Skip to main content

The Speaker Authority Problem for Context-Sensitivity (Or: You Can’t Always Mean What You Want)


Context-sensitivity raises a metasemantic question: what determines the value of a context-sensitive expression in context? Taking gradable adjectives as a case study, this paper argues against various forms of intentionalist metasemantics, i.e. that speaker intentions determine values for context-sensitive expressions in context, including the coordination account recently defended by King (Philos Perspect 27:288–311, 2013; Noûs 48(2):219–237, 2014a; in: Burgess, Sherman (eds) Metasemantics: New essays on the foundations of meaning, Oxford University Press, Oxford, pp 97–118, 2014b). The paper argues that all intentionalist accounts face the speaker authority problem, that speaker intentions are just the wrong sorts of things to determine the standards for gradable adjectives in context. The problem comes to light when we look at cases in which speakers have idiosyncratic, false beliefs that cause their proper communicative intentions to come apart from the non-intentional features of context like the question under discussion, facts about the world, practical goals, and prior linguistic discourse.

This is a preview of subscription content, access via your institution.


  1. 1.

    I am following Glanzberg (2007, 2009), and King (2013, 2014a, b) in calling this a metasemantic question about context-sensitivity. Semantic questions are descriptive; they describe what the values of linguistic expressions are. By contrast, metasemantic questions are foundational; they ask by what means a linguistic expression comes to have that value. Articulating the metasemantic question in a theory neutral way is difficult. King contrasts the cases he is interested in (i.e. everything aside form the automatic indexicals) as ones in which the standing meaning of an expression needs supplementation in context to secure a value. To understand the contrast he has in mind consider ‘I’ versus ‘that’. ‘I’, relative to a context, refers to the speaker of that context; all we need to get this is the character (i.e. standing meaning) of ‘I’ and a context. But ‘that’ doesn’t just get a value relative to a context by virtue of its standing meaning; the context needs to be supplemented with something like an intention or demonstration. King’s metasemantic question is regarding what exactly the context needs to be supplemented with for these types of non-automatic context-sensitive expressions to get a value in context. But this way of phrasing things depends on a particular view of contexts. If contexts include the intended (or demonstrated) objects, then a demonstrative does get a value based solely on its standing meaning and a context. The metasemantic question is then best phrased as a question about what determines the (relevant aspects of) the formal context. How one thinks of the question depends on their view of contexts and the relationship between context and content, and I don’t think we have to make a decision on this matter to pursue the question at hand. I will make no distinction between the various views in what follows.

  2. 2.

    As noted in fn.1, how exactly to articulate the question and the answer to it depends on a full view of the relationship between situation of utterance, formal context, and content. But we don’t need to take a stand for present purposes, so I will use the term ‘context’ neutrally, not meaning to invoke a particular formal notion or exclude features of the actual situation of utterance.

  3. 3.

    See Rett (2015), chapters 2 and 3, for an excellent overview.

  4. 4.

    While thinking of context-sensitivity in this way makes the question posed particularly clear, what follows is also relevant to other views of context-sensitivity. More generally, we can think of the meta-semantic question as: what determines context-sensitive content in context?

  5. 5.

    This sort of maneuver is employed by those who take the comparative form of the adjective to be basic, and therefore the semantics of ‘tall’ is not by itself appropriate in the positive form.

  6. 6.

    There is also a related set of views, defended by, e.g., Ludlow (1989) and Stanley (2002a) that propose a variable at logical form for a comparison class which doesn’t determine a degree (the semantics for gradable adjectives does not involve degrees on these views). For our metasemantic purposes, we can class these kinds of views as being of the third type, since the sort of context-sensitive variable that needs to be determined is a property variable for a comparison class, and the metasemantics is not affected by whether that in turn determines a degree standard or not.

  7. 7.

    This does not exhaust all possible views, but it does cover the most popular options.

  8. 8.

    In fact, it is almost ubiquitously assumed, but it is difficult to demonstrate this with examples. Notable exceptions to this are Gauker (1997, 2010), which argue for an objective notion of context, and Glanzberg (2007, 2009), which argue for an indirect metasemantics.

  9. 9.

    King puts these criteria in terms of objects, since he takes referential demonstratives as his paradigmatic case. Talk in terms of objects does not seem appropriate for the case at hand, or when talking about context-sensitive expressions more generally. He also uses the term supplementive as a theory-neutral term for context-sensitive expressions that do not automatically get a value given their character, rather than talking in particular about unpronounced variables.

  10. 10.

    See for example Stanley and Szabó (2000), Stanley (2002b, 2005b).

  11. 11.

    By ‘core indexical’, Stanley means what I referred to above as automatic indexicals, like ‘I’ or ‘now’.

  12. 12.

    Note that throughout, Stanley is assuming a comparison class kind of context-sensitivity for gradable adjectives, so what speaker’s intentions determine is a comparison class (a property).

  13. 13.

    This is the issue taken up in Glanzberg (2007), in which he argues that gradable adjectives have an indirect metasemantics in response to Richard’s arguments. The way that Glanzberg in fact applies the notions of direct and indirect metasemantics line up fairly well with my intentionalist/non-intentionalist distinction. That is, his examples of direct metasemantics involve intentions directly determining the value of the context-sensitive element, and the indirect one is a balance of non-intentionalist features plus speaker intentions. In principle, the distinctions cross-cut each other, since we could have a direct metasemantics in which one non-intentionalist feature of context directly determines the value of a context-sensitive element, and an indirect metasemantics in which intentions play a large or necessary role (though by definition of an indirect metasemantics, they won’t be sufficient for determining content).

  14. 14.

    DeRose (2008) is also interested in the analogy between knowledge ascriptions and gradable adjectives, but in dispelling a myth that the context-sensitive element of gradable adjectives is always and only a comparison class (i.e. “tall for an F”, where F is some property) that can be easily articulated in context.

  15. 15.

    For example, Schiffer (1977, 1981, 2016) and Neale (2005, 2007, 2016).

  16. 16.

    It is worth quoting Neale (2016) in full on this point, to get the perspective of thorough-going intentionalist who accepts something like the definition of speaker intentions that I gave above. Neale comments on Kaplan’s Carnap/Agnew case, in which Kaplan intends to be demonstrating a picture of Rudolph Carnap when saying “That is a picture of one of the greatest philosophers of the twentieth century”. Unbeknownst to him, the picture normally on his wall has been replaced by one of Spiro Agnew, and so when he points to the wall behind him, he indicates a picture of Agnew with his gesture. Neale writes:

    [Kaplan’s] communicative intention concerns o [the picture of Carnap], so the answer to the question is clear: Kaplan is referring to o. The fact that his addressee might (though need not) think he is referring to the picture of Agnew, call it \(o^{\prime }\), is completely irrelevant. As is the fact that a reasonable, attentive interpreter might (but need not) take him to be referring to \(o^{\prime }\). (Sometimes you refer to o and everyone else thinks you are referring to \(o^{\prime }\). Life’s tough. But not so tough that we can’t have an intentional account of what speakers refer to with demonstratives.) (p. 283)

    To see that this is quite different from how King is thinking about the case, one need only look to King’s extensive discussion of the case in King (2013).

  17. 17.

    King (2014b, p. 116). King uses the term “demonstrative” here, but it is explicitly in a context in which he means it to apply to all context-sensitive expressions, including gradable adjectives.

  18. 18.

    King also has other, independent reasons for preferring this semantics.

  19. 19.

    As Neale argues, this is similar to the meaning-intention problem that Schiffer (1992, 2005) raises for speakers’ intentions regarding modes of presentation in things like belief ascriptions.

  20. 20.

    This is something on which King (2013, 2014b) goes into considerable detail.

  21. 21.

    Of course, we then encounter the problem of borderline cases at the border of the range; perhaps a probability distribution would work better.

  22. 22.

    I use the notation \(\textit{adjective}_{\textit{degree}}\) as a shorthand for that degree being the threshold for the property in question. It should be noted that gradable adjectives like ‘tall’ are unidimensional, in that they are only ever measured along a single dimension (i.e. height, in this case). Others are multidimensional, like ‘large’, ‘happy’, or ‘rich’, in that they can be measured across multiple scales, depending on the context, e.g. ‘large’ when predicated of a city can mean large in population or large in area, or some combination of both. Like many others who write on the topic, for the purposes of this paper I’m assuming that the multidimensionality issues have been resolved before dealing with the context-sensitivity introduced by the need for a standard of comparison. It is beyond the scope of the present paper to consider what the resolution of the multidimensionality of gradable adjectives shows (if anything) about the metasemantics of gradable adjectives in context. For more on the multidimensionality of gradable adjectives see Kennedy (2013) and McNally and Stojanovic (2014).

  23. 23.

    The notion of question under discussion that is at play here has to be something more objective than the most well-known conception of a question under discussion, due to Craige Roberts [see Roberts (2004)]. On her view, the question under discussion acts as a partition on the common ground, but since in these types of cases the common ground is defective, i.e. the interlocutors are unbeknownst to each other employing different notions of tallness, it’s not clear how to implement this view. In any case, the defectiveness of the common ground has little to do with the problem with applying this conception of a question under discussion for the present purposes, as we’ll see when we turn to the next cases. The point is that what seems to matter for the determination of the content of the unpronounced variable is an objective notion of question or topic under discussion, i.e. how it relates to the relevant aspects of the world rather than the common ground. This is the notion of question under discussion I wish to employ. I will return to this in Sect. 4.

  24. 24.

    This might count as a case of Gertrude having conflicting intentions. For someone who doesn’t have false beliefs about the heights of average adults, having in mind the property of being an adult versus being one of the people one normally sees will determine similar properties and degree thresholds. However, since these come apart for Gertrude, depending on how we describe her intentions, we get very different results. On some versions of King’s theory, having conflicting intentions results in there being no semantic value. But see King (2013) for a more detailed discussion of how to resolve conflicting intentions, at least for the case of demonstratives.

  25. 25.

    I give further arguments against the compatibility of an intentionalist metasemantics and the comparison class semantics when discussing Case 4 below.

  26. 26.

    For another example, consider someone saying “New York City is best for rich people”, but where the speaker thinks that making $40,000 a year with no other assets counts as rich.

  27. 27.

    Note that King can’t simply give up the common ground criterion—this would make discharging one’s responsibility for successful communication much too difficult, since speakers would not be able to rely on any of the mutual presumptions between speaker and hearer. Spelling everything out as though there are no mutual presumptions between speaker and hearer is simply not feasible, nor is it an accurate description of what actually goes on in conversation.

  28. 28.

    This is clearly a toy case to elicit the clearest intuition possible. But it is entirely plausible that it be common ground between interlocutors that one holds a particular standard for a gradable adjective that the other one doesn’t share. People regularly argue about things like standards for richness or what counts as the bedroom being cold at night, without coming to an agreement. Someone can know perfectly well what their husband means when he complains that the bedroom is cold without agreeing on the standard for coldness. (In fact, this is a threshold that a couple could know quite precisely—one partner thinks that the thermostat being set at 68\(^{\circ }\) Fahrenheit or below is the threshold for cold while the other thinks it is 72 \(^{\circ }\)F.)

  29. 29.

    Some gradable adjectives might be more suitable for very local, personal standards (in certain contexts) than others. For example, what counts as a cold or hot room temperature is likely a very local matter, as in, decided by the person or people in the room and no one else. But this might be due to the fact that what counts as a cold or hot room is simply what feels cold or hot to its inhabitants, and not because it is a matter of speaker’s intentions. Note that what counts as a cold or hot oven, for example, does not depend on the feelings or communicative intentions of the person or people in the room.

  30. 30.

    The question of what content is expressed relative to the context is independent of the question of what the speaker meant and what the interpreter understood. Neale (2016) (pp. 273–4, 282–3) claims that no philosophical puzzle arises when what a speaker means and what a hearer interprets doesn’t match. I agree with this, but not with his conclusion that appealing to speaker meaning is all we need when it comes to context-sensitivity. It is not a philosophical puzzle, but there is an interesting philosophical question distinct from speaker meaning and hearer interpretation: what was the content of what the speaker said? As the examples illustrated, this question doesn’t just arise when there is speaker meaning/ hearer interpretation mismatch.

  31. 31.

    Thanks to an anonymous reviewer for pointing out this way of explaining this intuition.

  32. 32.

    I have made no argument here that there should be a single metasemantic story, even limiting our focus to context-sensitive expressions. I think we need to look at each kind of expression on a case by case basis.

  33. 33.

    The foregoing non-intentionalist metasemantics also obviates the need to have a pluralistic semantics for gradable adjectives, as argued by Keith DeRose (see the example in Sect. 2 above), since even if someone only knows that Susie is tall for her age without knowing anything about her age or average heights for kids her age can still express that Susie is tall for a 5 year old or that Susie is \(\hbox {tall}_{45^{\prime \prime }}\), depending on the QUD and the practical goals of the conversation.


  1. Bartsch, R., & Vennemann, T. (1972). The grammar of relative adjectives and comparison. Linguistische Berichte, 20, 19–32.

    Google Scholar 

  2. DeRose, K. (2008). Gradable adjectives: A defence of pluralism. Australasian Journal of Philosophy, 86(1), 141–160.

    Google Scholar 

  3. Gauker, C. (1997). Domain of discourse. Mind, 106, 1–32.

    Google Scholar 

  4. Gauker, C. (2010). Contexts in formal semantics. Philosophy Compass, 5, 568–578.

    Google Scholar 

  5. Glanzberg, M. (2007). Context, content, and relativism. Philosophical Studies, 136, 1–29.

    Google Scholar 

  6. Glanzberg, M. (2009). Not all contextual parameters are alike (unpublished manuscript).

  7. Kennedy, C. (2007). Vagueness and grammar: The semantics of relative and absolute gradable adjectives. Journal of Semantics, 30(1), 1–45.

    Google Scholar 

  8. Kennedy, C. (2013). Two sources of subjectivity: Qualitative assessment and dimensional uncertainty. Inquiry, 56, 258–277.

    Google Scholar 

  9. King, J. (2013). Supplementives, the coordination account and conflicting intentions. Philosophical Perspectives, 27, 288–311.

    Google Scholar 

  10. King, J. (2014a). Speaker intentions in context. Noûs, 48(2), 219–237.

    Google Scholar 

  11. King, J. (2014b). The metasemantics of contextual sensitivity. In A. Burgess & B. Sherman (Eds.), Metasemantics: New essays on the foundations of meaning (pp. 97–118). Oxford: Oxford University Press.

    Google Scholar 

  12. Ludlow, P. (1989). Implicit comparison classes. Linguistics and Philosophy, 12(4), 519–533.

    Google Scholar 

  13. McNally, L., & Stojanovic, I. (2014). Aesthetic adjectives. In J. Young (Ed.), Paper draft for the volume semantics of aesthetic judgement.

  14. Neale, S. (2005). Pragmatism and binding. In Z. Szabó (Ed.), Semantics vs pragmatics. Oxford: Oxford University Press.

    Google Scholar 

  15. Neale, S. (2007). On location. In M. O’Rourke & C. Washington (Eds.), Situating semantics: Essays on the philosophy of John Perry (pp. 251–393). Cambridge: MIT Press.

    Google Scholar 

  16. Neale, S. (2016). Silent reference. In G. Ostertag (Ed.), Meaning and other things: Essays in honor of Stephen Schiffer. Oxford: Oxford University Press.

    Google Scholar 

  17. Rett, J. (2015). The semantics of evaluativity Oxford studies in theoretical linguistics. Oxford: Oxford University Press.

    Google Scholar 

  18. Richard, M. (2004). Contextualism and relativism. Philosophical Studies, 119(1), 215–242.

    Google Scholar 

  19. Roberts, C. (2004). Context in dynamic interpretation. In L. Horn & G. Ward (Eds.), Handbook of Pragmatics (pp. 197–220). Oxford: Blackwell.

    Google Scholar 

  20. Schiffer, S. (1977). Naming and knowing. Midwest Studies in Philosophy, 2, 28–41.

    Google Scholar 

  21. Schiffer, S. (1981). Indexicals and the theory of reference. Synthese, 49, 43–100.

    Google Scholar 

  22. Schiffer, S. (1992). Belief ascription. Journal of Philosophy, 89, 499–521.

    Google Scholar 

  23. Schiffer, S. (2005). Russell’s theory of descriptions. Mind, 114, 1135–83.

    Google Scholar 

  24. Schiffer, S. (2016). Gricean semantics and reference. In G. Ostertag (Ed.), Meaning and other things: Essays in honor of Stephen Schiffer (pp. 493–523). Oxford: Oxford University Press.

    Google Scholar 

  25. Stalnaker, R. (1978). Assertion. Syntax and Semantics, 9, 315–332.

    Google Scholar 

  26. Stanley, J. (2000). Context and logical form. Linguistics and Philosophy, 23(4), 391–434.

    Google Scholar 

  27. Stanley, J. (2002a). Making it articulate. Mind & Language, 17(1–2), 149–168.

    Google Scholar 

  28. Stanley, J. (2002b). Nominal restriction. In G. Peter & G. Preyer (Eds.), Logical form and language (pp. 365–390). Oxford: Oxford University Press.

    Google Scholar 

  29. Stanley, J. (2005a). Knowledge and practical interests. Oxford: Oxford University Press.

    Google Scholar 

  30. Stanley, J. (2005b). Semantics in context. In G. Peters & G. Preyer (Eds.), Contextualism (pp. 221–253). Oxford: Oxford University Press.

    Google Scholar 

  31. Stanley, J., & Szabó, Z. (2000). On quantifier domain restriction. Mind & Language, 15, 219–261.

    Google Scholar 

  32. Szabó, Z. (2001). Adjectives in context. In I. Kenesei & R. M. Harnish (Eds.), Perspectives on semantics, pragmatics, and discourse: A festschrift for Ferenc Kiefer (pp. 119–146). Amsterdam: John Benjamins.

    Google Scholar 

Download references


Thanks to Christopher Gauker, Jeff King, and several anonymous referees for helpful feedback on earlier drafts of this paper. Thanks also to audiences at the CUNY Cognitive Science Speaker Series, the Context-Relativity in Semantics conference at the University of Salzburg, The Ninth International and Interdisciplinary Conference on Modeling and Using Context in Larnaca, Cyprus, Jeff King's graduate seminar at Rutgers University, and Meaning & Other Things: A conference celebrating the work of Stephen Schiffer at NYU for helpful discussion of earlier versions of this paper.

Author information



Corresponding author

Correspondence to Karen S. Lewis.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Lewis, K.S. The Speaker Authority Problem for Context-Sensitivity (Or: You Can’t Always Mean What You Want). Erkenn 85, 1527–1555 (2020).

Download citation