A recent idea about the nature of propositions is to reverse the order of explanation between the properties of propositions, such as being true or false, and the acts by which agents grasp them. Hanks and Soames have proposed theories of this sort (Hanks 2015; Soames 2015); see also Soames (2010), Hanks (2011, 2013, 2019), Hom and Schwartz (2013), King et al. (2014, chapters 6, 9, 12), Collins (2018), Recanati (2019) and Reiland (2019).

My topic is Hanks’ and Soames’ different accounts of predication. According to Soames, predication is neutral in the sense that an agent can predicate a property F of an object a without being in any sense committed to a’s being F. This is contrasted with judging that a is F, which involves making the predication and also affirming it. The affirmation, not the mere predication, is what carries the commitment (Soames 2015, 15–29).

Hanks denies that predication is neutral in this sense (Hanks 2015, 22). Furthermore, Hanks argues that the very idea of neutral predication is incoherent (Hanks 2015, 35). On this basis Hanks rejects Soames’ theory of what propositions are. Hanks then presents his theory as the best of those that adopt the guiding idea about reversing the order of explanation once Soames’ view has been rejected.

I will argue that Hanks cannot rely on his argument against Soames because an analogous argument could be made against Hanks’ own theory of questions and orders.

It will be helpful to have the core of both theories available to refer to. I will focus on Hanks’ theory, as presented in Hanks (2015), because the details of his view matter to the argument in a way that the details of Soames’ view do not. Let ‘Frank’ be the name of a certain goose and ‘honks’ a predicate that expresses the property of honking. The English sentences (1), (2), and (3) are used to say that Frank honks, ask whether Frank honks, and order Frank to honk, respectively.Footnote 1

  1. (1)

    Frank honks.

  2. (2)

    Does Frank honk?

  3. (3)

    Frank, honk!

Hanks’ theory is that each sentence is associated with a proposition of a different type: predicative, interrogative, or imperative.Footnote 2 Propositions are identified with the type of action performed by those who utter the sentences. The propositions have constituents, which are themselves types of act. Following Hanks’ notation, let ‘Frank’ denote the act of referring to Frank, and ‘HONKS’ denote the act of expressing the property of honking. Acts of predicating, questioning, and ordering are denoted by ‘\(\vdash \)’, ‘?’, and ‘!’, respectively. The propositions themselves can then be represented by combining these symbols.

  • \(\vdash \langle {\mathbf {Frank}},{\mathrm {HONKS}}\rangle \)

  • \(?\langle {\mathbf {Frank}},{\mathrm {HONKS}}\rangle \)

  • \(!\langle {\mathbf {Frank}},{\mathrm {HONKS}}\rangle \)

What this notation brings out is that predication is involved only in the first of these propositions: the one expressed by (1). Performing the other actions does not require predicating, and the type/proposition does not contain it. This is a good result if predication is not neutral, because the people who perform these actions are not committed to Frank’s honking. People who ask questions may be committed to other things, such as certain propositions being answers to their question, but these are distinct commitments.

This can be contrasted with an influential view about propositions defended by Searle, among others (Searle 1969; Recanati 2013). This is the traditional view about propositions. On this view, there is just one proposition, that Frank honks. The actions represented in Hanks’ notation should be thought of instead as actions directed at that proposition: asserting it, asking whether it is true, ordering that it be made true. Another way to put this is that the traditional view, but not Hanks’, is committed to a version of the content–force distinction (Geach 1965; Hanks 2007; Recanati 2013). As Hanks sees it, this is a relic of the Fregean way of thinking about propositions which made sense in that setting but should be rejected by proponents of the new style of theory (Hanks 2015, 40–41).Footnote 3 Hanks usefully distinguishes between two versions of the content–force distinction: the taxonomic and the constitutive (Hanks 2015, 18–19). On the taxonomic version of the view, the content–force distinction is ‘the view that speech acts with different forces all share the same truth-conditional contents’ (Hanks 2015, 19). As Hanks notes, this was not Frege’s own mature view: in Frege (1956) he identifies the content of assertions and questions, but distinguishes this from that of orders (Hanks 2015, 9 and 19). On the constitutive version of the view, the content–force distinction amounts to the claim that ‘there is nothing inherently assertive about the propositional contents of assertions’ (Hanks 2015, 9). As Hanks notes, it is the constitutive claim that is at the heart of his disagreement with Soames.

Soames’ view is not committed to one proposition being the content of (1), (2), and (3), as on the traditional view. Soames has suggested that propositions are to be distinguished from questions and directives (Soames 2019, 1382). However, Soames is committed to a form of neutral predication which unifies the content of, e.g., (1). So, Soames is committed to the constitutive but not the taxonomic version of the content–force distinction. Soames does not take this to be an unfortunate relic of the Fregean way of thinking about propositions. So, this is a key decision point in the development of a non-traditional theory of propositional content.

It is worth noting what Hanks and Soames agree on. They agree that there is an act of predication, and they agree that this act is sufficient to unify the constituents of propositions. Unifying, in this sense, entails at least whatever it is that gives propositions truth conditions. Hanks and Soames disagree about whether predication is neutral (or even whether neutral predication is a coherent notion).

Hanks makes an important general point about the content–force distinction; I expect that this question will guide the next stage of the debate over the nature of propositions. Rather than engage with these issues, my project here is to criticise Hanks’ direct argument against Soames’ commitment to neutral predication. So, my ambitions are limited. I am not claiming even to resolve the issue of whether Hanks’ or Soames’ view is the best version of their approach to propositions. Instead, I limit myself to refuting Hanks’ argument that Soames’ version of the view is incoherent.

Hanks sets up the argument in these terms:

... I do not understand Soames’s notion of neutral predication. I do not understand what it would be to attribute a property to an object while remaining completely neutral about whether the object has that property. To attribute a property to an object is to characterize the object as being a certain way. How is it possible to do that and yet not take any stand at all about whether the object is that way? That seems incoherent. (Hanks 2015, 36)

Hanks reasons as follows. Take a simple act of predicating, such as predicating F of a, where the object does not have the property. According to Soames, the act of predication has truth conditions. And this one is false, because a does not have F. Hanks writes:

Now suppose that a is not F. If so, then this act of predication is false and hence the agent did something incorrect. The agent made a mistake. But how could the agent make a mistake if she took no stand one way or the other about whether a is F? It is incoherent to suppose that an agent can make a mistake by predicating F of a while taking no position at all about whether a is F. (Hanks 2015, 36–37)

Hanks presents the argument in the following form, prefaced by ‘Begin by supposing that S performs a pure act of predicating F of a and that a is not F’ (Hanks 2015, 37).

  1. 1.

    S’s act of predication is false.

  2. 2.

    S’s act of predication is incorrect.

  3. 3.

    S made a mistake.

  4. 4.

    S must have taken a position about whether a is F.

  5. 5.

    S’s act of predication was not neutral.

We are supposed to draw the general conclusion that no act of predication can be neutral, at least not if acts of predication are to have truth conditions. Hanks then considers a response made by Soames to this argument which is to deny premise 1, and claim that token acts of predication do not have truth conditions, but types do (Hanks 2015, 37, footnote 9).

To deal with that response, Hanks later presents the following argument that does not refer to token acts of predication (Hanks 2015, 39). (The scenario is the same: S predicated F of a, and a is not F.)

  1. 1.

    S inaccurately represented a as F.

  2. 2.

    S made a mistake.

  3. 3.

    S must have taken a position about whether a is F.

  4. 4.

    S’s act of predication was not neutral.

I will focus on this argument because it is the one that Hanks relies on, and that Soames would be most interested in resisting.

If this is to be used as an argument against neutral predication, which is the use Hanks wants to put it to, then premise 1 must be read in such a way that both sides of that debate will agree to it. So, ‘S represented a as F’ must mean just that S predicated F of a: if representing entailed something stronger, then the target of the argument could just reject this premise. And, ‘inaccurately’ must mean just that a is not F. I will take premise 1 in that way; I think that everybody should accept it. I also take it that premise 3 follows from premise 2 and premise 4 follows from premise 3. So, the claim that must be evaluated is that premise 2 follows from premise 1.

The argument can be put in even more schematic terms, to allow us to see its structure. I will use Hanks’ notation for the combination of acts directed at objects and properties, and use X as a placeholder for acts such as predication.

  1. 1.

    S performed \({\mathrm {X}}\langle {{\mathbf {a}}},{\mathrm {F}}\rangle \), and a is not F.

  2. 2.

    S made a mistake.

  3. 3.

    S must have taken a position about whether a is F.

  4. 4.

    S’s act of X was not neutral.

The target of Hanks’ argument, i.e., Soames, or any other proponent of neutral predication, will now want to reject the transition from premise 1 to premise 2. Their idea will be that, on their view, it does not follow from S performing an act of predication when the targeted object does not have the expressed property that S made a mistake.

Why does Hanks think that premise 1 entails premise 2? I take it that his idea is the following. The act that we are interested in, representation, in this case, must be sufficient to unify a proposition suitable to be expressed by (1). That proposition must have truth conditions. Hanks then reasons that any act which unifies the act of referring to a and expressing F into a proposition with truth conditions must be such as to entail commitment to a’s being F. Such a justificatory story is required to secure the transition from premise 1 to premise 2 in a way that will threaten the neutral predication view.

This, I take it, is what Hanks has in mind with his metaphor of sorting a among the Fs (Hanks 2015, 22–23). The idea is that such an action has conditions for being accurate, which both suggests that the type can have truth conditions but also that performing it when a is not among the Fs is a mistake.

The problem with the thought just presented is that, if we accept it, an analogous argument to the one Hanks gives against neutral predication can be constructed against neutral questioning and ordering. I will argue for that conclusion in the remainder of the paper.

Interrogative propositions are propositions, on Hanks’ view: ‘Asking is one of the basic ways in which we combine objects with properties’ (Hanks 2015, 188). So, they enjoy a kind of propositional unity of the kind that needs to be explained for all propositions. What this means, in the present context, is that their constituents determine ‘answerhood conditions’ (Hanks 2015, 197). For example, the answer to the question asked with (2) is ‘yes’ if and only if Frank honks (whether, e.g., Sarah the sheep bleats is irrelevant). Given that the constituents of the various predicative and interrogative propositions are identical, and identically arranged, the difference between the two propositions can consist only in the presence of the predicative and interrogative acts represented by ‘\(\vdash \)’ and ‘?’. The view is therefore committed to the latter act providing propositional unity to the interrogative proposition. Furthermore, this act is sufficient for creating a unity in which a relationship between the constituents is encoded; that is the relationship that one must understand in order to understand what counts as an answer to the question. An example with a transitive verb may be helpful here: the question ‘Does Frank love Sarah?’ is understood only when the order of the constituents is understood. The act that unifies the question asked with (2) is neutral, in the relevant sense, because someone who performs it is not committed to Frank’s honking. Whether or not Frank honks, the agent who has asked the question has succeeded in combining Frank and the property of honking in a way that determines answerhood conditions that are to do with Frank and honking, without thereby being committed to Frank’s honking.

The problem for Hanks’ view is that the argument, presented as 1–4, used against neutral predication could be deployed against neutral questioning. Here is how that argument might go: Suppose that someone has asked whether Frank bleats, and he does not. This is a polar question which has a yes/no answer; it has answerhood conditions which determine which is the right answer. The agent has combined Frank and BLEATS into an interrogative proposition to which the answer is yes if and only if Frank bleats. So, the agent has performed some X on Frank and BLEATS which is sufficient for answerhood conditions. Because Frank does not bleat, however, the agent has made a mistake. The conclusion, absurdly, would be that the act of questioning is committal with respect to what is asked about. Of course, the obvious response is to deny that in the case of interrogative propositions the move from premise 1 to premise 2 is permissible, and to do so on the basis that the action X doing the unifying work here is one that does not carry commitment.

This brings out the core of the problem with Hanks’ argument. In all three sorts of proposition there is something which unifies the constituents, and which generates properties such as accuracy and answerhood conditions. At least some of these things, e.g., whatever does the unifying work in an interrogative proposition, are not committal, otherwise someone who asks whether Frank honks would be committed to Frank’s honking which would be absurd. So the notion of something which does this sort of unifying is not incoherent because it happens in some cases.

The problem for Hanks’ view now is that there is no obvious explanation of why would it be incoherent to hold that something noncommittal unifies predications. The justificatory story has been undercut by considering the other actions which unify complex acts into propositions. Someone who thinks that something noncommittal unifies predicative propositions will reject any version of Hanks’ argument 1–4 where X is replaced with their proposed noncommittal unifier. They will also deny that being assessable for accuracy is sufficient for being committal, and they will appeal to the analogy with questions and orders where it is possible to assess whether the answer to a (polar) question is affirmative, or whether a order has been carried out.

Hanks could at this point simply insist that his view just is that predicative propositions are unified by an act of predication which is committal. Furthermore, he might claim that this is essential to any predicative unifier. However, this cannot be his response in the particular context of an argument for such a view, based on the incoherence of the alternative view defended by Soames.

This creates a dilemma for Hanks. He can either deny that any noncommittal combining action can be sufficient for propositional unity, in which case he will have to deny that there are interrogative propositions etc. It is part of Hanks view, and indeed Soames’, that there are interrogative propositions (Hanks 2015, chapter 9). To deny this would be to accept the taxonomic content–force distinction which Hanks rejects.

Alternatively, Hanks can accept that there are neutral combining actions that are sufficient for propositional unity. It does not follow from accepting the second horn that predication is one of these neutral combining actions, but it is hard to see how Hanks can deny that this possibility is coherent.

Note that my argument does not target the idea that there are different types of propositions unified by different acts. My point is that Hanks has to accept that some of these acts are neutral. If so, he cannot claim that it is incoherent that the one that unifies predicative propositions is neutral.

I will now consider two possible replies, the second of which I think is suggestive of an interesting line of inquiry. Firstly, one might try to adapt what Hanks says about predication that does not carry commitment to the interrogative case; this would be a way to defend the first horn of the dilemma. Hanks claims that there are contexts which cancel the commitment of predications. This allows Hanks to respond to the objection to his view that not all acts of assertion of complex propositions carry commitment to all the atomic propositions embedded in those complex propositions (Hanks 2015, chapter 4; Hanks 2019).

One example Hanks gives is that of disjunctive propositions, represented below (Hanks 2015, 106).Footnote 4

  • \(\vdash _\uparrow \langle (\sim \vdash \langle {\mathbf {Frank}},{\mathrm {HONKS}}\rangle ,\sim \vdash \langle {\mathbf {Sarah}},{\mathrm {BLEATS}}\rangle ),{\mathrm {DISJ}}\rangle \)

The idea is that the context created by disjunction is one where the force of the acts is cancelled. This is represented by ‘\(\sim \)’.

This line of response to my argument is a dead end even if it is the right way for Hanks to respond to the problem that he developed it to solve. Firstly, on at least one immediately obvious way of implementing the idea, shown below, this looks like a way to reinstate the taxonomic content–force distinction that Hanks rejects.

  • \(?\sim \vdash \langle {\mathbf {Frank}},{\mathrm {HONKS}}\rangle \)

This is because the interrogative proposition embeds an act of predication. This is not the same as saying that there is a core neutral proposition in common with all three types; the embedded proposition here is not neutral, so this is not the constitutive force–content distinction. But, the act of questioning now involves an act directed at a (predicative) proposition which is also the sort of proposition involved in a corresponding act of assertion, so this is the taxonomic force–content distinction.

A second worry, internal to Hanks’ theory, is that this response would commit Hanks to saying that every question or order creates a cancellation context. That is because, necessarily, questions and orders do not carry commitment. There would be no ‘pure’ acts of questioning or ordering which do not create such contexts. However, Hanks wants to take the ‘pure’ cases of the actions as foundations for our understanding of the cancelled cases; it is hard to see how this could work if pure examples not only do not but could not occur (Hanks 2019, 1386–1388).

A second, much more interesting, response is available. I have argued that interrogative propositions require that there is something that is (i) sufficient to unify propositions, and (ii) neutral in the sense at issue between Hanks and Soames. This shows that Soames’ view is not incoherent, but it does not show that it is right.

Hanks will presumably maintain that the right theory is one on which three different acts unify the three different sorts of proposition and, given what he says about predication, only two of these acts are neutral. The question this raises is why someone might prefer this view, with three sorts of act, to one where there is just one (neutral) sort of act. This would be a version of the taxonomic content–force distinction.

Another possibility is to reject, along with Hanks and Soames, the taxonomic content–force distinction but accept, against Hanks and along with Soames, the constitutive content–force distinction. On this sort of view, at least for predicative propositions, the act by which they are unified can be factored out as a distinct act from their affirmation or endorsement. Soames describes his view about this in Soames (2015, 18–19) and Soames (2019, 1370–1371).

Why might one prefer one of these views to the others? One clear motivation would be that one had a prior commitment to a view about whether acts such as asserting or questioning are simple or complex, i.e., whether asserting is a matter of unifying the constituents of a proposition and also doing something else, or whether there is just one action. If they are simple, then Hanks’ view is better. If they are complex then Soames’ view is better, and the overall view might be more like the traditional view than Hanks proposes. And, if Hanks’ view is better then there is no need for a neutral act of predication because the simple, non-neutral act will be what provides unity to the proposition.

All options are live, because there is no obvious way to decide whether the relevant acts are simple or complex. Both sides of the debate accept that there are complex actions. Introspection does not seem like a reliable guide in these cases.

In conclusion, I reject Hanks’ argument against Soames’ idea that predication is neutral because the same argument can be deployed against Hanks’ account of the unity of interrogative propositions. An exactly parallel argument could be made regarding imperative propositions. I do not think that this settles the debate between Hanks and Soames because the question that my argument brings into focus about the simplicity or complexity of acts of expressing propositions has not been resolved.