Testimony is an important source of knowledge: we form beliefs by taking in what others tell us.Footnote 1 If these beliefs are true (and certain other conditions are met), then those beliefs constitute knowledge.Footnote 2 However, not every belief we acquire from others talking to us is a belief acquired through testimony. From a speaker’s utterance, a hearer can often acquire knowledge about (say) a speaker’s accent or personality. This is not knowledge acquired through testimony.

What sort of utterances are appropriate material for testimonial belief acquisition? Or, in the language of speech acts: which speech act can appropriately perform the epistemic magic of testimony? The default answer is: assertion. Can other utterances also function as testimony? In particular I ask: can, for example, questions, complaints, or commands contain testimony about what the speaker took for granted in making these speech acts? I shall answer these questions in the affirmative. One testifies not only to what one asserts, but also to what one semantically presupposes, as I argue. The very same features which make assertions paradigmatic testimony are also features of semantic presupposition. By parity of reasoning then, if assertions are testimony so are semantic presuppositions.

Most work in epistemology takes for granted that explicit assertion is the paradigmatic vehicle of testimony. It is sometimes even assumed to be the only such vehicle. Below, I challenge this commonplace assumption and expand the category of testimony. For now, the following are paradigmatic examples of testimony:

  1. (1)

    I have a sister,

  2. (2)

    Your name is Phyllis.

  3. (3)

    Elizabeth II. is the queen of the UK.

Paradigmatic testimony is said to come in the form of explicit “statements” (Coady, 1994, 42; Graham, 1997, 227), “tellings generally” (Fricker, 1995, 396) or “assertoric tellings” (Fricker, 2012, 62), “people telling us things” (Audi 1997, 406; Sosa, 1991, 219), “exchange of information through telling and being told in everyday life” (Moran, 2005, 8), “someone tell[ing] us that p” (Owens, 2006, 105).Footnote 3 For some, assertion is not only the paradigmatic vehicle for testimony, but the exclusive vehicle. Elizabeth Fricker explicitly excludes presupposition, as we will see (2012), and Sanford Goldberg notes that “in order to constitute a case of testifying (or a case of one person “telling” another something), a speech act must have the assertoric force of straight assertion” (2015, 76).Footnote 4 So the idea of testimony as statement, assertion, or telling seems to underpin a large variety of accounts of testimony (I quoted from the works of both evidentialists and non-evidentialists; from both reductionists and non-reductionists).Footnote 5

Statements and assertions are utterances in the declarative mood. While ‘assertion’ seems the more technical term in philosophy of language, I take it that ‘statement’ refers to roughly the same sort of utterance. A telling is a specific kind of assertion: “Telling (…) is making a statement with the understanding that here it is your word that is to be relied on” (Moran, 2005, 8). Or a telling is a speaker’s assertion where the speaker seeks to share p with an “intended audience, whom she believes to be ignorant, or possibly ignorant, as to whether P” (Fricker, 2006, 596). Tellings are therefore more directly aimed at inducing belief in a hearer. Putting these specifics of tellings aside, we can say that assertion is usually regarded as the paradigmatic testimonial utterance. This is a legitimate move because, as we will see, what makes tellings testimony on these accounts depends on their assertoric character.

But assertion isn’t the only thing we can do with words nor is it the only thing that functions as testimony—or so I will argue. Many of our speech acts take things for granted. (1)–(3) above have a strange, unnatural feel to them while actual conversation is more likely to contain sentences like the following:

(1*) (to a friend) I’m sorry, I’m running late [for our coffee date]; I’ve had to pick up my sister from the airport today.Footnote 6

(2*) (Mom to Dad, while daughter Phyllis listens) Little Phyllis can crawl so well already; surely, she will walk soon.

(3*) (a news reporter) Queen Elizabeth II. of the UK will visit Ghana today.

What was asserted in (1)–(3) is semantically presupposed in (1*)–(3*): that I have a sister, that the daughter’s name is Phyllis, that Elizabeth II. is Queen of the UK. All six examples are assertions, but (1*)–(3*) are more complicated assertions: they contain presuppositions within them. Especially (2*) seems more natural than (2): it seems unlikely behaviour to come up to your small child and explicitly tell her what her name is. Instead, Phyllis will likely pick up on her name when it is used as it is in (2*), rather than declared. But perhaps (1*)–(3*) still strike us as slightly artificial. We can model them even closer on natural language:

(1**) (to a friend) I’m sorry, I’m running late; haven’t I told you I had to pick up my sister at the airport?

(2**) (Mom to Phyllis) Oh, well done, my darling Phyllis, you are doing so well; now try to get up on your two feet, Phyllis! Can you do that for me?

(3**) (a critic of imperialism) Ah, the UK’s Queen Elizabeth II. flew over again from her nice London palace to inspect the slums of Ghana—is this the 1950s?

(1**) contains a question and perhaps a complaint, but the question still conveys the content of (1), among other things. (2**) is an expression of encouragement and seems to also contain a command of some sort but still conveys that the hearer’s name is Phyllis as (2) did. (3**) has turned into a rhetorical question in a sarcastic tone. It can be uttered in a context where everyone knows full well all those propositions conveyed within it—its purpose is not to convey content but to express annoyance. Yet, (3**) could still convey to an ignorant hearer that the UK’s queen is Elizabeth II., the content of (3). The last six examples are cases in which the propositions of (1)–(3) respectively are conveyed via semantic presupposition. The last three are further cases where these propositions aren’t embedded within assertion, but within different speech acts. On the dominant picture underlying large parts of the debate, only (1)–(3) would qualify as testimony of the propositions expressed in those. (1*)–(3*) would testify to the content of the main assertion about the airport, crawling and walking, and the visit to Ghana. (1**)–(3**) would not be testimony, or paradigmatic testimony, on the dominant picture. But it seems that from all nine example utterances an audience could come to believe the propositions of (1)–(3) respectively. Might not all nine cases then be testimony? I argue that this is so. (1*)–(3**) are instances of testimony for the very reason that (1)–(3) are.

To make this claim, we first need a better grasp of why assertion is usually taken to be paradigmatic testimony. In Sect. 2, we will see that this isn’t just to make thinking about testimony easier: some have argued that by asserting (and even only by asserting) a speaker presents herself as knowing p and therefore licenses the hearer to form a belief that p from the speaker’s say-so. So assertions become a necessary component to the magic of testimony. Elizabeth Fricker (2012), Richard Moran (2005), and Sanford Goldberg (2015) explicitly defend this view about the exclusivity of assertion while the idea that assertion is paradigmatic testimony is in the background of a wide variety of work on testimony (see list above). It is because of their explication of what is elsewhere merely assumed that I focus specifically on Fricker’s, Moran’s, and Goldberg’s arguments. Fricker’s argument about a knowledge norm for assertion (and Goldberg’s about an epistemic norm) will further be my core target as they offer the most detailed defence of the assertion-as-testimony account. It would seem to follow from their arguments that presupposition fails to work the testimonial magic. I take issue with this claim. In Sect. 3, I argue that semantic presupposition is exactly like assertion in its testimonial powers. I show that, whatever reasons we have for thinking that assertion can be testimony, those are equally reasons for thinking that presupposition can be testimony. This means that—by the very lights of dominant assertion-based accounts of testimony—presupposition should also be testimony. While I develop my argument specifically in opposition to Fricker’s, Moran’s, and Goldberg’s arguments for assertion-testimony (and Fricker’s against presupposition-testimony), my point applies more widely: as we saw, assertion is currently regarded as the paradigmatic, if not actually the only way, to testify. I refute both these claims. Section 4 will discuss and reject one objection to my argument about a common knowledge norm.

I suggested that our last six cases (1*)–(3**) more closely represent how people actually speak. Yet, before we dive into the argument about assertions, presuppositions, and testimony, we may want to ask: why is it that people speak in these presupposition-heavy ways?

1 Why Presuppose?

Fricker gives two reasons for opting to be implicit: economic considerations of brevity and epistemic considerations of avoiding commitment (2012, 68). We will see that the second reason is false: presupposition actually doesn’t carry with it less commitment than assertion. This enables both assertion and presupposition to function as testimony. But presupposition has advantages in addition to Fricker’s first point that it conveys information quickly and briefly. Presupposition can elegantly change the social landscape, it is good at inducing belief, and it can help the speaker avoid challenge or complaint. All these may be good reasons for speakers to presuppose rather than assert.

First, a speaker can easily change the social landscape with presupposition. Depending on the context, “I had to pick up my sister from the airport” can function to either include or exclude those who didn’t yet know that I have a sister. In some contexts, with some intonations, it can welcome a previously ignorant speaker into a group of friends who (now) all know about my family. In other contexts, with another intonation, it can exclude the previously ignorant speaker, making clear to them that everyone but them is already good enough friends with me to know about my family. Surely, other utterances, like the explicit “I’m not friends with you!,” also have exclusionary force; yet, presupposition seems able to possess this force particularly elegantly and implicitly without direct confrontation.

Secondly, presupposition might be particularly good at inducing beliefs in one’s audience. This is so for at least the following reason: I must already assume or entertain the content of an utterance’s presupposition so that I can even be in a position to understand the utterance’s main content. Here is an example: without assuming that Phyllis is his daughter’s name, the father in sentence (2*) can make little sense of his partner’s utterance. He might wonder: “Who is this Phyllis? Why should I care about her crawling and walking abilities?” Elisabeth Camp explores this idea of a presupposition’s priority for the workings of metaphors (2017, 51, 53–57): metaphors presuppose particular perspectives or characterisations that provide a frame necessary for understanding the metaphor. Here is one of Camp’s examples:

But, soft! what light through yonder window breaks?

It is the east, and Juliet is the sun. (Romeo and Juliet, 2.2.2–3)

Only if I accept the presupposed characterisation of the sun as bright and warm (rather than hostile and burning hot) can I make sense of Romeo’s metaphorical assertion that Juliet is the sun. If I presuppose another, more hostile, characterisation of the sun, I will be confused as to why Romeo delivers these lines in a loving tone. Entertaining the content of the presupposition is therefore a necessary prerequisite to adequately considering the main point of a speaker’s utterance. True, entertaining is different from accepting: during a theatre performance I might entertain that the sun is warm and bright only to later return to my belief as an astrophysicist that the sun is actually hot and hostile. Nevertheless, getting an audience to entertain something is already a significant achievement, as Mary Kate McGowan for example has shown for the accommodation of salience and presupposition (2004). And entertaining is a first step towards accepting and believing. A speaker who seeks to get her audience to believe controversial content would therefore be well advised to pack it into a presupposition, rather than to outright assert it. In this way, she is able to sneak this controversial content into the speaker’s belief-set through a “back door,” as Rae Langton has put it (ms ). Marina Sbisà (1999) drew attention to this point some 20 years ago: in an analysis of presupposition use in Italian newspapers she outlines how presupposition is frequently employed to spread controversial, ideological content.

Sbisà (1999) also notes another interesting feature of presupposition: a presupposition is more difficult to challenge than an explicit assertion. There are at least two reasons for this. First, challenge can be linguistically difficult. As Jason Stanley remarks in his work on propaganda and not-at-issue content, this is because “denying the speaker’s claim is naturally understood as denying what is asserted, while agreeing with what is presupposed” (2015, 135–6). To deny a speaker’s presupposed content, a hearer needs to be more inventive. Here is a classic example, as raised by David Lewis (1979, 339) and since discussed by Rae Langton and Caroline West (1999, 309; Langton, 2018, 153):

Even George Lakoff could win.

— Whaddya mean, even?

A simple “You’re wrong” would not have done the trick. That would have denied that George won, it would not have denied that he is an unpromising candidate. (More on this central feature of presupposition in Sect. 3.)

But there is a second sense in which challenging a presupposition is difficult. As Rae Langton put it, this is a normative, rather than linguistic sense (2018, 160–161). Sbisà writes: “the speaker is required to take [the presupposition] for granted, unless he or she wants to challenge the author's reliability and thus the communicative relationship” (1999, 14). A hearer might prefer not to violate such norms of conversational politeness unless absolutely necessary. This difficulty to challenge a presupposition relates to the previous point about easily induced belief: when one hearer avoids being uncooperative and rude in challenging a speaker’s presupposition, it is more likely that other members of the audience accept the presupposed proposition. Nobody has alerted them to its potential falsity.

Taken together, presuppositions have many advantages: they can elegantly include or exclude parts of one’s audience, they seem particularly good at inducing belief, they are difficult to challenge, and they can do all this often with less words than assertions to the same end. These advantages may explain the frequent use of presupposition. This is speculation: empirical work could better establish why speakers actually presuppose. What we can see from our speculation, however, is that presuppositions are especially advantageous for a speaker who can say all sorts of things, make others believe them, and need not be too afraid to be held accountable. Fortunately, however, the balance of power isn’t entirely tipped towards the speaker’s side. If we understand that presuppositions can function as testimony, as I show now, the hearer has to hold the speaker accountable for presupposed falsehoods (even if acting on this right may be difficult).

2 Testimony and Assertion

What is so special about assertion that it (or even only it) has been taken to function as testimony?Footnote 7 To answer this question let us look at a specific case of belief-induction from testimony.

Speaker S asserts (1) to hearer H: “I have a sister.” H, who subsequently exclaims “Ah!”, comes to believe that S has a sister. S’s testimony served as a source of belief for H when H assumes S has spoken truly, infers that the content of S’s utterance is true, and comes to believe S’s proposition on the basis of this evidence. This is one—evidentialist—story of testimonial belief induction. H uses S’s utterance as evidence for H’s belief. Here is another—non-evidentialist—story: when S utters (1), S assures to H the truth of (1) and H, much like accepting a promise, can accept this assurance and so believe S’s proposition. Here, H doesn’t use S’s utterance as evidence but simply takes S’s word for it. In the good case, testimony does not only induce belief, but knowledge.Footnote 8

But why is H licensed to believe the proposition in (1) just on the basis of S’s assertion (1)? S could easily have lied and H would have formed a false belief. For both evidentialist and non-evidentialist the fact that S uttered an assertion (at least in part) licenses H to form a belief according to S’s say-so. For some non-evidentialists other conditions must also hold: H must for example have additional evidence as to S’s reliability. Putting these additional conditions aside, let  us focus on assertion and its power to license belief-formation.

S’s assertion gives H a special guarantee that S spoke truly. Fricker (an evidentialist) says that by asserting a speaker “overtly and fully commits herself to vouching for P as true, and presents herself as knowing that P” (2012, 63). Similarly, Moran (a non-evidentialist) claims that by asserting a speaker “presents himself as accountable for the truth of what he says” and “offers a kind of guarantee for the truth” (2005, 11). This is not a guarantee that p will definitely be true. Rather, it is a guarantee that should p turn out to be false the speaker is liable for that. Equipped with this guarantee, a hearer can come to believe that S has a sister on the basis of S’s testimony. Goldberg (a different kind of non-evidentialist) agrees: by asserting a speaker is “attesting to the truth of a proposition,” she is “presenting a proposition as true in such a way as to implicate one’s own epistemic authority on the matter” (2015, 77).

Committing oneself, presenting oneself, attesting, or offering a guarantee are intentional acts: I cannot present myself as knowing without intending to present myself as such. At least Fricker and Moran repeatedly stress this point explicitly (Fricker, 2012, 64, 70; Moran, 2005, 6, 15, 19).

But the relevant sense of intentional act must be a weak sense: we make assertions all the time and not in all cases do we explicitly plan to portray ourselves as knowing. Neither is presenting ourselves as knowing always our “primary reason” for asserting, following Donald Davidson on intention (1963). Rather, we assert p with the primary reason of passing on information, indicating the right direction to a stranger on the street, etc. Yet—and perhaps this gets closest to the invoked sense of intention—for this to succeed it must seem to the stranger that I am speaking from knowledge: only then will she follow my directions. So if I speak with the primary reason of giving right directions, that entails that I want to present and so present myself as speaking from knowledge. In this way, Fricker’s and Moran’s claim about assertion as intentionally presenting oneself as knowing can be sustained.

Why does assertion come with this special guarantee, with this overt and full commitment? Moran seems to simply postulate that it does. Asserting that p just has the normative power of vouching for the truth of p, much like promising that p just has the normative power of vouching that one will do p. Fricker and Goldberg, on the other hand, say more about this guarantee that comes with assertion: it has to do with the fact that assertion is constituted by an epistemic norm. I look in particular at Fricker’s knowledge norm (2012), but the same argument would hold for Goldberg’s more general epistemic norm (2015). This explanation about norms will become important for my argument about presupposition as testimony. I will show that presupposition is constituted by a similar norm and therefore comes with exactly the same guarantee as assertion.

Fricker starts with the assumption that speech acts are “constituted, and its occurrences made possible” by implicit norms (2006, 593). The norm constituting assertion is a knowledge-norm, as Fricker notes, following Timothy Williamson: “one must: assert p only if one knows p.” (Williamson, 2000, 243).Footnote 9 So if S had asserted “I have a sister” without knowing it, she would have violated the norm. A constitutive norm states, first, what is essential to the speech act it constitutes: a speech act without this norm would be a different speech act (Williamson, 2000, 239). Secondly, a constitutive norm determines what individuates the speech act it constitutes: “necessarily, assertion is the unique speech act A whose unique rule is the C rule” (Williamson, 2000, 241). So if a knowledge norm constitutes assertion, then only assertion is governed by such a norm—constitutive norms individuate. Thus, if knowledge being the norm of assertion explains why assertions guarantee the truth of the asserted proposition, then only assertions come with such a guarantee. But if the presence of this guarantee is necessary for testimony, then only assertion can function as testimony. This seems part of Fricker’s reasoning for her claim that only assertion is testimony.

One further consequence of the knowledge norm for assertion—aside from a guarantee the speaker gives—is this: if the speaker has not spoken correctly, a hearer has grounds to “complain to S if p turns out to be false” (Fricker, 2006, 594; cf. Goldberg, 2015, 80). This feature can be explained by the knowledge norm for assertion: if a hearer can complain about a speaker’s knowledge norm violation of asserting what she didn’t know, a hearer can also make a weaker complaint when the speaker asserts what is false. A knowledge norm (or an epistemic norm generally, following Goldberg) entails a truth norm. H can therefore complain to S when (1) is false: “What do you mean? That can’t be true! I know for a fact that you are an only-child!?”Footnote 10 Such a complaint will count as a criticism of the speaker’s assertion—in asserting a false proposition, she violated the knowledge norm. The speaker must take the blame for her wrong assertion.

Complaining might sound harmless, but it need not be. Aside from complaining to the speaker, a hearer has multiple other, more severe options: she can stop being friends with the speaker; she can tell others about the speaker’s unreliability and socially exclude her; if they work together, the hearer can complain to their boss who could have the speaker fired; in particularly serious cases a court might even persecute her. In light of this, speakers have at least some interest in asserting the truth some of the time and hearers have relative security that they will.

On this picture, only assertion is testimony for three connected reasons: only assertion is governed by a knowledge norm, only in the case of assertion does a hearer have grounds to complain about falsity, and only through assertion does a speaker offer a guarantee for the truth of her proposition to the hearer. I argue that all three reasons are also reasons for thinking that presupposition can be testimony. They are falsely presented as exclusive features of assertion.

Fricker clarifies her claim about the exclusivity of assertion-testimony: assertions are labelled “explicit primary messaging” and contrasted with “implicit secondary messages”; this second class cannot function as testimony.Footnote 11 To simplify, we can say that these two are on different sides of an “explicit-implicit boundary” (2012, 90). Here is an example of an implicit secondary message:

(1***) (in a conversation about whether being an only-child or having siblings is better) It sucks to have a sister.

This implicitly conveys that the speaker has a sister and has made bad experiences having a sister. For Fricker, implicit secondary messages of this sort are not testimony because they are (1) deniable (2012, 87–8) and (2) vague (2012, 88). If a message is deniable, a speaker offers no guarantee for it nor can a hearer complain if what she perceived to be the message is false. In (1***) a speaker can deny that she communicated to have a sister, rather she might say that she only had in mind the terrible experience her friend had with his sister. If a secondary message is vague, no knowledge norm attaches to it. Perhaps the speaker of (1***) herself hasn’t decided whether she wanted to convey that she has a sister or that her close friend has one.Footnote 12

Fricker lists presupposition as an example of implicit secondary messages (2012, 90). While I agree with Fricker that deniability and vagueness are genuine problems and that (1***) is therefore probably not testimony, I believe that these two problems don’t apply in the case of presupposition. But instead of arguing that presuppositions are neither deniable nor vague,Footnote 13 I make a stronger claim: I show that Fricker’s exact reasons for the testimonial function of assertion are also reasons for the testimonial function of semantic presupposition. Presupposition, too, is constituted by a knowledge norm, a hearer has grounds to complain if the conveyed proposition is false, and a speaker offers a guarantee for the truth of the presupposed proposition.

3 Testimony and Presupposition

The sentences in (1*)–(3*) both assert and presuppose things. (1*), for example, asserts that the speaker had to pick up her sister from the airport today, when it presupposes that she has a sister. This presupposition is marked by a trigger: “my sister.” It is therefore a semantic presupposition, while the whole sentence pragmatically presupposes that the speaker has mastered the English language (at least to some extent). Such—pragmatic—presuppositions have “nothing to do with the meanings of any of those words” in sentence (1*) (Beaver & Geurts, 2014). Semantic presuppositions, on the other hand, have to do with word meaning: they are marked by triggers which contain presuppositions “as part of the[ir] conventional meaning” (Beaver & Geurts, 2014).Footnote 14 The assertoric content of a sentence then builds on these semantic presuppositions.Footnote 15

A number of tests have been put forward for identifying presuppositions (although all face problems and exceptions).Footnote 16 I briefly present only von Fintel’s “hey, wait a minute test” (2004, 271) to give the reader a better grasp of the idea of semantic presupposition. It goes like this: I can challenge the presupposition in (1*) with “hey, wait a minute, I didn’t know you have a sister.” But “hey, wait a minute, I didn’t know that you’re picking her up at the airport today” isn’t as smooth. The retort “Well, yeah, that’s why I’m telling you now” seems apt. So the test is able to distinguish presupposition from assertion. Assertions can be challenged by a hearer’s simple “No!” while challenging a presupposition requires more round-about moves like “hey, wait a minute…” As we saw in Sect. 1, this is an advantage for a speaker who wants to avoid being challenged.

The problem with von Fintel’s test is that it doesn’t only single out semantic presupposition. It seems to also pick out other implicit messaging, like pragmatic presupposition (“hey, wait a minute, I didn’t know you spoke English”) or conventional implicatures (“I’m picking up my sister who’s flying in from Florida.”—“Hey, wait a minute, I didn’t know your sister was flying in from Florida”).

But let’s not spend too long on the linguistic intricacies of presupposition—my focus is epistemological. The reader may permit me one final note about the kind of linguistic thing a presupposition is. Assertion is a speech act and testimony is an epistemic source. Presupposition, I suggest, is something like assertion, but not quite its own speech act. This is because presupposition doesn’t stand on its own feet: it is a part of a larger utterance. I will therefore borrow Manuel García-Carpintero’s (2020) classification for whom (drawing on Searle (1969)) presupposition is an ancillary speech act, auxiliary for the performance of other speech acts such as assertion (as in (1*)) or question (as in (1**)). This classification has the benefit that presupposition, too, can have a constitutive norm, much like assertion. This insight about a constitutive norm will get going my argument about presupposition as testimony.

Following Fricker and Goldberg (and to some extent Moran), an assertion is testimony for three reasons, as we saw above: a constitutive epistemic norm, a hearer’s right to complain if the asserted p is false, and the speaker’s offered guarantee for the truth of the asserted content. I agree that these three reasons enable assertion to be testimony, but I disagree with the claim that assertion is therefore the central or even exclusive vehicle of testimony. Below, I argue that for presupposition exactly the same three reasons also hold. A road-map will help: leaning on García-Carpintero’s (2020) recent work in philosophy of language in response to Williamson (2000), I first argue that presupposition, too, is constituted by a knowledge norm. We saw that the two other reasons for assertion as testimony—complaint and guarantee—are related to this first reason about a knowledge norm. These two reasons therefore also follow for presupposition, as I show next. We might worry that the same norm now seems to individuate both assertion and presupposition, so in a further step I calm these worries. I conclude: presupposition can function as testimony, just like assertion.Footnote 17

3.1 Knowledge Norm

Williamson has three arguments that assertion is constituted by a knowledge norm, all of which also work for presupposition, as I show following García-Carpitnero’s recent work. First, Williamson invokes ordinary ways of speaking. To express doubt in an utterance of (3) we ask “How do you know?” or more aggressively “You don’t know that!” (2000, 252f.). This indicates the existence of a knowledge norm: only then can we make these challenges. But the same works for presupposition, as García-Carpintero (2020, 280) shows. Let’s take (3*) “Queen Elizabeth II. of the United Kingdom will visit Ghana today.” The presupposition can be challenged by: “How do you know it’s Queen Elizabeth II.? I thought the sovereign’s name is Charles III.”

Secondly, if we assume that assertion is constituted by some norm, we will likely arrive at a knowledge norm. Let’s start with a weak norm: a truth norm. But a truth norm for assertion amounts to a warrant norm: to make sure that I adhere to the truth norm I need to have a warrant for the truth of p; but a warrant for the truth of p is just a warrant for p. Yet, the warrant norm is not yet strong enough. I am not allowed to assert “Your lottery ticket didn’t win” before I know this to be the case, even though I do have a warrant for that assertion (winning the lottery is highly unlikely), Williamson observes (2000, 251). The same holds for presupposition, as García-Carpintero points out (2020, 284). In “Let’s leave, there’s no point in waiting for them to announce that your ticket didn’t win” I presuppose that the ticket didn’t win, but request that we leave. I am presupposing wrongly because I don’t yet know whether the ticket won, even though it’s highly unlikely it did. As in the previous case, my friend can complain: “You don’t even know that yet!” So a knowledge norm must also apply to presupposition.

Thirdly, a knowledge norm for assertion explains what is wrong with the Moorean paradox: “A, and I don’t know that A” (Williamson, 2000, 253–4). The same inadequacy can be detected for presupposition, following García-Carpintero (2020, 284) and Langton (ms 2017): “I’ve had to pick up my sister from the airport today, and I don’t know that I have a sister” is odd. This must mean that presupposition is constituted by a knowledge norm, analogous to assertion.Footnote 18

One might object to this argument as follows. In all the examples I gave so far, the relevant presuppositions were embedded within assertions. One might suspect that we found that presuppositions are constituted by a knowledge norm only because they all occurred within assertions and because assertions are constituted by such a norm. So there is no knowledge norm connected to presupposition itself.Footnote 19 I believe that this worry is mistaken: it is not the case that one must know what one presupposes only if the presupposition is embedded within an assertion. Presupposition is constituted by a knowledge norm independently of assertion’s knowledge norm. We can see this when we look at (1**)–(3**): these are three examples of presupposition outside of assertion. And yet, using the three tests outlined above—ordinary conversation, lottery cases, and Moorean paradox, we find that a speaker ought to also know those presuppositions embedded in questions, complaints, or commands. Presupposition is therefore constituted by a knowledge norm, just like assertion is.

3.2 Guarantee

The knowledge norm for assertion makes it so that assertions come with a guarantee for the truth of their utterance. If this is the case, then a knowledge norm for presupposition must yield the same result: presupposition too must come with a guarantee for its truth. Fricker, Goldberg and Moran motivate this claim by showing what one does when one asserts: one offers a guarantee for the truth of one’s asserted proposition, commits oneself to its truth, and so presents oneself as knowing it. This language of commitment, presentation, and guarantee describes just as well what one does when presupposing—or so I show now.

A speaker of (1*) or (1**) offers a guarantee that she has a sister: this guarantee comes with her words “my sister” and its force as a presupposition. She commits herself to this proposition: if she later asserts that she doesn’t have a sister, she can be criticised for her contradictory claims and for her flakey commitment. So the speaker presents herself as knowing that she has a sister. We can see this because Williamson’s test works for her: “You don’t know that you have a sister” critiques the speaker’s presentation as a knower.

But does the speaker really present herself as a knower of the proposition that she has a sister? In (1*) does she not rather present herself as needing to pick up her sister from the airport? In (1**) does she not rather present herself as wondering whether the hearer already knows about her sister’s arrival and as annoyed by her sister? This is analogous to a worry I had about intention in the case of assertion: do we really present ourselves as speaking from knowledge when we assert? Might we not rather present ourselves as doing whatever the primary point of our utterance was? If we face the same worry again here, we ought to respond analogously. Above we said that, in the case of assertion, in presenting ourselves as doing whatever the primary point of our utterance was, we also present ourselves as knowing what we assert. So the same must hold for presupposition. Let’s take a concrete case: in presenting oneself as needing to pick up one’s sister from the airport, a speaker also presents herself as knowing that she has a sister. Otherwise she would not be able to present herself as needing to pick her up. This means that by presupposing a speaker presents herself in the same way and so offers the very same kind of guarantee for the truth which Fricker, Goldberg and Moran find in assertion.

Assertion’s guarantee for the truth enables assertion to be testimony: a hearer is now licensed to form belief p from the speaker’s say-so. In parallel fashion this means: presupposition’s guarantee for the truth enables presupposition to be testimony. The hearer is licensed to come to believe the presupposed p just the same.

3.3 Complaint

Let’s test this result against our last reason for an utterance’s testimonial function: a hearer’s ground to complain if the asserted proposition is false. This ground to complain also exists when the presupposed proposition is false. Take again the example of the sister which we looked at above. There, H could complain if S falsely asserted that she has a sister. The very same complaint is just as apt if S falsely presupposed that she has a sister and asserted that she had to pick her up:

(1*) (to a friend) I’m sorry, I’m running late [for our coffee date]; I’ve had to pick up my sister from the airport today.—What do you mean? That can’t be true! I know for a fact that you are an only-child!?

Note that complaining may be more complicated here than it is for assertion. As we already saw in Sect. 1 and later in II with von Fintel’s test, a hearer needs to specify that they are complaining specifically about the presupposition. A general expression of disagreement (“No!”) would be taken to disagree with the main content of the utterance, in this case with the assertion that the speaker is late. For presuppositions, complaining about falsity may be more involved, but it is as apt as it is for assertions.

This clearly demonstrates: we have three related reasons—about knowledge norm, guarantee, and complaint—to acknowledge presupposition’s testimonial capacities. The dominant understanding on which assertion is paradigmatic testimony is incomplete as it overlooks presupposition. It also means that those who explicitly argue that only assertion is testimony (I looked particularly at Fricker, Moran, and Goldberg) are mistaken: they wrongly exclude presupposition.Footnote 20

4 Common Knowledge Norm

One small modification is needed to make what I presented here fully coherent. As it stands there is a problem: assertion and presupposition are testimony because they are both constituted by a knowledge norm; but, following Williamson, a constitutive norm is unique to a particular speech act. So either (i) the knowledge norm is not unique or (ii) one of either assertion or presupposition is not constituted by it. I think we can defend both options in ways in which the problem about unique norms is averted, and so my argument about testimony by presupposition stands.

First, we can defend (i). For Williamson, “assertion is the unique speech act A whose unique rule is the C rule” (2002, 241; my italics). This claim is different from the claim that assertion is the unique item whose unique rule is the C rule (which for Williamson is the knowledge rule). Assertion may be the only speech act that is governed by a knowledge rule, but things other than speech acts may also be governed by thea knowledge rule. I suggested that presupposition is not a full speech act of its own, but rather something García-Carpintero calls ancillary speech act. In this way, we can say that Williamson is right about the individuating knowledge norm for the speech act that is assertion while I am also right about presupposition’s knowledge norm.

Second, we can defend (ii) in a way that also does not harm my argument about presupposition as testimony: we will see that a new norm for presupposition entails the old knowledge norm.

Von Fintel’s “Hey, wait a minute!” test points towards a potential difference between the knowledge norm for presupposition and that for assertion: in order to presuppose p not only the speaker has to know p, the audience has to know p too. Otherwise an audience’s “Hey, wait a minute, I [as a member of the audience] didn’t know…” would not be a complaint. García-Carpintero therefore suggests a “common knowledge “ norm for presupposition: “For one to presuppose p in a context is correct if and only if p is common knowledge in that context” (2020, 272).Footnote 21 Something is common knowledge in a context iff it is known by all individuals involved in that context: both speaker and audience. A common knowledge norm combines two norms into one: a norm for speaker knowledge and a norm for audience knowledge.

This solves our problem about the uniqueness of constitutive norms. Assertion and presupposition are constituted by different constitutive norms, one unique to assertion, the other unique to presupposition. However, we are now faced with a new problem about the testimonial function of presupposition. If an audience ought to already know what a speaker presupposes, how can this presupposition function as testimony, ideally inducing new belief or knowledge? If the common knowledge norm is adhered to, a presupposition could only convey what is already known anyway. If the common knowledge norm is adhered to, it seems presupposition cannot successfully function as testimony. I defend García-Carpintero’s common knowledge norm against this challenge.

We can make at least the following two points to this end. First, even when the content of a presupposition is common knowledge between a speaker and her intended audience, this presupposition may function as testimony for an unintended audience, a bystander who accidentally overhears the conversation. Example sentence (2*) illustrates this well. Here, Phyllis’ mother tells her father: “Little Phyllis can crawl so well already; surely, she will walk soon.” The mother’s intended audience is Phyllis’ father and since both mother and father know their daughter’s name, the common knowledge norm is adhered to. When Phyllis overhears the conversation, grasps the presupposition, and learns that her name is Phyllis, she is not the intended audience. Yet, the presupposition functions as testimony for her. Her mother did not violate the common knowledge norm and still her presupposition was testimony.Footnote 22

This example demonstrates that presupposition can function as testimony even when the common knowledge norm for speaker and audience is complied with. Presupposition can also function as testimony when the norm for audience knowledge is violated—a less severe violation than violating the norm for speaker knowledge, as we will see. In such a case, an audience may accommodate a speaker’s norm violation by quietly adopting the content of her presupposition. García-Carpintero leans on Lewis’ (1979) notion of accommodation and Stalnaker’s idea of an audience “tacitly recogniz[ing]” the presupposition (1974/1999, 51) to suggest exactly this: informative presuppositions may be accommodated, elevated to common knowledge (or belief) by a sympathetic audience (2020, 285–7). In this case, the common knowledge norm is violated in some sense—before accommodation the audience didn’t know. But, following García-Carpintero, it is adhered to in a different sense—after accommodation the audience now knows. What García-Carpintero brands accommodation could be called successful testimony on my account of presupposition as testimony.

From my example sentence (1*) “I’m sorry, I’m running late [for our coffee date]; I’ve had to pick up my sister from the airport today” we can see that accommodation of a presupposition will often seem appropriate, that perhaps violating the norm for audience knowledge isn’t so severe. An audience’s challenge “I didn’t know you had a sister” seems weak and slightly pedantic. Perhaps it is best understood as a curious, puzzled remark. It seems far less serious than a different challenge: “How can you know she’s actually your sister? Didn’t you say your mom lived in this polyamorous commune when she had you both?”Footnote 23 The first challenge—about audience knowledge—accuses the speaker of not being sufficiently sensitive to her audience’s epistemic state, while the second—about speaker knowledge—accuses the speaker of speaking falsely and conveying falsehoods to her audience.

Lack of sensitivity for other’s epistemic state may be forgiven, but speaking falsely is a serious norm violation. The damage the latter can do is far greater than that of the former. Only in that latter case can falsehoods be spread. Accordingly, the sanctions for violating the speaker knowledge norm are more severe than those for violating the audience knowledge norm. A speaker who speaks falsehoods may be excluded from a social group, may be trialed in court, may lose their job (when for example as a journalist they wrote falsehoods). A speaker who presupposes what an audience didn’t yet know may be met with a puzzled remark—“Hey, wait a minute, I didn’t know you had a sister! You’ve talked so much about your parents but never about her.” This remark registers the speaker’s norm violation, but there are no further sanctions. The epistemic norm for audience knowledge will routinely be violated in ways exemplified here and nobody will make too much fuss about it. This is in line with Williamson’s idea about constitutive norms: “Breaches of the rules of a game, language, or speech act may even be common,” as he notes (2002, 240). Later, he adds: “We are often quite relaxed about breaches of the rules of a game which we are playing. If the most flagrant and the most serious breaches are penalized, the rest may do little harm” (2002, 259). In precisely this way, breaches to the norm for speaker knowledge are often serious and will be penalised. Breaches to the norm for audiences knowledge may do little harm and are hardly penalised. Yet, that breaches to a norm are common does not mean that no such norm exists. Von Fintel’s ‘hey, wait a minute…’ test works as a test for presupposition precisely because presuppositions are governed by such a norm for audience knowledge.

Taken together, we can say that presupposition is constituted by an asymmetric knowledge norm. It consists of two distinct norms: for speaker knowledge and for audience knowledge. The norm for speaker knowledge is binding, violations are severely sanctioned. The norm for audience knowledge is less binding, violations are more likely tolerated. The binding norm for speaker knowledge enables presupposition to function as testimony since speakers thereby vouch for the truth of their presupposed proposition. The less binding nature of the norm for audience knowledge makes it possible for presupposition to actually convey messages to one’s audience via testimony since it enables speakers to get away with violating this norm.

My argument stands. We saw that assertion is testimony because it is constituted by a knowledge norm (for the speaker), because a speaker thereby offers a guarantee for the truth of her assertion when asserting, and because a hearer can complain if the asserted proposition is false. The very same holds for presupposition—only that its constitutive knowledge norm may be a common knowledge norm, made up of a binding norm for speaker knowledge and a less binding norm for audience knowledge.Footnote 24 This argument amends the dominant picture on which assertion is paradigmatic testimony and it challenges the view that only assertion is testimony.

In Sect. 1, we surveyed some advantages of presuppositions: they have social inclusionary as well as exclusionary power; they seem particularly good at inducing belief; and they are difficult to challenge—both for linguistic and normative reasons. These features give a presupposing speaker a lot of epistemic and social power. Yet, once we understand presuppositions as testimony we can see how an audience has some counterbalancing power. A speaker may presuppose what they don’t know in order to avoid challenge and induce belief in her audience through the back door. Yet, an audience need not allow them that trick. Due to presupposition’s knowledge norm, anyone who presupposes is accountable for what they presuppose. The right to hold them accountable lies with the audience.