:

• (6) Its pizer is all whopperjawed.

Now, you might ask what a pizer is and what it is for it to be whopperjawed. But you may just as well say “Oh, okay,” accepting (6) without understanding it, then go on to ask about the other place you were considering.

In this latter case, you’ve been stupefied, but not because there was some content which you were incapable of understanding. In the speaker’s dialect, pizer just means porch and whopperjawed means crooked or misaligned.³⁰ You can, of course, grasp the proposition that the cottage’s porch is crooked. Unlike the Byers case, it’s (6)’s form rather than its content that prevents you from understanding it. This is stupefying by form.

Linguistic ignorance is not the only way stupefying by form can occur. Someone may have put things in such a convoluted way that processing it would take more effort than one can or is willing to expend. Or they may have simply mumbled. Again, in cases like these, one might follow up with a request for clarification, but often one instead accepts without understanding.

These two kinds of stupefying correspond with ways awareness operators like $A_i^c$ have been interpreted in the economics and computer science literatures.³¹ Above we noted that besides the attentional sense discussed in linguistics, it is sometimes understood in other ways. Indeed, usually the notion is introduced as being about “lack of conception” (Schipper 2015, p. 77), using examples like this one:³²

How can someone say that he knows or doesn’t know about p if p is a concept he is completely unaware of? One can imagine the puzzled frown on a Bantu tribesman’s face when asked if he knows that personal computer prices are going down!

(Fagin and Halpern 1988, p. 40)

This is very close to stupefying by content. One can just as well imagine the Bantu tribesman saying that yes, of course he knows that, despite not being able to grasp the relevant content due to lack of the relevant concepts.

The other way ‘awareness’ is sometimes interpreted is as an ability to process, on which “an agent is aware of a formula if he can compute whether or not it is true in a given situation within a certain time or space bound” (Fagin and Halpern 1988, p. 41). Work on this typically focuses on issues of computational complexity and resource bounded algorithms for computing truth in interpreted formal languages, but for application to natural language discourse one could incorporate linguistic ignorance, facts about human sentence processing, and features of the form of an utterance which can affect its interpretability in a given context. In this sense of awareness, you lacked awareness of (6), even though you are aware of it in the conceivability sense.³³

Observe that both of these new kinds of unawareness imply the inattentional sense of unawareness we were working with in §1. One cannot attend to the content of a sentence in the relevant sense if one cannot process the sentence or if one cannot grasp the sentence’s content. This means that when someone has been stupefied by either form or content, one accepts a sentence without attending to its content.³⁴

And note that the unawareness of the stupefied concerns the at-issue content of those utterances which they are accepting. Byers’ incomprehension stemmed from not having the concepts for what is expressed by ionizing processes or alpha rays, the meaning of which contributes squarely to the at-issue content of (1). Because of this, he could not have attended to this content. And because you didn’t know the meaning of pizer and whopperjawed, you couldn’t attend to the at-issue content of (6). Assuming that acceptance of these utterances resulted in additions of their at-issue contents to common ground, as acceptance of assertions generally does, stupefying involves addition of at-issue contents to common ground without their being attended to. This would mean that the At-Issue/Attention Link does not hold, and we should not expect inattentional short-circuiting to happen only through not-at-issue content.

2.2 Stupefying and Common Ground

We may doubt, though, that these contents really are added to common ground, despite the fact that they apparently have been asserted and affirmed. I will raise and address two reasons for so doubting.

The first concerns the relation between speech acts and addressee understanding. On one prominent view, going back to Austin (1962, pp. 116–117), in order for a speaker to perform an illocutionary act like warning, ordering, asking, or refusing, the force and content of their utterance must get uptake from their audience: its force and content must be understood.³⁵ The actor on the stage yelling “Fire!” does not really warn the audience if the audience does not recognize the utterance as a warning.³⁶ Nor would they have warned the audience if the audience took it as a warning but understood it to be a warning about the presence of a ϕ-er. In these cases, there is no warning, only an attempt to warn.

In cases of stupefying, the content of the utterance is not understood, and so does not get uptake. This means that the illocutionary act has failed to be performed. There was no assertion after all, just an attempt at assertion. But if the content was not asserted, nobody has even proposed that it be added to the common ground, thus we shouldn’t think that it actually has been added to it. So goes the objection.

Depending on how we draw the boundary between illocutionary and perlocutionary acts, I am willing to grant that understanding may be required for some kinds of illocutionary acts to be performed. However, I do not grant that uptake is required for the performance of all illocutionary acts. It seems clear that you were told that the house has a whopperjawed pizer and even clearer that the speaker asserted that it does.³⁷ We might still think uptake is required for an assertion to count as fully successful, but it seems not to be required for an utterance to be an assertion.³⁸ We must either deny that uptake is required for assertion or else weaken the notion of uptake so that it does not require understanding of the content. Either way, the objection is answered.

The second objection is this: if the addressee doesn’t grasp the at-issue content, then they don’t believe it. But if they don’t believe it, it can’t be one of the things believed by all the participants of the conversation, and so can’t be part of the common ground, even if an addressee has affirmed an assertion of that content.

Recall, though, that it’s not belief that is necessary for common ground status, but acceptance. And while belief may require grasp of the relevant content,³⁹ acceptance should not.⁴⁰ We do need to have something that connects the agent to the content they accept, otherwise it would be a mystery why they would count as accepting this content rather than some other. But this can be satisfied in various ways, such as the addressee thinking of the content under the description “whatever the speaker just said, whatever it means”. Similarly for whatever extra is required for making the acceptance suitably public. We can commonly believe that we all accept whatever the speaker just said, even if not all of us understand what it meant. So it is possible for there to be contents in the common ground that not everyone in the conversation can grasp.⁴¹

Nevertheless, even granting that it’s possible, we might still doubt that the at-issue content of (1) or (6) do in fact go into the common ground when addressees affirm them without understanding. Perhaps we can make do with less. Here are a couple options for doing so. We might take the least common denominator view, according to which all that goes into the common ground in these cases is whatever parts of the assertion’s content that both the speaker and the addressee recognize (or perhaps: recognize in a suitably public way).⁴² So in the case of (6), all you understand of the utterance is that the cottage has some thing with some property that makes it unsuitable for you to stay there. And all Byers gets from (1), presumably, is that Radithor does something that will bring about astonishing results.⁴³

Alternatively, we might adopt the metalinguistic view, which says that in cases of stupefying, what gets accepted isn’t the at-issue content of the assertion, but rather an associated metalinguistic content, like (7).⁴⁴

• (7) That utterance of “Its pizer is all whopperjawed” is true, whatever it means.

With either the least common denominator or metalinguistic view, the at-issue content of the utterance doesn’t go into the common ground. This would mean stupefying would not be a way of inattentional short-circuiting with at-issue content, since the content that the speaker can’t attend to never makes it into the common ground. If this were right, we could retain the At-Issue/Attention Link. We should reject these alternative views, however, since there is good reason to think that the at-issue content does get into the common ground.

There are various ways to act as if some content is true for the purposes of a conversation. Some of these will involve understanding the content and acting in ways that would be rationalized by belief in such a content, like taking out an umbrella because one accepts that it will rain soon. But for conversation, the primary ways of acting as if some content is true are tied to linguistic conventions. One important way to act as if p is true is to assert something that means p (i.e., that has p as its at-issue content). Another is to assent to someone else’s assertion of something that means p. In cases of stupefying, one assents to assertion of a sentence that means p, so one acts as if p is true for the purposes of the conversation. So one accepts that content. And this assent is suitably public, so one’s acceptance is enough to get that content into the common ground, given that the speaker also publicly accepts it. So we should reject views which imply stupefying doesn’t get the at-issue content of the accepted assertions into the common ground.⁴⁵

Here’s another argument, primarily against the least common denominator view. Explaining presupposition licensing is one of common ground’s central theoretical roles. One way to test for whether something has made it into common ground, then, is to check whether it can thereafter be presupposed without accommodation. And stupefied acceptance licenses the presuppositions we’d expect if their at-issue contents had been added to the common ground. For example, (8) is a natural follow- up question you (or some other participant of the conversation) might ask after you’ve accepted (6).

• (8) Does the other cottage down the road have a whopperjawed pizer, too?

This presupposes that something salient besides the other cottage down the road has a whopperjawed pizer. Presumably it is licensed here because the proposition that the first cottage has one is already part of the common ground, rather than through presupposition accommodation. Having already accepted (6), it would be strange to challenge the additive presupposition of (8). So it is important that we take the common ground to have been updated with the at-issue content of (6), rather than something more general.

The objector would likely try to apply the same strategy to reinterpreting the effects of (8) that they applied to (6). The least common denominator proponent might claim that this question is just asking whether there’s something wrong with the other cottage, and so has a similarly broad presupposition, one that would be satisfied by what they take to have been added by (6). But this cannot be right. Contrary to what this proposal predicts, (8) could not be correctly answered by (9).

• (9) Yes, its air conditioning is broken.

The metalinguistic view fares better here. The reinterpretation of (8) would be that it is asking whether “has a whopperjawed pizer” applies to the other cottage, and presupposes that it applies to something else salient. With the acceptance of the metalinguistic interpretation of (6), this presupposition would be satisfied, but (9) would still be a bad answer, as desired.

There’s an additional reason to reject the metalinguistic view, however. According to this view, when the addressee doesn’t understand the at-issue content of an utterance, the at-issue content doesn’t get added to the common ground, only metalinguistic information does. But understanding an utterance is a matter of degree. Consider Burge’s case of a man who thinks he has developed arthritis in his thigh, but seems to have a number of other reasonable and correct beliefs about arthritis (Burge 1979, p. 77). When he was told by his doctor beforehand that he has arthritis in his wrists, he no doubt understood what was said less well than his doctor did, but understood it better than Byers could understand claims about alpha rays. We can imagine a range of interlocutors with understandings of assertions declining gradually from the doctor’s understanding of what they said, down to Burge’s confused man, all the way down to the minimal understanding that Eben Byers and the visitor to Ocracoke Island had of the utterances they accepted, and there will be no clear point at which we go from an interlocutor who understands to one who doesn’t.⁴⁶

Understanding is a matter of degree, but an expression’s being given a metalinguistic interpretation is not. So if we go with the metalinguistic view, we are left with two unappealing options. One, the radical metalinguistic view, is to say that the at-issue content is never what gets directly added to the common ground when an assertion is made and accepted, only the corresponding metalinguistic content is. To keep this from being obviously wrong, we need to add that other contents which the interlocutors (publicly) take to follow from the truth of the meta-linguistic content also get into the common ground as secondary effects. So this option can allow that interlocutors exchange information about non-linguistic matters. Be that as it may, the radical metalinguistic view is still unattractive. Typically when we declare that p, we’re directly asserting that p, not that the linguistic form we just produced is true, leaving it to the interlocutors to infer the consequences as they wish. Conversations seem to be primarily in material rather than formal mode. The radical metalinguistic appears to be a distortion of how conversations work.

The alternative modest metalinguistic view is to say that despite the smooth decrease, at some degree of understanding (perhaps not the same for all conversations), indicating acceptance switches from adding the at-issue content to common ground to adding the associated metalinguistic content instead. Those taking this option would be stuck with the unenviable task of saying why there would be this sudden flip and producing evidence indicating that it does happen.

I don’t mean to claim that the least common denominator view or either variant of the metalinguistic view is entirely hopeless. It might be worth trying to develop versions of them further and to try to give them some independent motivation. But as things stand, I take it to be a significant cost to have to accept any of them. Better to take stupefying, like other instances of assented-to assertions, to involve addition of at-issue content to the common ground.

In my view, then, stupefying results in additions to common ground of at-issue contents which have not been attended to by the addressee. So I conclude the At-Issue/Attention Link is false, which is where the account of §1 went wrong. We can fix it by dropping it. This allows us to see stupefying as a potential means of inattentional short-circuiting which works for at-issue content. One can get at-issue content into the common ground without that content’s being attended to, since you can get this content into the common ground without your interlocutor understanding it well enough to attend to it.

2.3 A Norm of Attention?

Though we should give up the At-Issue/Attention Link, it isn’t totally off-track. It does seem that in accepting through saying Okay or Oh, I see, etc., one typically does act as if one understands and is attending to the relevant content, as one acts as if one believes or knows the content of what one asserts. How can we account for this without the At-Issue/Attention Link?

One way is by appeal to a norm of correctness for acceptance, rather than some automatic updating of attention of the kind required for the At-Issue/Attention Link. Lying shows us that lack of belief doesn’t always prevent assertion, but we still think that there’s some conversational norm which requires belief in or knowledge of what one asserts.⁴⁷ This is perhaps why asserting is a way of representing oneself as believing. Similarly, stupefying shows us that lack of attention doesn’t always prevent acceptance, but we might still think that there’s some conversational norm which requires attention to the content one accepts. This would explain why it seems that accepting is a way of representing oneself as attending. I suggest that we replace talk of automatic public attention with a normative requirement.

Pursuing this line of thought, we could propose the following.

Attentional Norm of Confirmation: i may confirm ϕ only if $A_i^{[\varphi ]}\varphi$.

That is, confirm only if you are attending to the at-issue content of what you’re confirming. We need not take this to be the only norm governing confirmation. You shouldn’t confirm something you know to be false, for example. Nor do we need to assume that it is a fundamental norm; perhaps it can be derived from some other norm(s). However, it is the norm we’d need most directly for understanding why acceptance seems to represent one as attending.

Conversational participants, it seems, tend to try to follow this norm and often one is in some sense criticizable if one violates it. If it emerges that one had confirmed something one didn’t understand, one can be chided: “Why did you agree if you didn’t understand what I was saying?”. However, I don’t think that there can be such a norm in general. In some conversations, the following kind of hedged confirmations can be an acceptable response to assertions one doesn’t understand.⁴⁸

• (10) Hmm, I don’t think I understand that, but okay.
• I have no idea what that means, but yeah, sure.

If the Attentional Norm held in general, we’d expect hedged confirmations like these to be always unacceptable, as the corresponding ones seem to be for putative norms of assertion.⁴⁹

• (11) It’s raining but I don’t know/believe it is.

So while assertion may well have a norm of knowledge or belief, we should not accept that there is an attentional norm for confirmation.

It’s important, though, that one can’t say these things in just any circumstance. If you’re told that it’s best not to stay at the house with the whopperjawed pizer, uncomprehending acceptance seems worse than accepting that you shouldn’t stay at some given house which you can already identify, because it has a whopperjawed pizer. When something about the pizer is cited as a reason for some piece of advice, the uncomprehending addressee can still accept the advice, if only on faith. But when it is used to specify what the advice is, hedged confirmation doesn’t work because one can’t take the advice in the way the conversation seems to require.

This suggests that rather than appealing to a norm of acceptance, we should try to derive the normative requirement for attention in most cases from more general principles of pragmatics and rationality, in a roughly Gricean way. Something like:

Attentional Cooperative Principle: Apportion your attention as is required by the accepted purpose or direction of the talk exchange in which you are engaged.⁵⁰

Like Grice’s Cooperative Principle, this is very general and rather vague. Though it would take a lot of substantive work, the hope is that we could derive from it the requirement for most cases of attending to what one accepts, while still allowing for certain exceptions.⁵¹ This would explain why accepting usually seems to involve acting as if one is attending—and why the At-Issue/Attention Link has been an attractive assumption to many—but nevertheless allows for the possibility of stupefied acceptance, even an openly stupefied acceptance.

Much work would be required to defend and flesh out the details of a proposal along these lines. All I will say here is that it seems plausible that we can account for the connection between at-issue content and attention without committing to the At-Issue/Attention Link.

2.4 Leveraging the Stupor

I have argued that stupefying is a way that at-issue content can be added to the common ground without being attended to, and so the At-Issue/Attention Link should be rejected. However, we’re not quite finished with the account of how stupefying can work as a means of conversational manipulation.

In §1.1, we considered a puzzle about how getting something in the common ground can be a means of manipulation, given that belief is not in general required for something’s being in the common ground. The solution was to appeal to conversational tone: in conversations with certain serious conversational tones, belief is required. As things stand, though, this is an uncomfortable fit with the account of stupefying I’ve offered. There are two problems that need to be addressed.

The first problem concerns how we should think about conversational tone. The conversations involving stupefying that we’ve considered seem to be serious ones, not mere joking or pretense. Yet it’s plausible that in at least some cases of stupefying, some content added to the common ground is not believed by all the interlocutors. A fortiori, it is not publicly believed by all the interlocutors. So if we are to maintain the view that stupefying gets an assertion’s at-issue content into the common ground, we’ll need to revise the notion of conversational tone.

Another common phenomenon gives us independent reason to revise the notion of conversational tone in the required way. Lying is one way that a speaker can get contents into the common ground of a serious conversation. Yet one does not typically believe one’s own lies. A fortiori, the lies are not publicly believed by all the interlocutors. So there is independent reason to think a serious tone doesn’t require belief by all interlocutors.

How, then, should we think of conversational tone? Distinguish between what kind of attitude an interlocutor takes towards the propositions in the common ground and what they convey about their attitudes towards the propositions in the common ground. I think we should take conversational tone to be determined by what interlocutors convey about their attitudes rather than what attitudes they in fact have. Here’s a first pass at a revised definition: a conversation has tone T when each interlocutor i publicly intends that all other interlocutors believe that i has attitude T towards the propositions in the common ground. No doubt there will be further complications that a more developed theory of conversational tone will need to deal with, but this simple revision solves our first problem. The liar and the stupefied can intend others to believe that they believe everything in the common ground, so lying and stupefying can take place in conversations with a tone of belief.

The second and more pressing issue for the question of how stupefying works is this: how can a proposition’s being in the common ground significantly direct someone’s conduct if they don’t understand that proposition? If I have no idea what a whopperjawed pizer is, how can getting me to accept that the cottage has one significantly influence my actions beyond the conversation itself?

As noted above, understanding is a matter of degree, and in many cases of stupefying there is at least some degree of understanding. One route to influence outside of the conversation is through the stupefied’s partial grasp of the proposition they accepted and now think is true, whatever exactly it means. The addressee may understand the proposition well enough to realize that they should behave in certain ways if it is true, and so go on to behave in those ways.

This is not the only way that stupefying can influence conduct, however. The other way of achieving influence I will discuss does not rely on what the addressee does understand, but instead takes advantage precisely of their failure to understand.

Stupefied acceptance can be leveraged into acceptance of propositions that the addressee does understand. The most straightforward way to do this is for the speaker to assert, presuppose, or otherwise imply a conditional with the stupefying content as its antecedent: if Radithor produces alpha rays that do such-and-such, then ingesting radium water is beneficial to your health; if that cottage has a whopperjawed pizer, then you shouldn’t stay there. There are some cases where one might have allowed oneself to be stupefied but resist such a conditional. One might do so if one already has strong reason to doubt the consequent, for example. But this can be difficult to do, given that one doesn’t understand the antecedent. Moreover, typically the same kinds of reasons that have led one to accept the antecedent without understanding it will also lead one to accept the conditional and, on that basis, the consequent. So if one is not just partaking in a serious conversation but also taking the conversation seriously and believing what’s in its common ground to the extent that one can, one will come out believing this consequent, which one does understand well enough to act on in a variety of ways.

Stupefied acceptance can thus leveraged into acceptance of contents that are understood well enough to direct conduct. This sort of leveraging is what leads the stupefied to buying and drinking Radithor, staying at the cottage down the road, taking out this loan, voting for that policy, and so on.

3.1 Why Speakers Stupefy

Stupefying is often done intentionally. Given that it can be a means of short-circuiting, it’s no wonder that it is attractive to hucksters, quacks, frauds, and the like. It is a tool for manipulation.

One particular attraction of stupefying to the unscrupulous is that it sometimes allows for a kind of bootstrapping of epistemic authority. One method is what could be called the Guru’s Gambit: confidently say some gibberish, using impressive-sounding esoteric vocabulary. Imply this is connected to areas of deep concern to your audience—relationships, health, money, etc.—and that the gibberish gives you knowledge of feasible ways they can make big improvements in these areas. Since you’re saying impressive-sounding things which they think they can’t understand but seem to make sense to you, you must know something they don’t. You must be a relative expert on the matter of quantum healing or whatever. And as an expert, you may have some extra insight into what it can do. Moreover, since they don’t understand what you’re saying, no objections come to mind. One is inclined, then, to accept both the stupefying assertions and the more quotidian ones based on them, like “My method can help you” and “You should buy my book”. This is helped along by a healthy dose of wishful thinking and, when others have already bought into it, the stupefied acquiescence of the surrounding crowd.⁵³

A similar kind of bootstrapping can help explain obscurantism, which involves making statements which are difficult for one’s audience to understand but have contents which can be expressed in simple ways that could be easily understood. It is an attempt to stupefy by form while giving the addressee the impression that they are being stupefied by content. Like the Guru’s Gambit, obscurantism can trick an addressee into assigning the speaker more epistemic authority than they deserve. This is why charlatans often make needless use of jargon and put things in overly convoluted ways.

In cases like these, it rarely matters to the speaker what content the stupefying utterances have. All that matters is that it can successfully stupefy, then be leveraged in the right ways. This is why attempts to stupefy so often involve bullshit, in Frankfurt’s sense of lacking a “connection to a concern with truth” (Frankfurt 1986, p. 125).

Intentional stupefying is not limited to the domain of posers and purveyors of bullshit, though. For various reasons, real experts sometimes deploy their knowledge in ways intended to stupefy their audience. Here is an example from Elena Ferrante’s Those Who Leave and Those Who Stay. Lina is telling Lenù, the narrator, about her work with computers, which were newfangled technology at the time.

She talked to me about ferrite cores, rings traversed by an electrical cable whose tension determined the rotation, 0 or 1, and a ring was a bit, and the total of eight rings could represent a byte, and that is a character. … Is that clear? she asked me every so often. I answered yes, weakly, but I didn’t know what she was talking about. I perceived only that she noticed that nothing was clear to me, and I was ashamed of this.

(Ferrante 2014, p. 262)

Lina is “bewildering [Lenù] with an incomprehensible jargon” (p. 261), apparently intentionally. It’s not obvious why she’s doing so, but one can imagine a range of motivations for acting like this: to show off, to put down, to not insult by dumbing things down. Perhaps she’s just excited to show her friend that she’s finally found work that is enjoyable and intellectually engaging. She could well respond to Lenù’s shame by telling her it’s okay if she doesn’t understand and thanking her for listening.

Intentional stupefying can happen in various ways, for various reasons. But not all stupefying is intentional, and indeed it often goes unrecognized even after it has happened. Here’s a real example of unintentional stupefying of some concern to bioethicists. In his study of patient consent to medical research, John Fletcher describes the case of Mrs. B., a hospitalized patient who took part in a non-therapeutic study of dyslipoproteinemia. Fletcher reports:

… [Mrs. B.] complained in her interview with me that she lacked knowledge about the purpose of Dr. A.’s study, and that she had difficulty understanding the technical parts. When I reminded her that she did have ample opportunity in the consent process to ask such questions of Dr. A., she said “Well, I didn’t want him to think I was stupid!” She had pretended to understand some things and did not question Dr. A. as much as she wanted.

(Fletcher 1976, p. 270)

Mrs. B. was, it seems, stupefied by some of the things Dr. A. said in their discussions leading up to her consenting to the study.⁵⁴ But this was not intentional on Dr. A.’s part. According to Fletcher, “Dr. A. actually believed Mrs. B. to be very well informed and highly curious about the technical side of medicine” (p. 272).

Experts often have a hard time assessing what non-experts know, since they tend to overestimate the amount of shared background knowledge.⁵⁵ They are thus a common source of unintentional stupefying. They are not the only source, though. Take someone who has bought into the quack’s explanation, excitedly repeating it to their doctor friend hoping for it to be understood and confirmed. The friend may be stupefied, but not intentionally so.

3.2 Why Addressees are Stupefied

We’ve seen that there are a number of reasons speakers make potentially stupefying assertions, intentionally or otherwise. Why, though, do addressees allow themselves to be stupefied? Why go along with it?

In a personal conversation, when someone makes an assertion, here are the options the hearer has:

Besides confirming, denying, or raising some question about the utterance without confirming or denying it, one can of course do all sorts of things. Blankly stare into space, assert something totally unrelated, scream and run away, or stop, drop, and roll. These are all classed under the option other, and are usually not options on an addressee’s radar in a normal conversation, and so are typically ignored.⁵⁶ What we want to know is why, in a case in which something has been asserted which they don’t understand, an addressee would pick confirm as opposed to deny or question.

It’s pretty clear why one usually wouldn’t deny the utterance: since one doesn’t understand it, it’s unlikely one will be able to justify acceptance of the negation of what was asserted. And given the sort of challenge that denial typically involves to the original speaker, it’s likely that this justification will be demanded. The same explanation goes, more or less, for raising some question that’s supposed to bear on the assertion’s truth. One’s lack of understanding makes it hard to know what questions would make sense to raise, and the answers to them will likely just put one in a new stupor, so asking them doesn’t accomplish much anyways.

What’s more interesting is why, when faced with some assertion one doesn’t understand, one doesn’t simply ask for clarification about what the utterance means. Why not respond to the incomprehensible assertions with something like (12)?

• (12) I don’t understand, could you explain that in other terms?
• What’s an ionizing process/pizer/ferrite core/etc.?

Such requests for clarification are often the sensible thing to do, and indeed they are often made.⁵⁷ But sometimes addressees don’t ask them, instead allowing themselves to be stupefied. They need not be irrational on this account, however, because clarification requests can have costs that may outweigh the potential benefits. The most common and important clarification costs, I think, stem from three facts: (i) they take time and effort, (ii) they express ignorance, and (iii) they sometimes appear to challenge authority.

Regarding (i), it can take quite a while to explain to someone all they need to know in order to understand the relevant terms. In extreme cases, such as those involving arcane technical vocabulary, it can take years to acquire the concepts needed to understand. But even when it would take just a few minutes or less to figure out what’s being talked about, this can still be a cost: one could be doing more useful things and so, one might worry, could one’s interlocutor. Sometimes this cost is minor enough to be negligible, but often it is not.⁵⁸ If it doesn’t seem important enough that one fully understand before accepting, allowing oneself to be stupefied may seem to be the best option so as to avoid the demands on time and effort.

As for (ii), we’ve already seen this cited as an explanation of stupefying: Mrs. B., recall, worried that Dr. A. would have thought her to be stupid had she asked what was necessary to avoid being stupefied. This is a common reason that addressees allow themselves to be stupefied: by pretending to understand, one saves intellectual face.⁵⁹

Moreover, in plenty of cases, ignorance can imply other failures which we wish to keep hidden. It is often one’s responsibility to understand certain things because one is a citizen, say, or a parent, or a doctor, or because one has promised to read and think about something. Displaying ignorance by asking for clarification, then, can also involve the cost of revealing that one has failed to fulfill certain obligations.

Finally, our identities are often tied up with what we know, partic ularly with our knowledge of language.⁶⁰ One may worry that asking someone to tell them what pizer means may out them as an outsider (or a dingbatter, as it would be put in Ocracoke Brogue), or that asking what supervenience means would show one not to be a bona fide analytic philosopher. Allowing oneself to be stupefied is sometimes a way one tries to fit in.

Another potential social cost of clarification questions is (iii): that they can challenge authority, or at least appear to do so. We’re all familiar with the aggressive “But what does that even mean?” clarification request meant less to find out what the speaker meant than to show everyone that they were talking nonsense. Clarification questions, even ones not meant in this way, can come across as challenges to the speaker, which can be costly. One way it can be costly is when the speaker is, in one way or another, more powerful than the one asking the clarification question. If their authority appears to be challenged, they or their associates might impose some social or even physical cost on the questioner.

Another way the appearance of challenge can be costly is through the harmful effects it has on the person who made the original assertion. We can’t always explain what we mean, even when we have said something perfectly meaningful. But someone’s not being able to explain what they meant can nevertheless make them appear incompetent, reducing their credibility. Similarly when one can but does not want to take the time to explain what one means. Clarification questions’ ability to undermine credibility means they can exacerbate what Miranda Fricker (2007, p. 1) calls testimonial injustice, which occurs “when prejudice causes a hearer to give a deflated level of credibility to a speaker’s word”.⁶¹ Triggering or worsening testimonial injustice is sometimes a cost of clarification questions. Stupefied acceptance is a way to avoid such costs.

How costly (i), (ii), and (iii) end up being varies a lot depending on the context: how pressed one is for time, what one is expected to know, the social positions of speaker and addressee, and so on. But it’s not hard to see how they can make requesting clarification into an unattractive option when compared with allowing oneself to be stupefied.

This is especially so when the costs of confirming are low, as they often are. One obvious way for the cost of confirming to be low is if not much rides on what ends up becoming joint discourse commitments. This is so when the conversational tone is not one of belief, or when one doesn’t care about the topic or respect one’s interlocutors enough to take the result of the conversation seriously. If people are just chatting to pass the time and achieve various social goals unrelated to a conversation’s content, there’s not much harm in explicitly accepting some assertion which one doesn’t understand. Hence the ubiquity of smiling and nodding at cocktail parties.

Another way for (expected) confirmation cost to be low is when what matters is not anything that depends on the hearer understanding, but just depends on whether what has been said is true. Sometimes the hearer has good evidence—and indeed can know—that what the speaker uttered is true, even without understanding.⁶² If my doctor tells me that levothyroxine will “diffuse into the cell nucleus and bind to thyroid receptor proteins attached to DNA”, activating “gene transcription and synthesis of messenger RNA and cytoplasmic proteins”, and that this is why she thinks it will help with some symptoms I have, it really makes no difference whether I understood what she said, so long as she’s right about it and about its connection to my symptoms.⁶³ And given her profession, I trust that she is right about this, so it seems like a low cost to just accept what she has said without understanding.⁶⁴ For similar reasons, Eben Byers may well have been rational in being stupefied into drinking Radithor.

3.3 What Is to Be Done?

We’ve seen some reasons why speakers might say what their addressees won’t understand and why addressees sometimes accept what has been said anyways. Sometimes speakers stupefy out of greed, malice, or carelessness, but not always. Sometimes addressees are stupefied out of vanity or obsequiousness, but not always. Stupefying can be a means of short-circuiting and manipulation for speakers and can be irrational for hearers, but it can also result from innocent, cooperative, and reasonable conversational moves. Not all stupefying is bad. We should probably not try to eliminate stupefying, then, but rather be wary of it. I’ll conclude by speculatively suggesting that some important goods are unattainable without stupefying. This would make any attempt to avoid stupefying altogether a serious mistake.

When we learn from others speaking with us, we usually do so by at least provisionally accepting what they say.⁶⁵ Some learning involves eliminating possibilities which we already realize, if only implicitly, are possibilities. This is well modeled by the familiar Stalnakerian picture of conversation. Other learning, though, involves conceptual expansion, adding to our stock of what we consider possibilities.⁶⁶ We emerge from it not, primarily, with a collection of new facts about how the world is, but rather new ways of thinking about how the world might be. These new ways of thinking are usually not easily definable from what one already understands. So if a teacher aims to impart this kind of knowledge through speech, they will typically do so by saying what we don’t yet understand. And to learn the new way of thinking from them, we provisionally accept what they say without understanding it, hoping that as we follow them and try thinking these thoughts on our own, things will begin to fall into place. When this happens, we gradually come to understand what we have already accepted. Learning from others in this way will involve stupefying.⁶⁷

In one of Malebranche’s dialogues, Theodore’s pupil Aristes finally accepts something Theodore has been saying and restates the view back to him. “Do you conceive all this quite distinctly?” Theodore asks, “Are you quite convinced of it by the reasons I have given you and by your own reflections?” Aristes admits he was stupefied:

ARISTES…. It seems that you excite sensations in me instead of producing clear ideas. I am using your language. In all honesty, I do not understand everything you are telling me. I see it, and a moment later I no longer see it. For I still only glimpse it. It seems to me you are right, but I am not understanding you too well.

(Malebranche 1688–1997, §III.viii)

As we’re learning new ways to think, we often only barely glimpse what someone is saying. To come to understand it, we accept it anyways, using their language until it becomes our own. If this is right, we need to live with stupefying. There are times when we should resist the temptation to stupefy and be stupefied, but other times it is necessary for imparting and acquiring new ways of thinking.⁶⁸