Reassessing truth-evaluability in the Minimalism-Contextualism debate

The debate between Semantic Minimalism and Radical Contextualism is standardly characterized as concerning truth-evaluability—specifically, whether or not sentences require rich contextualization in order to express complete, truth-evaluable contents. In this paper, I examine the notion of truth-evaluability, considering which kinds of mappings it might require from worldly states of affairs to truth-values. At one end of the spectrum, an exhaustive notion would require truth-evaluable contents to map all possible states of affairs to truth-values. At the other end, a liberal notion would require only that truth-evaluable contents map at least one possible state of affairs to at least one truth-value. I show that both Minimalists and Radical Contextualists rely on some intermediate, moderately strict notion of truth-evaluability, falling between these two poles. I consider four ways in which such a notion could be defined. However, I argue that each of these is ultimately implausible, giving us no reason to favour a moderately strict notion of truth-evaluability over the liberal alternative. This suggests that the debate must shift to more moderate ground; rather than concerning the in principle possibility of truth-evaluable contents, it fundamentally hinges on their explanatory value. More generally, paying close attention to the notion of truth-evaluability allows us to tease apart distinct strands in the Minimalism-Contextualism debate, and gain a better appreciation of what is at stake.


Introduction
The notion of truth-evaluability features centrally in the debate between Semantic Minimalism and Radical Contextualism. Minimalists argue that well-formed sentences in language express complete, truth-evaluable contents, with only a very limited role for the contexts in which those sentences are uttered. 1 Thus, Borg (2012, p. 3) writes: According to minimal semantics, natural language sentences mean things, the things they mean are in some sense complete (that is to say, they are propositional, truth-evaluable contents), and these literal meanings are determined entirely as a function of the lexical elements a sentence contains together with its syntactic form. Cappelen and Lepore (2005) defend a similar view, arguing that sentences can express complete, truth-evaluable propositions, without pervasive contextual influences. 2 Radical Contextualists, on the other hand, deny that we can sensibly talk about sentences expressing complete, truth-evaluable contents. They argue instead that contexts must always play a substantial role, such that only richly contextualized utterances of sentences can express complete, truth-evaluable contents. According to Travis (2008, p. 70): The new view might be called the speaking-sensitive view of words, or lan-guage…[Words'] fixed, language-contributed semantics must, in general, be supplemented if they are to be properly assessable as to truth; that is, if they are to count either as true or as false. Nor is such supplementation provided automatically by some fixed stock of 'indexical features '. 3 This in-principle claim, that minimal contents are necessarily incomplete, distinguishes Radical Contextualism from its more moderate neighbour. Moderate Contextualists accept that there can be minimal truth-evaluable contents but claim that these have no theoretical or practical importance in their own right. Instead, the contents of primary importance are held to be those which the speaker intuitively communicates; and these will often be highly context-dependent. Recanati (2010) advocates such a view, under the label of 'Truth-Conditional Pragmatics'. He writes: There is no doubt that one can define 'what is said' in such a way that only weak pragmatic effects can affect what is said. If we do so, Minimalism comes out true by definition…What the truth-conditional pragmatist means by 'what is said' corresponds to the intuitive truth-conditional content of the utterance. 1 Specifically, that role is exhausted by providing values for a limited set of genuinely context-sensitive elements, as will be discussed further in Sect. 3. 2 I will focus mainly on Borg's version of Minimalism. However, in Sect. 3, I will mention some ways in which it differs from that of Cappelen and Lepore. 3 I will focus on the version of Radical Contextualism put forward by Travis (2006Travis ( , 2008 although I note that Carston (2002) makes a similar claim.
1 3 Synthese (2021Synthese ( ) 198:2765Synthese ( -2782 According to [Truth-Conditional Pragmatics], the intuitive truth-conditions of an utterance are affected by free pragmatic processes. (Recanati 2010, p. 12, emphasis in original) 4 I will argue in Sect. 6 that a closer analysis of the notion of truth-evaluability makes it difficult to maintain a distinction between Radical Contextualism's in-principle claim and Moderate Contextualism's practical relevance claim. For now, though, let us return to the dispute between Minimalism and Radical Contextualism. Both sides have explicitly positioned truth-evaluability at the heart of their disagreement. As Borg (2012, p. 46) puts it 5 : Where minimalism differs from occasionalism then is over the question of whether [content recoverable without rich appeal to the context of utterance] can itself be thought of as genuine, truth-evaluable content. Travis…is adamant that it can't… It is this commitment to the idea that any content recovered simply on the basis of lexico-syntactic features alone (without any rich appeal to the context of utterance) is destined to fall short of propositional, truth-evaluable content that makes the occasionalist a true opponent of the minimalist.
Responding to Cappelen and Lepore, Travis conceptualizes the debate similarly: [Cappelen and Lepore] assign truth precisely that role in semantics which [Radical Contextualism] denies it. They insist that (unproblematic) declarative sentences-'Pigs grunt', 'There are French women in Chicago', 'The oboe is double-reeded'-and even more problematic ones-'Giraffes are tall', 'Sid has had enough'-express, as such, 'minimal propositions'. Whatever else a minimal proposition is, it is (for them) something truth-evaluable, and, as a rule, truth-valued. That, then, is the main issue between [Semantic Minimalism] and [Radical Contextualism]. (Travis 2008, pp. 151-152, emphasis in original) In this paper, I do not seek to defend either a Minimalist or Contextualist stance. Instead, I will argue that it is a mistake to characterize the debate between Minimalism and Radical Contextualism in terms of truth-evaluability. I will show that, once this is acknowledged, the disagreement turns out to concern the usefulness of minimal truth-evaluable contents, rather than their existence per se.
The structure of the paper is as follows: in the next section, I introduce a spectrum of possible notions of truth-evaluability. In Sects. 3 and 4, I show that both Minimalists and Radical Contextualists rely on moderately strict notions of truthevaluability, falling somewhere between the liberal and exhaustive extremes. In Sect. 5, I consider-and reject-four possible candidates for a moderately strict notion of truth-evaluability. I conclude, in Sect. 6, by arguing that the debate should 1 3 not be framed in terms of the truth-evaluability of contents but what various truthevaluable contents do for us.
2 What truth-evaluability requires I start from the assumption that what makes a content truth-evaluable is, in some sense, its ability to map worldly states of affairs to truth-values. 6,7 A naïve view would require truth-evaluable contents to map all possible states of affairs to truth-values. Take, for example, the following sentence: (1) Serena Williams is American According to the naïve view's exhaustive criterion, for sentence (1) to express a truth-evaluable content, states of affairs must either be ones in which Serena is American (and which are therefore mappable to TRUE) or ones in which that is not the case (mappable to FALSE). Crucially, no states of affairs may remain unmapped.
The naïve view, however, is typically considered to be implausibly stringent by philosophers on both sides of the debate. Borg (2012, p. 108) writes: On this approach then a proposition p must suffice to determine, for every possible world, whether p is true or not in that world. As it stands, this condition is too strong (for instance, there are issues with worlds where objects fail to exist, worlds where fundamental laws are very different to those of this world, and perhaps with vagueness). Travis (2008, p. 134) similarly points out: "Our language is not rigid. It does not foresee its own application in, or to, every circumstance in, or to, which it may prove to need to be applied". Building on this, he writes: Take an arbitrary collection of statements, for each of which there is a way which is the way it said things to be. Is it guaranteed that every member of the collection is either true or false? Does that follow merely from those statements being the ones they are? Or from each having the feature that there is such a thing as how it said things to be? My answer is no. (Travis 2008, pp. 142-143) It seems clear, then, that neither Minimalists nor Radical Contextualists wish to deploy an exhaustive notion of truth-evaluability. Instead, they are thinking of truthevaluable contents as achieving only some partial mapping of possible states of affairs to truth-values. The task is to say what this partial mapping should be.
One possibility lies at the opposite end of the spectrum. A liberal notion of truthevaluability would count any content as truth-evaluable so long as it mapped at least one possible state of affairs to at least one truth-value. Returning to (1), there need only be some possible state of affairs in which Serena is American, or one in which that is not the case. 8 In Sects. 3 and 4, I will show that neither Minimalists nor Radical Contextualists can be thinking of truth-evaluability in this liberal way. Instead, both of their accounts depend on some stricter notion. Although there is relatively little discussion in the literature of how to define such a notion, in Sect. 5 I will consider four alternatives. These require truth-evaluable contents to map, respectively: (i) at least one possible state of affairs to TRUE; (ii) many, or most, possible states of affairs to truth-values; (iii) possible states of affairs to truth-values in an intuitively correct way; (iv) actual states of affairs to truth-values.
Jumping ahead for a moment, I will argue that none of these options is plausible. I will propose instead that we revert to the most liberal way of defining a mapping from states of affairs to truth-values.

Truth-evaluability in minimalism
Minimalists appeal to truth-evaluability at two junctures in their account: first, to demonstrate that some sentences require (very limited) relativization to contexts, in order to express truth-evaluable minimal contents; and, second, to defend against any further contextual influences on such contents.
According to Minimalists, the truth-evaluable, minimal contents expressed by sentences are determined compositionally; roughly-speaking, by the meanings of words (as given by their entries in the lexicon) and the way they are combined (according to grammatical rules).
In typical cases, the contributions that sentence components make to minimal contents are held to be independent of the particular context in which they are uttered. Nonetheless, Minimalists standardly recognize a set of exceptional elements 1 3 in language that are genuinely context-sensitive. These include, at least, indexicals (like 'I' and 'you'), demonstratives (like 'this' and 'that'), a small set of other, open-class expressions (like 'here', 'yesterday', and 'actual') and tense-markers. Such context-sensitive elements can have different extensions in different contexts; for example, the expression 'she' can refer to different (female) individuals, when uttered by different speakers on different occasions.
Minimalists deny that sentences containing these kinds of context-sensitive elements can express truth-evaluable contents independently of context. Thus Borg writes: To claim that pragmatics has no role to play at all within the semantic realm is very likely to lead to the claim that the subject matter of semantics is subpropositional or non-truth evaluable content. The reason for this is pretty easy to see: consider the sentence 'I'm here now'. Unless one is able to look to a specific context in which this sentence is uttered it will be impossible to deliver values for the context-sensitive terms 'I', 'here' and 'now'. Yet without such values any content we can recover for the sentence (perhaps in terms of a Kaplanian character) is bound to fall short of propositional, truth-evaluable content. (Borg 2012, p. x) Take the following sentence 9 : (2) She is American The context-independent, linguistic meaning of sentence (2) places some conditions on its truth. Specifically, it requires that there is at least one female individual who is American; we can express this as follows 10 :

Context-Independent Content (CIC): 'She is American
Minimalists consider CIC to fall short of being a truth-evaluable content because it does not specify which individual-the actual referent of 'she' in the context of utterance of (2)-must have the properties of being female and American.
Relatedly, CIC is not a traditional truth-condition: it captures only necessary but not sufficient conditions for the truth of sentence (2) (the direction of the conditional travels only from left to right). This means that CIC is incapable of mapping any states of affairs to TRUE. Instead, for any state of affairs that includes at least one female American (thus satisfying the necessary condition expressed by CIC), the truth-value of the content remains indeterminate (neither definitely true 9 In this sentence, 'she' is to be understood deictically rather than anaphorically. This ensures that its referent cannot be determined purely linguistically, within the discourse context, but only by appeal to features of the extra-linguistic context. 10 According to some accounts (2) may express a truth as long as the speaker's intended referent is American, even if that individual turns out not, in fact, to be female. Thus, the necessary condition in the consequent of CIC reduces to: Ǝx(x is American). If preferred, readers can simply plug in this alternative condition, and substitute 'American' for 'female American' throughout the ensuing discussion. nor definitely false). These considerations are taken as evidence against contents like CIC being truth-evaluable. 11 Note, though, that CIC is capable of mapping some states of affairs to FALSEspecifically, those in which there are no female Americans. The context-independent, linguistic constraints provided by sentence (2) are sufficient on their own to rule out such states of affairs. Recall too that the liberal notion of truth-evaluability described in Sect. 2 above requires only that at least one possible state of affairs be mappable to at least one truth-value. If we were to adopt this notion, then, CIC would count as a truth-evaluable content (just in virtue of its being 'falsity-evaluable', as it were). Since Minimalists deny that CIC is truth-evaluable, it is clear that they cannot be operating with the liberal notion of truth-evaluability. Instead, they are assuming something stricter. In Sect. 5, I will consider some potential candidates. For now I continue setting out the positive Minimalist view, leading to its second appeal to truth-evaluability.
Minimalists argue that a sentence like (2) can only express a truth-evaluable content once relativized to a context of utterance. Specifically, that context must determine the referents of context-sensitive elements like 'she'. To characterize the truthconditions of minimal contents, Borg (Borg 2004(Borg , pp. 165-166, 2012 adopts Higginbotham's conditionalized t-sentences (Higginbotham and Segal 1994, pp. 92-93). For sentence (2), this gives us something like the following:

Minimal Content (MC):
If the speaker of 'She is American' refers with the utterance of 'she' therein to x and to nothing else, then this sentence, as uttered in this context, is true if and only if x is American.
MC depends on a specific, contextualized utterance of (2), in order to incorporate x as the actual referent of 'she'. 12 This is what supposedly allows MC to provide sufficient as well as necessary conditions on truth. Since MC is a standard biconditional truth-condition, it maps states of affairs to TRUE if x is American; and to FALSE if x is not American. In this way, Minimalists standardly acknowledge some role for context in determining minimal contents.
A crucial qualification is that the role for context is restricted to providing values only for genuinely context-sensitive elements of the sentence. There is some disagreement between Minimalists as to how these genuinely context-sensitive elements should be defined. Cappelen and Lepore (2005) primarily seek to limit the quantity of context-sensitive elements in language (more or less to those enumerated by Kaplan (1989)). Borg (2004Borg ( , 2012 instead prioritizes a qualitative constraint, 11 Note that CIC differs from two other kinds of contents in the neighbourhood. First, it differs from a content expressible by the sentence, "There are some female Americans", for which the condition '∃x(x is female ∧ x is American)' would be not only necessary but also sufficient. That content would map any state of affairs containing at least one female American to TRUE. CIC also differs from Perry's 'reflexive content' (Perry 2012). The reflexive content of (2) would typically determine a value for x; the important point is that it would do so via a relation between x and the utterance of (2), not via the direct reference of 'she' to x. 12 Specifically, 'she' is supposed to acquire the value, x, via direct reference. This means that MC2 is not a reflexive content in Perry's sense (Borg 2012, p. 136, footnote 27). requiring context-sensitivity to be traceable back to lexico-syntactic features of the sentence. The key idea here is that any context-sensitivity must already be built into the lexical entry or grammatical rule. 13 , 14 Setting this wrinkle aside, Minimalists typically agree that a minimal content like MC is truth-evaluable as it stands, without further contextual enrichment. The claim that there are complete, truth-evaluable, minimal contents of this kind constitutes Minimalism's second appeal to truth-evaluability. As will be discussed in the next section, it is also the focus of the Radical Contextualist critique.
Before moving on, though, it should be noted that Minimalists may well have other reasons for accommodating limited contextual influences on minimal content, aside from considerations of truth-evaluability. In particular, they point out that certain context-sensitive elements of language are referring expressions, and rigidly designate their referents (see, for example (Borg 2012, pp. 134-142)). Whereas a content like MC captures this feature of the expression 'she', CIC does not; in that sense, CIC might be deemed 'incomplete'.
It would take us too far afield to evaluate the substance of claims about which expressions are rigid designators. Instead, I want only to show that such claims are orthogonal to the current discussion. The question we are concerned with here is whether or not a content like CIC is truth-evaluable, not whether it fails to capture certain facts about the meaning of the expression 'she'. Minimalists have claimed that one problem with a content like CIC is that it falls short of truth-evaluability; it is this claim I wish to interrogate. I allow that a different incompleteness challenge may be mounted against CIC, on the basis that it fails to respect certain meaning facts. Indeed, this is an approach I discuss further in Sects. 5 and 6. However, once the Minimalist takes this path, any appeal to truth-evaluability, as such, simply falls away. 15 13 In this way, it is maintained that minimal contents flow directly from a sentence's formal, compositional, semantic meaning. As we will see in Sect. 5, though, they need not correspond to communicated contents, which may also depend on pragmatic factors. As Kratzer (2012, p. 4

) puts it:
Words, phrases, and sentences acquire content when we utter them on particular occasions. What that content is may differ from one context to the next. It is the task of semantics to describe all those features of the meaning of a linguistic expression that stay invariable in whatever context the expression may be used. This invariable element is the meaning proper of an expression. 14 Borg (2004Borg ( , 2012 additionally argues that reasoning about speakers' subjective states should not enter into the process of deriving minimal content. Cappelen and Lepore do not impose this constraint (Cappelen and Lepore 2005, pp. 147-149). 15 In Sect. 5 I will also reject attempts to incorporate such meaning facts within the very definition of truth-evaluability itself.

Truth-evaluability in Radical Contextualism
Radical Contextualists claim that minimal contents, and not just the likes of CIC, inevitably fall short of truth-evaluability. Extending the Minimalist's argument against CIC, it is argued that the context-independent, linguistic meanings of sentences always underspecify truth-evaluable contents. 16 Travis writes: The driving force of [Radical Contextualism] is this idea: the open sentences of language speak of ways for things to be which admit of understandings… This blocks truth-conditional semantics. For suppose I say, 'The sentence "Sid grunts" is true iff Sid grunts'. Either I use that last 'grunts' on some particular understanding of being a grunter-one understanding among many-or I do not. If I do, then I assign the sentence a property it does not have. For it does not speak of being a grunter on any special understanding of this. But if I do not, then I fail to state any condition under which anything might be true. Being a grunter on no particular understanding of being one is just not a way for Sid to be. In brief, the choices here are falsehood or failure to say anything. What would be needed to block this result are ways for things to be, which one might speak of, and which do not admit of understandings. (Travis 2008, p. 159, emphasis in original) Similar reasoning can be applied to the expression 'American' in our sentence (2). Imagine, for example, a tennis player who is an American citizen but has nonetheless entered a tournament as part of a team that is unaffiliated to America. In a context in which the player's citizenship is at stake, an utterance of (2) might be considered true. However, in a context in which her team affiliation is at stake, an utterance of the same sentence might be considered false. According to the Radical Contextualist, there are no available understandings of expressions like 'grunt' and 'American', in the absence of specific contexts of utterance. 17,18 From the Radical Contextualist's perspective, then, the minimal contents posited by Minimalists fall short of truth-evaluability, since they fail to recognize the full 17 In response, Minimalists typically reject this claim outright. They argue, for example, that such expressions pick out properties common to all individuals of whom they may be literally and truthfully predicated. See, for example, (Cappelen and Lepore 2005, chapter 11) and (Borg 2012). For opposing views, see (Chomsky 2000), (Pietroski 2005(Pietroski , 2018, MacFarlane 2007). 18 Since the issue here concerns the satisfaction of predicates like 'grunts' and 'is American', it would be possible to reframe the discussion in terms of the satisfaction-conditions of such predicates, rather than the truth-conditions of sentences that contain them (I am grateful to an anonymous reviewer for raising this point). In order to address the Minimalist and Contextualist arguments head-on, I have opted to retain the standard framing in terms of truth-evaluability. However, I do not believe anything of substance hangs on this decision; the same considerations that favour a liberal notion of truth-evaluability (enumerated in Sect. 5) would also favour a liberal notion of satisfiability, such that a predicate would count as satisfiable so long as at least one thing were ruled into, or out of, its extension. extent of contextual effects. With respect to sentence (2), the minimal content can be thought of as placing only a necessary condition on the truth of an utterance of (2). Specifically, it requires that the referent of 'she' is American on some understanding of 'American'-an understanding which could potentially be made available in some context or other. The truth-condition must therefore be pared back to the following conditional: Minimal Content 2 (MC2): If the speaker of 'She is American' refers with the utterance of 'she' therein to x and to nothing else, then this sentence, as uttered in this context, is true only if there is some understanding of 'American' on which x is American.
Whereas the Minimalist's MC assumes that there is some context-independent understanding of 'American', the Radical Contextualist denies that any such understanding is available. In the absence of an appropriate, contextually-determined understanding of 'American', MC2 fails to provide a value for that expression. As before, this leaves us with a content that lacks a sufficiency condition for its truth. It is incapable of determining whether x is American, and therefore cannot map any states of affairs to TRUE. Instead, states of affairs in which x is American on some possible understanding remain unmapped to either truth-value. This is taken to show that MC2 falls short of truth-evaluability.
Nevertheless, again, MC2 is capable of mapping some states of affairs to FALSE-specifically, those in which x is not American, on any possible understanding. This seems compatible with Travis's account; in the long passage quoted below, he argues that linguistic meaning places important constraints on expressible content: I think that the English 'is round' speaks, as such, of being round. So I think it speaks of that on every use of it which is (a case of) speaking proper English. I am not dissuaded from this view by the fact that, on different such speakings of it, it will make different contributions to the truth-conditions of wholes of which it may, then, be part. There is a fact about that bit of English which I hope I can capture in that way. It is that the meanings of those words constrain, in a particular way, what you can say on an occasion in using them-even if they do not narrow things down to just one thing. I can, occasion permitting, call the squash ball round and speak truth of it even as it begins its rebound off the wall. But I cannot so easily call it round and speak truth of it just in case it barks, or is made of lead, or is on fire. My idea is: I can use 'is round', on an occasion, of a ball, to say what I would say on that occasion in calling it round; and (ceteris paribus, perhaps) that is all I can use those words to say of it. Further, to say what I just did is to say how the meanings of those words constrain their use. I cannot use those words of the ball (speaking English, so that they mean what they do) and thereby say the ball to be on fire, unless there is an understanding of being round on which to be round is to be on fire.
(Which, so far as I can see, there is not.) I do think this is a pretty stringent constraint on what you can use 'is round' to say in speaking English. (Travis 2008, pp. 175-176, emphasis in original) I interpret the linguistic constraints described by Travis as ruling out some possible states of affairs. 19 As we have seen, this is all that is needed for contents to count as truth-evaluable on a liberal notion of truth-evaluability. Radical Contextualists, then (like Minimalists) must be operating with some suitably stricter notion of truthevaluability, given their denial that contents like MC2 are truth-evaluable.

Stricter notions of truth-evaluability
In the previous two sections, I demonstrated that neither Minimalists nor Radical Contextualists can be working with a liberal notion of truth-evaluability. Instead, both camps are relying on something stricter. There has been little explicit discussion of what that stricter notion is. In this section I will consider four possibilities that require truth-evaluable contents to map, respectively: (i) at least one possible state of affairs to TRUE; (ii) many, or most, possible states of affairs to truth-values; (iii) possible states of affairs to truth-values in an intuitively correct way; (iv) actual states of affairs to truth-values.
I will argue that none of these works. Of course, given the form of the argument, this does not prove that no moderately strict notion of truth-evaluability is available, since more promising candidates might yet be identified. However, by eliminating some initially plausible possibilities, I aim to shift the burden of proof to those who rely on such a notion, to provide a workable definition.
Starting with (i), we saw that contents like CIC and MC2 were incapable of mapping any possible states of affairs to TRUE (even though they map some to FALSE). The first suggestion, then, would require truth-evaluable contents to map at least one possible state of affairs to TRUE, not just to FALSE. In other words, this proposal stipulates that the truth-condition must be a standard biconditional, preventing mere 'falsity-evaluability' from entailing truth-evaluability.
The problem with this proposal is that it would automatically deny truth-evaluability to necessary falsehoods. By definition, these cannot map any states of affairs to TRUE (no state of affairs could correspond to the right-hand side of a biconditional truth-condition). Take, for example, a sentence of the following form: (3) x is F and not-F. 19 Likewise, if we chose to think about things in terms of predicate satisfaction, a predicate's invariant satisfaction-condition would act as a constraint, or filter, preventing at least some things from being included within its extension on any occasion of use (and therefore ensuring that at least some possible applications of the predicate would express falsehoods). This is despite the fact that the satisfactioncondition might not define any very specific extension that the predicate may be used to pick out on a particular occasion of utterance.
Let's assume that the lexical entry for 'F' has no context-sensitivity built into it; and that both occurrences of 'F' in (3) obtain identical understandings in the context of utterance. In such a scenario (3) will express a contradiction, mapping no states of affairs to TRUE. Nevertheless, I take it that both Minimalists and Radical Contextualists would consider it to express something truth-evaluable (just false). 20 It might be objected, however, that necessary falsehoods are relevantly different from the contents expressed by sentences like (1) and (2). An obvious difference is that necessary falsehoods map all possible states of affairs to FALSE, leaving no possible worlds unmapped. It might be suspected, then, that the problem with the liberal notion of truth-evaluability is that it leaves various possible worlds unmapped to either truth-value.
As discussed in Sect. 2, neither Minimalists nor Radical Contextualists endorse an exhaustive notion of truth-evaluability, so the problem can't be a failure to map all possible states of affairs to truth-values. Perhaps, though, contents like CIC and MC2 map too few states of affairs to truth-values. This brings us to proposal (ii), which would require truth-evaluable contents to map many, or most, possible states of affairs to truth-values. However, any such criterion is at best vague and at worst unworkable. It is vague as long as the requirement to map 'many' states of affairs to truth-values fails to specify how many. Any attempt to impose such a threshold would seem straightforwardly arbitrary. A deeper problem is that, assuming there are infinitely many possible states of affairs, it may not even make sense to talk about some proportion of them being mappable to truth-values.
Another way of interpreting the worry is that the contents expressible with sentences like (1) and (2) (unlike necessary falsehoods) should, intuitively, map certain kinds of states of affairs to TRUE-specifically, those in which the individual being referred to is, in fact, American (whether 'American' is thought of as expressing a property context-independently, or only on a context-specific understanding). This brings us to suggestion (iii). The idea is that truth-evaluable contents must map states of affairs to truth-values in an intuitively correct way, i.e. in a way that reflects what a speaker could reasonably mean by using those words.
I take it that CIC and MC2 would not count as truth-evaluable on this basis. A speaker uttering (2) is reasonably taken to refer to some particular (female) individual by using 'she'; and to predicate some property of that individual by saying that she is 'American'. Thus, intuitively, the content of (2) should map to TRUE those 20 Note that I am not requiring Radical Contextualists to sign up to the idea that sentence (3) expresses something necessarily false independently of context. I have explicitly allowed for the possibility that it does so only following rich contextualization, as long as both occurrences of 'F' obtain identical understandings in the context. It is important here to distinguish between the context of utterance, which determines the content of (3), and the context of evaluation, which determines whether or not (3) is satisfied.
(3) may express a necessary falsehood only in some contexts of utterance, but wherever it does so, it will be false in all contexts of evaluation. This point comes up again below, in relation to option (iv). states of affairs in which the individual in question has the relevant property; and it should map to FALSE those in which she doesn't. 21,22 Nevertheless, intuitions about appropriate mappings to truth-values would need to be of the right kind for this to get going. For the Minimalist, this would mean showing that they concern the minimal content of the sentence, rather than some contextually richer speech act content that gets communicated by the sentence in use. In general, Minimalists are careful to point out that minimal contents need not (and typically won't) answer to all of our intuitions about what is communicated by an utterance. As Borg (2012, p. 48) puts it: These minimal contents provide the literal meaning of sentences relativized to contexts of utterance, though, the minimalist acknowledges, they do not usually provide the intuitive contents of speech acts involving those sentences. 23 To see the point, take the following sentence, as uttered by a tennis spectator to her companion: (4) Williams bounced the ball three times and served an ace.
What the spectator intuitively communicates is that Williams bounced the ball three times and then served an ace. However, the minimal content of the sentence is consistent with Williams having performed those actions in the reverse order, i.e. serving an ace and then bouncing the ball three times. Minimalists thus place a wedge between a sentence's minimal content (based on compositional semantics) and the wider (pragmatic) speech act contents of utterances.
At most, then, minimal contents aim to capture only some subset of intuitions about what speakers could reasonably mean, namely those which reflect the constraints imposed by context-independent, linguistic meanings. These would need to be carefully delineated before the Minimalist could appeal to what speakers of (2) intuitively use 'she' to mean. Specifically, it would need to be shown that the use of 'she' to refer to a particular female individual relates to the expression's lexical entry rather than to use-specific facts. 24 21 I note that the question of how widely these intuitions are shared is ultimately an empirical matter. 22 This may be one way of cashing out a suggestion we find in Borg, that one might consider a content truth-evaluable "just in case it makes a ruling in a range of clear-cut possible scenarios" (Borg 2004, p. 238). Note that Borg does not endorse this suggestion; she argues only that if it were adopted, minimal contents would count as truth-evaluable. 23 Similar points are made by Cappelen and Lepore (2005) and García-Carpintero (2013). It is worth noting that each of these philosophers is careful to emphasize the very important role played by speech act contents in a theory of linguistic communication. Their claim is therefore not that we can, or should, disregard such contents. Rather, it is that they should be held apart from minimal contents, which play a distinct theoretical role. 24 Radical Contextualists may face a similar problem if they wish to distinguish the content directly expressed by the utterance from other contents that are merely indirectly expressed, or implicated. Specifically, it would need to be shown that the use of 'American' to pick out a context-specific property relates to what the speaker of (2) directly expresses, rather than indirectly implicates. There is some evidence that Travis sees value in maintaining such a distinction; see, for example, his discussion in 'On What is Strictly Speaking True' (Travis 2008, chapter 1). Assuming that the relevant intuitions can be pinned to a sufficiently restricted content, still they seem to arise from meaning facts that are quite distinct from the requirements of truth-evaluability. For example, it is because 'she' is taken to be a certain kind of referring expression that the content of (2) is expected to concern the referent of 'she' (and to be true if and only if that particular individual is American). Rather than stipulating that truth-evaluable contents must respect such facts, it seems more parsimonious to appeal to the facts directly when arguing for contextual relativization. As discussed in Sect. 3, then, the Minimalist might argue that contents like CIC are incomplete, on the basis that the lexical entries for expressions like 'she' demand contextually-determined referents. Yet this is entirely consistent with CIC being truth-evaluable. The argument that CIC fails to capture our intuitions about the invariant meaning of the expression 'she' is independent of that content's status as truth-evaluable or not. In sum, nothing is gained by building independently motivated intuitions into the notion of truth-evaluability. 25 Finally, option (iv) would require any truth-evaluable content to map actual states of affairs to truth-values. Again, this would preclude contents like CIC and MC2 being truth-evaluable. For example, given that the actual world does contain female Americans, CIC fails to map any actual states of affairs to truth-values (not even to FALSE).
This condition is also implausible, however. At the same time as denying the truth-evaluability of the sentence 'She is American', it would deem the structurally similar sentence 'She is demoniac' to be perfectly truth-evaluable, in virtue of its being mappable to FALSE in the actual world (assuming there are in fact no demons!). This seems to carve up the territory in the wrong way, for both the Minimalist and the Radical Contextualist. 26,27 To sum up, I currently see no reason to favour a moderately strict notion of truthevaluability over the liberal alternative. The liberal notion has the advantage of being modest without being vacuous; it successfully captures the idea that truthevaluability concerns mappings of states of affairs to truth-values, without imposing additional, unjustified, constraints on the nature of those mappings. Pending the identification of more promising candidates for a moderately strict notion, I conclude that we should take contents like CIC and MC2 to be truth-evaluable, in line with the liberal notion.

Conclusions and implications
Where does this leave us? First, acknowledging that contents like CIC are truthevaluable undermines the Minimalist's claim that sentences like (2) require relativization to contexts, in order to be truth-evaluable. (As we will see below, however, those sentences may still be argued to require contextual relativization on separate grounds.) Likewise, acknowledging that contents like CIC and MC2 are truth-evaluable neutralizes the Radical Contextualist's claim that sentences can only ever express truth-evaluable contents once richly contextualized. (Again, though, as will be discussed below, it may still be denied that such contents have any explanatory value.) A third point, which falls out from the first two, is that the debate between Minimalism and Radical Contextualism cannot fundamentally concern truth-evaluability. I have suggested that both camps should agree that minimal contents are truth-evaluable. As I will argue below, I think the debate between them ultimately concerns the practical significance of minimal contents.
In fact, Minimalists and Radical Contextualists shouldn't only agree that minimal contents are truth-evaluable but also that contents like CIC are truth-evaluable. In other words, there is a level of truth-evaluable content that is even more minimal than Minimalism's minimal content. I will call this 'marginal content'. 28 Although I think Minimalists should recognize that marginal contents like CIC are truth-evaluable, I note that this is consistent with their maintaining that minimal contents like MC retain an important role in semantic theorizing. Such contents, recall, capture just those contextual effects which are anticipated by an expression's lexical entry, or a sentence's combinatory syntax. Thus, I believe the key task facing the Minimalist is to demonstrate the explanatory value of minimal contents, which hold apart these 'weak' or 'mandatory' contextual effects from the 'free' contextual effects that influence wider speech act contents. Minimalists have indeed sought to meet this challenge; see, for example, (Borg 2004(Borg , 2012(Borg , 2017 and (Cappelen and Lepore 2005). Although I do not attempt to assess their arguments here, they seem to me to be the cornerstone of any defence of Minimalism.
On the other side, I have argued that Radical Contextualists should also recognize both minimal and marginal contents as truth-evaluable. In other words, they 28 The proposal here is in some ways similar to that of García-Carpintero (1998), which identifies minimal semantic content with context-independent, linguistic meaning. However, García-Carpintero assumes that this content is not truth-evaluable. Others who restrict 'semantics' to an entirely contextindependent level of meaning similarly take it to fall short of truth-evaluability (in some or all cases)see, for example (Carston 2002) and (Harris in press). My purpose here has been to show that semantic constraints of the kind envisaged are, in fact, sufficient for truth-evaluability. Of course, though, we can (and should) posit additional levels of meaning in order to capture other aspects of what is communicated by uttered sentences.
should drop the in-principle claim that sentences cannot express truth-evaluable contents independently of context. That means Radical Contextualism must take up a more moderate position, objecting to minimal contents purely on the basis that that they lack practical relevance, since they generally fail to capture what speakers communicate.
Indeed, this line of argument is commonly pursued by Contextualists of all stripes. As we saw in Sect. 1, Moderate Contextualists acknowledge that sentences express truth-evaluable, minimal contents, while arguing that these contents have no significance, in their own right. Thus, whereas appeals to truth-evaluability have standardly purported to distinguish Radical Contextualism from Moderate Contextualism, they in fact end up obscuring a key point of agreement between the two views.
A key task facing both Moderate and Radical Contextualists, then, is to argue (against the Minimalist) that mandatory contextual effects have no privileged status in a theory of meaning but, instead, the contents of theoretical interest are those which can be freely influenced by the context. For some arguments to this effect, see (Carston 2002) and (Recanati 2004(Recanati , 2010. Again, I cannot assess these here but I suggest that they are central to the dispute between Minimalists and Contextualists, whether of a Radical or Moderate bent. It is only possible to recognize this point once we dispel the spectre of truth-evaluability. The upshot, I suggest, is that we should focus on what the contents posited by competing theories do for us, not whether they are truth-evaluable. 29 Before closing, it is worth pausing on my claim that Radical Contextualism collapses into a more moderate position. This may seem to contradict arguments put forward by Cappelen and Lepore (2005) and Borg (2012) to the effect that it is Moderate Contextualism that is doomed to collapse into Radical Contextualism. 30 I believe that the disagreement here is, in large part, only apparent. In this paper, I have focused narrowly on the different parties' claims about truth-evaluability; it is with respect to these claims that I have argued Radical Contextualism must move towards a more moderate stance. In contrast, Cappelen and Lepore (2005, chapters 3-5) focus on Contextualists' claims about the number of context-sensitive expressions in language. They argue that, once the Moderate Contextualist begins to 29 I have not been able to say much here about why we might want to appeal to marginal contents. To do so would require much further discussion. However, taking inspiration from the arguments put forward by Minimalists in their defence of minimal contents, I suggest that marginal contents could potentially perform important logical, psychological, and socio-linguistic functions. For example, such contents seem to support certain logical properties and relations: two marginal contents will be logically equivalent to one another if they map the same worlds to TRUE and FALSE and leave the same worlds unmapped; meanwhile, sets of contents will be logically consistent if their mappings to truth values are mutually compatible (and logically inconsistent otherwise). One might also explore the potential psychological role of marginal contents as the product of a distinct set of deductive mental processes. And such contents may prove to have socio-linguistic functions, as those we fall back on, or hold speakers liable for, in certain communicative situations. These points, of course, would require a considerable amount of further argument, which I cannot provide here. I note them merely to suggest that marginal contents may turn out to do something for us too, in addition to being merely conceptually possible. 30 I am indebted to an anonymous reviewer for helping me see this.
expand the set of context-sensitive expressions, there is no stable place to stop, until all linguistic expressions are included. Thus, taking the first step on a Contextualist path inevitably sends one down the slippery slope to a far more radical position. Whether or not they are right about this, it remains entirely consistent with what I have said; sliding towards the more extreme position on their scale, concerning the pervasiveness of context-sensitivity, does not imply taking a radical stance on the question of whether minimal contents can, in principle, be truth-evaluable. As I have attempted to show, a sentence can still express something truth-evaluable, despite its containing context-sensitive elements, just so long as its invariant meaning filters out some possible states of affairs.
Meanwhile, Borg (2012, pp. 35-38) considers Contextulists' claims about 'sharpening' or 'precisifying' contents, in order to capture what a speaker communicated in context. Whereas Moderate Contextualists standardly take such sharpenings to be possible, Radical Contextualists have argued (and Borg agrees) that the process is, in principle, never-ending; the only way to stop making increasingly fine-grained distinctions is by acknowledging that context plays an ineliminable role in contentdetermination. She thus identifies a different way in which Moderate Contextualism seems inexorably to slide toward a more radical position.
Again, though, this would be consistent with my argument; even if one is forced to accept that only richly contextually-embedded contents capture what a speaker communicated, this would not imply that such contextualization is required for truthevaluability. Thus, for example, according to Travis's occasion-sensitive account, it is impossible to sharpen communicated contents in a way that would allow them to be represented, either in language or in thought (this distinguishes his approach from many other forms of Contextualism). However, this dimension of his view is quite independent of the claim that only contextually-embedded communicated contents can be truth-evaluable. Again, it is the latter claim that has been my target here.
The current treatment has sought to shine a spotlight on the appeals to truth-evaluability that are standardly made by Minimalists and Contextualists. In so doing, it arrives at a subtler appreciation of the debate: on some dimensions, such as those discussed by Borg, Cappelen and Lepore, Moderate Contextualist positions arguably collapse into more radical ones; however, when it comes to claims about truthevaluability, it is the Radical Contextualist's in-principle claim that gives way to a more moderate claim concerning explanatory value. In sum, by paying close attention to the notion of truth-evaluability, we are able to tease apart distinct strands in the Minimalism-Contextualism debate, and see how they pull in different directions.