Norms of Constatives

According to the normative approach, speech acts are governed by certain norms. Interestingly, the same is true for classes of speech acts. This paper considers the normative treatment of constatives, consisting of such classes as assertives, predictives, suggestives, and more. The classical approach is to treat these classes of illocutions as species of constatives. Recently, however, Simion (Shifty Speech and Independent Thought: Epistemic Normativity in Context, Oxford University Press, 2021) has proposed that all constatives (i) are species of assertion, and (ii) are governed by the knowledge norm. I defend the classical treatment of constatives and show that Simion’s conclusion is untenable. No taxonomy of speech acts can accommodate such a view. More importantly, we can test whether a particular speech act is an assertion or not. I propose five tests of assertion, the passing of which is a necessary condition for being an assertion. Some constative speech acts fail these tests. Thus, contrary to Simion, not all constatives can be regarded as species of assertion.


3
One way to unpack this claim is to say that an assertion is the only speech act governed by this norm. Thus, if we want to discriminate assertions from other speech acts, we can ask whether the speech act in question is governed by KNA. This is a crucial question because the speech act of assertion performs many unique functions. For instance, only assertions are considered to be lie-prone (Stainton 2016). Thus, without separating assertions from other speech acts, we cannot have a definition of lying.
The normative view can be extended beyond assertions and applied to other speech acts or classes of them. Thus, just as we can propose a norm of assertion, we can do so for the class of assertives. For instance, Bach & Harnish (1979) propose that such speech acts as declaring, maintaining or propounding belong to assertives. These illocutions, just as assertions, are generally performed by means of flat-out declarative statements (Williamson 2000, 258). It is why such speech acts are often considered to be indistinguishable from assertions. The standard approach is to treat the notion of assertion as denoting a broader class of illocutions, to which the above speech acts belong. To put it in normative terms, if all of them are assertions, all of them are governed by KNA. Consider, on the other hand, conjectures. Contrary to assertions, they are standardly performed in an explicitly performative way, i.e., "I conjecture that p." More importantly, conjectures are not subject to KNA, i.e., one's conjecture is epistemically appropriate when one does not know that p. Rather, conjectures are considered to be distinct speech acts governed by a different norm.
Recently, Simion (2021, ch. 7) has proposed a unificatory treatment of all constatives as governed by KNA. Constatives are basically all speech acts that are made in a declarative mood, like assertions, predictions, conjectures, suppositions, and retractions. Here is her reasoning in a nutshell: CA All constatives are species of assertion. KNA Knowledge is the norm of assertion. KNC Knowledge is the norm of all constatives.
To individuate particular constatives, Simion adds special conditions to each class. I argue that Simion's unificatory view is untenable in this form. It unnaturally stretches the notion of assertion and misrepresents what constative communication is about. I show that Simion's proposal fails on two levels. Firstly, none of the available taxonomies of speech acts can accommodate her view. Secondly, and more importantly, we can empirically test the crucial premise CA in Simion's reasoning, and if we do so, it turns out to be unjustified. If Simion is right that all constatives are species of assertion, then all constatives should pass tests that show whether a particular speech act is an assertion or not. However, it is not the case. I show that some constatives deliver mixed results and some fail these tests. The classical approach in which assertions are species of constatives remains the only option. Furthermore, properly applying these tests can bring us closer to answering the bigger question, namely, "What is an assertion?".
The plan is as follows. After reconstructing Simion's proposal, I move to the two mentioned critiques of Simion. I focus especially on the latter strategy that provides a universal way of testing assertions. There is no single criterion that can separate assertions from non-assertoric speech acts. However, passing the submitted criteria provides necessary conditions for assertoric speech. In the final section, I propose a modification of Simion's proposal. Although it does not satisfy Simion's original purpose, it allows expansion for assertoric speech to certain classes of constatives. I also put her view into a broader context by showing that even the most inclusive proposals among norms of assertions are not as radical as Simion's view. Simion's (2021, ch. 7) claim that knowledge is the norm of constatives (KNC) 2 is based on two premises, i.e., (i) all constatives are species of assertion (CA), and (ii) knowledge is the norm of assertion (KNA). However, to capture the detailed characteristics of all constatives is more complicated. The core of her proposal is taken from Bach and Harnish's taxonomy that consists of 15 classes of constatives. I will not argue against any specific subnorms that Simion proposes; instead I will restrict myself to discussing only certain classes. Table 1 presents some examples in the format proposed by Bach & Harnish (1979, 42-44).

Simion's Account
Bach and Harnish operate in an intentionalistic framework, i.e., what distinguishes speech acts is the speaker's intention. Simion takes a different route and proposes a "normative transfer." This transfer consists of two steps. The first is CA, i.e., the claim that all constatives are species of assertion. Simion motivates her claim by arguing that this is uncontroversial since "on the Bach and Harnish taxonomy, all constatives are actually species of assertion" (2021, 91). For instance, predictives are supposed to be treated as assertions about the future, while descriptives are seen as assertions concerned with property attribution. Consider how Simion justifies counting suggestives as a species of assertion: … let us dub q the proposition 'There is reason, but not sufficient reason, to believe that p.' The above account of suggestives then becomes: Suggestives are expressions of (i) the belief that q, and (ii) the intention that H believe that q. As such, suggestives with content p collapse into assertions with content q. (2021,92).
In sum, suggestives that p are assertions that q, i.e., assertions that there is reason, but not sufficient reason, to believe that p.
Here we enter the second step of the normative transfer in which Simion proposes specific norms for each class of constatives. If all constatives are species of assertion, they are all governed by the norm of assertion. Simion favours KNA and generalises it to all constatives, hence she obtains KNC. In practice, the norm 1 3 of a particular class of constatives will be knowledge plus some other conditions. Table 2 presents the norms of classes listed above (Simion, 2021, 94-95).
It is worth noting that some of the norms do not mirror KNA in a straightforward sense. Take two examples: predictives and suggestives. Predictives can be treated as special cases of assertions, i.e., assertions about the future. According to Simion, my predictive (e.g., a prediction) that it will rain tomorrow is proper if I know that it will Table 1 Selected classes of constatives according to Bach & Harnish (1979) Assertives (affirm, allege, assert, aver, avow, claim, declare, deny (assert... not), indicate, maintain, propound, say, state, submit) In uttering e, S asserts that p if S expresses: i. the belief that p, and ii. the intention that H believe that p Descriptives (appraise, assess, call, categorise, characterise, classify, date, describe, diagnose, evaluate, grade, identify, portray, rank) In uttering e, S describes 0 as F if S expresses: i. the belief that 0 is F, and ii. the intention that H believe that 0 is F Concessives (acknowledge, admit, agree, allow, assent, concede, concur, confess, grant, own) In uttering e, S concedes that p if S expresses: i. the belief that p, contrary to what he would like to believe or contrary to what he previously believed or avowed, and ii. the intention that H believe that p Predictives (forecast, predict, prophesy) In uttering e, S predicts that p if S expresses: i. the belief that it will be the case that p, and ii. the intention that H believe that it will be the case that p Suggestives (conjecture, guess, hypothesise, speculate, suggest) In uttering e, S suggests that p if S expresses: i. the belief that there is reason, but not sufficient reason, to believe that p, and ii. the intention that H believe that there is reason, but not sufficient reason, to believe that p Suppositives (assume, hypothesise, postulate, stipulate, suppose, theorise) In uttering e, S supposes that p if S expresses: i. the belief that it is worth considering the consequences of p, and ii. the intention that H believe that it is worth considering the consequences of p Table 2 The results of normative transfer applied to classes of constatives from Table 1 KN-Assertives One's assertive with content p is epistemically permissible only if one knows that p KN-Descriptives One's descriptive with content 0 is F is epistemically permissible only if one knows that 0 is F KN-Concessives One's concessive with content p is epistemically permissible only if one knows that p, where p is contrary to what he would like to believe or contrary to what he previously believed or avowed KN-Predictives One's predictive with content p is epistemically permissible only if one knows that it will be the case that p KN-Suggestives One's suggestive with content p is epistemically permissible only if one knows that there is reason, but not sufficient reason, to believe that p KN-Suppositives One's suppositive with content p is epistemically permissible only if one knows that it is worth considering the consequences of p rain tomorrow. The subject of the knowledge norm is the content of the utterance. However, in the case of suggestives, the knowledge norm is about one particular aspect of the utterance, i.e., that I have a reason (but not a sufficient one) to believe that p. If I make a suggestive (e.g., a conjecture) that it is raining, the subject of the knowledge norm is not the content of my utterance, but the quality of the reason that I have for this claim. Thus, although Simion claims that all constatives are subject to KNA, the knowledge norm is applied to particular norms in different ways. Before going further, I think it will be helpful to see the crucial difference between Bach and Harnish's and Simion's approach. Consider Table 3. Bach and Harnish propose a classical taxonomy in which we have a general category, i.e., constatives, that encompass smaller categories. For this reason, assertives or predictives are species of constatives. Each class of constatives consists of particular speech act types. Bach and Harnish make space for differences between speech act types in a particular class of constatives. Simion, on the other hand, argues that we should think about constatives as species of one particular speech act, i.e., assertion. As a consequence, all constative speech acts should be regarded as assertions. Consider that for Simion, assertion is both the main category that encompasses all constative speech acts and a particular speech act that belongs to assertives. There is also another significant departure from classical taxonomies of speech acts. Simion focuses solely on classes, not particular speech acts. Because of that, there is no meaningful distinction between the class of suggestives and particular suggestives like conjecturing or hypothesising.
I will not argue against any particular norm that Simion proposes. Instead, I will focus on the first step in her argument, i.e., CA. If constatives are not a species of assertion, then any norm of assertion, even the weakest one, cannot be the norm of constatives. In the next section, I show why the normative transfer does not have a theoretical basis.

A Speech-Act-Theoretic Critique of Normative Transfer
In this section, I argue that Simion's claim that constatives are species of assertion (CA) is untenable. Firstly, I show why this misrepresents Bach and Harnish's proposal. Secondly, I indicate why such a view is inconsistent with any taxonomy of speech acts. Let me start by showing what the problem is with Simion's interpretation of Bach and Harnish's view. They say that "In general, a constative is the expression of a belief, together with the expression of an intention that the hearer form (or continue to hold) a like belief" (1979,42). When we compare this to the characteristic of assertives (i.e., an expression of a belief that p and an intention that the hearer will believe that p), we can see a significant resemblance. From this comparison, Simion derives CA.
When we read carefully the above characteristic of constatives, we can see that it is fairly general. It merely says that a constative must satisfy two conditions and that they must be in sync. There are two things to notice here. Firstly, Bach & Harnish (1979, 44-45) state that there is a lot of variability in the strength of belief both inside particular classes and between them. For instance, the belief expressed with a predictive is weaker than the belief expressed with an assertive. Secondly, there are classes that do not require even a weak belief that p. As they argue, "suggestives and suppositives that p are not expressions of belief that p" (1979,46). They repeatedly emphasise the differences between classes of constatives to show that all classes are species of constatives. Even though in most cases the speaker must believe that p, we still have clear outliers in the whole taxonomy. Thus, Bach and Harnish's taxonomy does not support Simion's generalisation that all constatives are species of assertion. 3 An analogous situation arises with other taxonomies of speech acts. What Searle (1975) calls representatives and Searle & Vanderveken (1985) call assertives roughly corresponds to Bach and Harnish's constatives. Neither specifies a detailed list of classes of constatives, but in both cases, it is clear that the main category (representatives or assertives) serves as an umbrella term for many distinct speech act types. In these cases, it is also unwarranted to claim that all representatives or assertives are species of assertion. We must be cautious when using this terminology: when Searle & Vanderveken (1985, 188) classify hypothesising or guessing as assertives, they do not mean that these speech acts are species of assertion, rather that they are species of constatives (in Bach and Harnish's nomenclature). Searle (1975, 354), in characterising representatives, claims that they all share some core features, like expressing belief and committing the speaker to something being the case. However, he makes a similar note to Bach and Harnish when he says that "The degree of belief and commitment may approach or even reach zero" (1975,355). An assertion may be a paradigmatic case of a representative (or assertive or constative), but all taxonomies of speech acts treat it as one speech act among others. The consensus is that assertions are species of constatives. 4

Norms of Constatives
Even if the taxonomy of Bach and Harnish, or other taxonomies of speech acts, do not support Simion's thesis, this does not mean that independently of any theory, constatives cannot be seen as a species of assertion. In fact, Simion suggests something like this when she says that her claim is "eminently plausible no matter one's theoretical commitments with regard to the nature of speech acts" (2021,96). 5 In the next section, I propose to test this hypothesis by applying various tests or criteria of assertion to some classes of constatives. The picture that emerges unequivocally shows that at least some classes of constatives are not species of assertion.

Tests of Assertion
There is a vast disagreement over which illocutions should count as assertions. 6 Furthermore, there are also multiple answers to the question "What is the norm of assertion?" TakeLying conjecture predictions. Some classify them as assertions (for instance, supporters of the truth norm), 7 some not (like advocates of KNA). Thus, various norms of assertion have different extensions: EXTENSION. Various norms of assertion denote different clusters of illocutions as belonging to assertions.
One of the functions of a constitutive norm of assertion is that it should separate assertions from other speech acts. However, EXTENSION shows that this is highly problematic. Ideally, delineating which speech acts are assertions should be done independently from any theoretical considerations. This is why, in this section, I apply certain widely recognised (but not yet used together) tests of assertion that are independent of any account of assertion. I start by introducing the tests and showing that any illocution that we want to call an assertion should pass them. Next, I put these tests to work. I divide examined constatives into three groups of cases: those that pass the tests, those that deliver mixed results, and finally, those that fail the tests. This makes clear that only some constatives pass these tests and thus can be considered as candidates for being an assertion. This indicates that independently from any theory of assertion, constatives are not a species of assertion.

Five Tests of Assertion
The empirical tests are based on widely accepted observations of how we make and react to assertions. There have been multiple explanations for these tests, but I am interested only in the fact that they can be used, especially when taken together, to differentiate assertions from other speech acts. 8 The first test concerns a specific way in which assertions can be challenged: 9 CHALLENGE. Assertions can be challenged by the "How do you know?" question. 10 For instance, if I assert that there is milk in the fridge, you can challenge my assertion by asking "How do you know?".
Secondly, there are certain constructions that are inappropriate for assertions. The most widely discussed are Moorean assertions, hence MOORE: MOORE. Assertions of the form "p, but I don't know that p" are judged as infelicitous. 11 In MOORE, one both asserts that p and denies that one knows that p, for instance, by saying "There is milk in the fridge, but I don't know that there is milk in the fridge." It is considered as infelicitous to firstly assert that p and deny it in the second conjunct.
The third test concerns probabilistic assertions: PROBABLE. Assertions based on merely probabilistic grounds are inappropriate. 12 Saying "Your lottery ticket did not win" without knowing the result of the lottery, even when the chance is very low, is inappropriate.
The fourth test is based on a close connection between assertions and lies: LYING. Assertions are lie-prone. 13 I can lie to you by asserting that milk is in the fridge because asserting something entails believing that p and lying necessarily involves insincerity, i.e., not believing that p.

Norms of Constatives
The final test concerns the relationship between assertions and retractions: RETRACTION. Assertions that turn out to be false are expected to be retracted. 14 This test considers one of the speaker's obligations in asserting something. If after my assertion that there is milk in the fridge you discover that this is not the case, it is reasonable to expect from me some kind of retraction. Retractions can be made in multiple ways, for instance by saying "Oh, sorry, you are right. I forgot that I drank it this morning." The proposed tests are accepted by almost all theories of assertion. Passing these tests is considered to be a necessary condition for being an assertion. However, it is crucial to emphasise that illocutionary acts that pass these tests should not be automatically treated as assertions. I will discuss this issue in the final section. Further, caution is required because some of these tests do not track merely assertoric content but also some forms of projective content, i.e., presuppositions and conventional implicatures. 15 Finally, Simion herself follows most of these tests when discussing the speech act of assertion. 16

Some Constatives Pass the Tests
In my analysis, I focus on the classes of constatives introduced in "Sect. 2." The first group consists of descriptives and concessives and I choose one speech act of each class, i.e., describe and admit, respectively. I deliberately choose one speech act that is naturally made in the form of a flat-out statement, i.e., we do not say "I describe that it's stormy and cold," but just "It's stormy and cold." The second speech act, on the other hand, naturally takes a performative form, i.e., "I admit that it's stormy and cold." Standardly, assertions are performed as flat-out statements. The case of admitting shows that performative speech acts can also pass the tests.
Let us start with CHALLENGE. One can be asked to describe the weather and respond with "It's stormy and cold." Similarly, someone who thought it was sunny can admit the error by saying "I admit that it's stormy and cold." In both cases asking "How do you know?" is a legitimate challenge.
Consider MOORE. Both statements are infelicitous: (1a) #It's stormy and cold, but I don't know that it's stormy and cold.
(1b) #I admit that it's stormy and cold, but I don't know that it's stormy and cold.
1 3 (1a) is straightforward since it resembles a classical Moorean assertion. However, (1b) does not work either; the fact that it is a performative statement is irrelevant.
For PROBABLE, imagine that we are in the most stormy place on Earth (which is probably Bogor on the island of Java where there are 322 storm days every year). It is still inappropriate to say, without looking out of the window, "It's stormy and cold outside," if this statement is based only on a high probability. To see this, consider that one may oppose this statement by asking "How do you know?" which can reveal that the speaker does not know what the weather is like outside. The same can be said about admitting. Imagine the following context: having watched the weather forecast a while ago I say that it is sunny. However, you confidently assert "No, it's stormy and cold!" Just to let you know that I agree with you, I respond with "Oh right, I admit that it's cold." If I knew that your judgement was made purely on probabilistic grounds, my admission seems inappropriate.
Both admissions and descriptions are lie-prone. If I say "It's stormy and cold outside" while believing that it is not so, I am lying. Similarly with admitting. It does not matter that my utterance is prefaced. By saying "I admit that it's stormy and cold" without believing this to be the case, I am lying. A prefaced admission is no less committal than a flat-out assertion.
Finally, both speech acts pass the RETRACTION test. If we go out and it turns out that it is not stormy and cold, I should retract my acts of describing and admitting that it was stormy and cold.
In sum, describing and admitting pass all the tests of assertion. This is the result that Simion expects. Assuming that these two cases are paradigmatic for descriptives and concessives, at least prima facie, these classes of constatives can be seen as species of assertions.

Some Constatives Deliver Mixed Results with Respect to the Tests
Let us now consider predictives, and focus on the paradigmatic speech act of this class, i.e., (explicitly performative) predictions. Simion says that Several of [classes of constatives] are species of assertion in a straightforward way that is not particularly interesting, nor should it be a matter of much controversy: predictives, for instance, are assertions about the future, while retrodictives are assertions about the past. (2021, 91).
There is no consensus about the question whether predictions are assertions or distinct kinds of speech acts. 17 I think that predictions give at best mixed results for the tests of assertion. We can observe the following pattern: when assertions are infelicitous, predictions are often appropriate. This is especially so for explicit predictions. Consider the following example for CHALLENGE: It is often the case, especially for short-term predictions, that "How do you know?" questions are appropriate. However, A could also respond to B by saying something like "Oh, I don't know of course, but can you see these clouds?" Consider the second case that shows that the epistemic threshold for predictions is, at least in some cases, weaker than for assertions.

Norms of Constatives
(3a) A: (I predict that) Humans will colonise Mars. (3b) B: ?How do you know? (3c) A: What do you mean? I cannot know that, I just think so because… With or without the performative verb, A makes a proper prediction. Further, because A does not know the future, one cannot know the content of one's prediction at the time of carrying out the act. Judgments whether B's challenge is appropriate, may vary. It is sufficient to see that CHALLENGE does not deliver clear results. Often, especially in the case of long-term predictions, weaker challenges are more appropriate (like "What makes you say that?" or "Do you really believe that?").
Consider MOORE with two sets of cases.
(4a) ?It will rain, but I don't know that it will rain. (4b) I predict that it will rain, but I don't know that it will rain. (4a') ?Humans will colonise Mars, but I don't know that Humans will colonise Mars. (4b') I predict that Humans will colonise Mars, but I don't know that Humans will colonise Mars.
(4a) seems infelicitous and hence passes the test. (4a') is more difficult to assess. Just as with CHALLENGE, short-term predictions can be evaluated more strictly and so are closer to assertions. Long-term predictions are usually more uncertain; hence, we expect less from them. (4b) and especially (4b') are felicitous and coherent (see Cariani 2020). Thus, at least some predictions fail this test.
It is felicitous to make a prediction based on probabilistic grounds. Consider that challenging an explicit prediction seems to be inappropriate: (5a) A: I predict that your ticket will lose the lottery. (5b) B: ?How do you know? (5c) A: I don't! This is just my prediction! Predictions in such contexts are felicitous because they express one's expectation about the future, rather than one's knowledge (Benton & Turri 2014). However, lottery predictions can be also made in a non-performative way. Such cases seem to be closer to assertions: (6a) A: Your ticket will lose the lottery. (6b) B: ?How do you know? (6c) A: I don't, but you know that your chances of winning this are close to zero.
It may be that some flat-out predictions are judged as assertions, but performative ones have lower epistemic standards.
LYING brings mixed results too. We can lie only if we do not believe in what we say. Predictions involve weak or partial belief. One can say "It will rain tomorrow" without believing so, with an intention to deceive the hearer. I grant that some may see deception and possibly a lie here. In the case of performative predictions, like "I predict that Humans will colonise Mars," the lying judgement can be much weaker but there is definitely room for deception.
I argued that we are expected to retract false assertions. Consider the case of false prediction: (7a) A: (I predict that) It will rain tomorrow. (The day after tomorrow when it is known that it was not raining.) (7b) B: ?You were wrong, it did not rain! (7c) A: ?Right, sorry, I take it back.
Both B's critique and A's response do not sound natural. As I noted, predictions can be challenged (usually in a weaker way than assertions) but because they concern the future, one cannot know their outcome. Thus, even when they turn out to be false, demanding their retraction sounds improper. 18

Some Constatives Fail the Tests
Let us move to the final group of classes of constatives that I want to discuss, i.e., suggestives and suppositives. I will consider the paradigmatic cases of each class, i.e., conjecture 19 and supposition, respectively.
Consider CHALLENGE. Imagine that a group of friends did not see the Euro 2020 final and one makes the following conjecture: (8) I conjecture that England won the Euro 2020. Simion rightly observes that conjectures require "access to one's reasons for conjecturing" (2021, 120). Thus, one may ask what are the reasons for making the conjecture (8). However, asking the "How do you know?" question is too strong and implies that 18 It may seem that predictions are unique in delivering such mixed results for the tests. However, there is a parallel class of constatives, i.e., retrodictions, that can be characterised as predictions about the past. Some treat these classes as two sides of the same coin, see, e.g., Weiner (2005) and Lackey (2007, fn. 33). I will not discuss this class here. 19 The results are even clearer for many other suggestives like guesses or speculations. The reason is that they are governed by weaker norms than conjectures. the inquirer did not notice that the statement is hedged. One can still inquire what the reasons are by asking, for instance, "What makes you say that?" In this respect, conjectures are similar to predictions. The reason for the failure of the test is not because conjectures are standardly made in a form of an explicit performative. A conjecture can also be made without a hedge, for instance, when it is embedded in a dialogue: (9a) A: What is your conjecture? Who won? (9b) B: England won.
Here, it is also inappropriate to challenge the conjecture with the "How do you know?" question. Suppositions behave differently. After saying.
(10) Suppose that England won the Euro 2020. The audience is not entitled to ask for any reasons or challenge (10). 20 Suppositions are non-committal speech acts, i.e., one does not need to have any reasons for p in order to state that p. 21 Demanding justification for supposing something can be rather confusing and is a sign of not playing the same language game. One may engage in a meta-critique of a supposition but this is not a critique of its content. Thus, for different reasons, both conjectures and suppositions fail the first test of assertion. In the case of the other tests, the results are analogous, i.e., conjectures do not pass the tests because, compared to assertions, they commit the speaker in a much weaker way, and suppositions do not commit the speaker at all.
Some may argue here that this critique mischaracterizes Simion's account. She proposes to treat conjectures as assertions that there is a reason, but not sufficient reason, to believe that p, and suppositions as assertions that it is worth considering the consequences of p. So, some may claim that the tests should be applied not to the flat-out assertions that p, as I do, but to assertions with the mentioned extensions. Thus, instead of asking "How do you know that England won the Euro 2020?" one should ask "How do you know that there is a reason, but not sufficient reason, to believe that England won the Euro 2020?" for conjecture and "How do you know that it is worth considering the consequences of England winning the Euro 2020?" for supposition when applying Simion's account. Amended in this way, the challenge of supposition sounds unnatural and overly strong. However, the challenge of conjecture now seems to work. When one conjectures (9), someone can challenge one's reason for making such a claim. As I will show, Simion's solution now seems to work for some of the tests, but still not for all of them. 22 20 Shields (2021) makes a similar point about stipulations that, I think, is transferable to suppositions: "To say, "That's false…" or "It's not the case that…" in response to a stipulation is infelicitous because it mischaracterizes what I am trying to do with my words…" (2021, 6). 21 Earlier I said that Bach and Harnish's constatives roughly correspond to other taxonomies of speech acts, like Searle's representatives. However, suppositions are notable exceptions. We do not find suppositions in Searle's taxonomy because they do not pass the basic condition for being representatives, i.e., they do not have a word-to-world direction of fit. Instead, they are often analysed as a form of pretence, see, e.g., Green (2000) and Recanati (2021). 22 I follow Simion's proposal throughout all the tests; however, I want to emphasise that her proposal waters down the tests of assertion. Simion aims to show that conjectures are species of assertion and, to do that, they must pass the tests in the version designed for flat-out assertions. Thus, even if the modified versions of the tests seem to work, they do not say anything about assertions, and thus do not prove that conjectures are species of assertion. Nevertheless, even following Simion's interpretation of the tests, conjectures do not pass all of them. Thanks to the reviewer for encouraging further clarity on this point.
(11b) ?Suppose that it's raining, but I don't know whether it is worth considering that it is raining.
Just as above, I think that the supposition sounds odd, but the conjecture seems to deliver the right result for the test.
Let us move to probabilistic statements. Intuitively, one can say, without knowing the results of the lottery, "I conjecture that you didn't win the lottery" or "Suppose that you didn't win the lottery" and not be criticised for making purely probabilistic claims. After all, one can justify one's conjecture by saying that the probability is very low, and supposition does not need any explanation. Are the results different when we take under consideration Simion's approach? I cannot see how we can justify suppositions but there seems to be a way out for conjectures. Some may argue that my conjecture about your lottery ticket is impermissible (hence, it passes the test) since if it would be permissible to conjecture that one ticket lost, one could make similar conjectures about all other tickets. Because conjecturing that all tickets lost is impermissible, it is also impermissible to conjecture that a particular ticket lost (see Simion 2021, 120, fn. 11).
Consider LYING and the following simple story: Lying Conjecture A group of friends did not see the Euro 2020 final. Liv says "I conjecture that England won the Euro 2020" without having any reason to believe that this is true. By making this conjecture, she wants to trick her friends into believing that she has some reason to believe that England won. Does Liv lie by making her conjecture? According to LYING, lying is restricted to assertoric content. 23 Hence, following the traditional view, Liv makes a conjecture, and therefore is not lying. Consider two reasons that strengthen LYING. Firstly, virtually all accounts of lying agree with a distinction between lying and misleading. Lying is possible only by claiming that something is true and requires a certain level of commitment (cf. Marsili 2020b). For instance, Carson proposes that "To count as a lie, a statement must be warranted to a certain minimum degree" (2006,302). Following this criterion, Liv is not putting forward her claim as true, rather she merely indicates that she has some insufficient reason to believe the content of her conjecture; she is misleading. The hedging explicitly indicates that her commitment is weaker than it would be for an assertion. Secondly, one useful criterion for lying is the so-called deniability criterion. It is based on the observation that lies cannot be sincerely denied. 24 For instance, when someone challenges my assertion "England won the Euro 2020," I cannot sincerely deny claiming that England won. However, when Liv's friend accuses her of lying, she can deny claiming that England won; she can say that she merely conjectured it and that she did not claim that she believed that England won, which would be true.
It is even harder to make any case for lying with suppositions. One could say "Suppose England won the Euro 2020" without believing that this is true. However, in supposing something, one does not undertake any commitment and one's belief regarding the truth of the stated claim is not relevant.
Even embracing Simion's switch from p to q does not help here. Consider that it does not matter for our lying judgement whether we think about conjectures as distinct speech acts with the content p or as assertions with the content q. In both cases, the content remains the same. Following Simion, Liv's conjecture that England won the Euro 2020 is an equivalent of the assertion that there is a reason, but not sufficient reason, to believe that England won the Euro 2020. In the case of lying, Simion's approach does not deliver different results. Liv still neither believes that p, nor puts p forward as true, nor commits to p, and if accused of lying, she could object and say that she never claimed that p. As above, suppositions are even more problematic. Assume that my supposition that England won the Euro 2020 is an equivalent of the assertion that it is worth considering the consequences of England winning the Euro 2020. Even if I do not believe that it is worth considering the consequences of England winning, I still do not pass any criteria of lying. I do not claim that something is true, I do not commit to the stated claim and, if accused of lying, I can sincerely deny that I ever stated that England won. Thus, even if Simion's account may help in passing some tests, it does not work in all of them.
Finally, we have RETRACTION. Following Simion's account, if conjectures should be treated as assertions that q, then q should be the subject of retraction. Thus, one should retract a conjecture not when it is false but when one did not have a reason to make it. Consider the following example: (13a) A: I conjecture that England won the Euro 2020. (13b) B: But you do not have any reason to say that! (13c) A: Right, sorry, I take it back. For Simion, suppositions are assertions that it is worth considering the consequences of England winning. So, the retraction is supposed to take place when one does not know that this claim is worth considering. In such a scenario, one can retract one's supposition. Table 4 summarises this section. There are clear cases of classes of constatives that pass the tests. It is important to emphasise that explicit performatives can pass the tests just as well as flat-out statements. This is especially important when we analyse the two remaining groups. Predictives in general give mixed results. However, there are predictions that fail the tests and this is sufficient to say that at least some predictives are not assertions. Suggestives and suppositives give clear results for the standard versions of the tests (although for different reasons): all of them fail the tests, and thus cannot be considered as assertions. Even if we would accept Simion's approach, suggestives and suppositives fail at least the LYING test.

3 5 A Pluralist Approach to Constatives
In this final section, I discuss what the results from the tests of assertion mean for Simion's view and propose its modification to fit into the data. Subsequently, I contrast her proposal with two alternative norms of assertion with radically different approaches to EXTENSION. This juxtaposition allows us to see how different norms of assertion deal with the outcomes of the tests.
In the previous section, I directly tested Simion's claim that all constatives are species of assertion. In principle, we can see that her thesis could be extended to descriptives and concessives, since speech acts that belong to these classes pass the tests of assertion. Furthermore, if knowledge is the norm of assertion, then it is also the norm of descriptives and concessives. Notice that among the speech acts that pass the tests of assertion, there are such that naturally take on the performative form, like admitting. Thus, even though assertions are standardly made by flat-out statements, they are not restricted to such statements. Suggestives and suppositives fail all the tests and some predictives also fail the tests. Even taking under consideration Simion's approach, suggestives and suppositives do not pass all the tests. To emphasise, the tests provide the minimal requirements that a speech act must satisfy to be considered an assertion. Thus, Simion's proposal that all constatives are species of assertion is untenable.
Can Simion's proposal be amended? I think so. Bach & Harnish (1979) propose more classes of constatives than I discussed, i.e., ascriptives, informatives, confirmatives, retractives, assentives, dissentives, disputatives, and responsives. Most of them, if not all, can be classified as species of assertion. Simion's unificatory claim can be modified to count most constatives as species of assertions, i.e., those that pass all the tests of assertion. This proposal fits the data, but of course, significantly weakens her proposal.
So far I was focused exclusively on answering the question "Which speech acts can be classified as assertions?" However, there is also the question "Why can those speech acts be classified as assertions?" A full discussion of this point is beyond the scope of this paper but here is a direction in which the answer could go. Speech acts like admitting can be treated as assertions because they entail the illocutionary force of assertion (Marsili 2020a). This does not mean that admitting something is merely asserting it. Rather, one can express different illocutionary forces in one act, i.e., one can both assert and admit something (Searle & Vanderveken 1985). Another way of framing it is treating such speech acts as special modes of assertion (Alston 2000, ). For instance, admitting can be seen as an assertion with the additional normative requirement that the speaker is in a certain position. Finally, I would like to juxtapose Simion's proposal with two alternative views that present different approaches to EXTENSION. Consider firstly norms of assertion with a broad extension. Some of these norms classify as assertions even those speech acts that fail the above tests. The most radical cases are context-sensitive norms where, roughly, the epistemic standards for proper assertions shift with changes in context. Take McKinnon's (2015) supportive reasons norm, according to which one's assertion that p is permissible if one has supportive reasons for p. 25 McKinnon argues that this norm gathers under one label all speech acts from guessing to guaranteeing. For McKinnon, all illocutions in which the speaker commits to the content, even in a minimal sense, should be considered as assertions. This of course goes against the verdict of the tests of assertion. Notice that even such a radical proposal has a narrower extension than what Simion proposes. McKinnon wants to capture suggestives (she mentions guessing, which is a weak suggestive), but excludes suppositives. The reason is simple: the latter ones are non-committal speech acts. Even in the light of context-sensitive norms, it is hard to justify the extension of Simion's view.
To complement this picture, on the other side of the spectrum, we have norms with a narrow extension. Such norms count only some speech acts that pass the tests of assertion as assertions. As an example, take García-Carpintero's (2004, 156) transmission of knowledge norm: TKNA. One must: assert p only if one's audience comes thereby to be in a position to know p.
Both KNA and TKNA take knowledge as the norm of assertion. However, while KNA assumes that what licences proper assertion is the epistemic position of the speaker, TKNA proposes that the epistemic position of the audience should be in the centre. In what sense does TKNA have a narrow extension? Consider the speech act of reminding. Taking, as an illustration, the LYING test, intuitively, one can lie by saying "I remind you about the dentist appointment" when one does not believe that there is any dentist appointment. Searle & Vanderveken (1985, 185) provide a handy definition in which reminding is explicitly treated as an assertion: "To remind is to assert to a hearer with the additional preparatory condition that the hearer once knew and might have forgotten the propositional content." In Simion's terms, reminding is a species of assertion. Notice, however, that TKNA does not treat reminding as an assertion. One can remind something that the hearer already knows, so when one reminds something, there may be no transmission of knowledge. In the heart of TKNA lies the idea that assertions essentially aim to transmit knowledge. Thus, at least for some norms, passing the tests of assertion is insufficient to count particular speech acts as assertions.

3 6 Conclusions
My main aim was to show that the unificatory treatment of constatives proposed by Simion is inadequate. In philosophy, particularly in philosophy of language, there is much focus on the notion of assertion because it performs certain unique functions. For instance, the notion of lying is explained in terms of assertion. If we were to accept Simion's extension of the notion of assertion to all constatives, we would either have to accept the possibility of lying with such speech acts as suppositions or abandon the foundational assumption in the debate on lying. Neither is satisfactory.
Even if most classes of constatives pass the tests of assertion, this does not mean that they automatically should be treated as species of assertion. We must be wary that these tests provide only minimal requirements for being an assertion. The classical taxonomies of speech acts give resources to distinguish particular classes of constatives from each other. I think that this is the right path to proceed. Instead of broadening the extension of assertion, and blurring the boundaries between speech acts, we can focus on their variety and on what is distinct in each of them. Such an approach can bring us closer to an adequate treatment of constatives.

Conflict of Interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.