Skip to main content

Strong Epistemic Possibility and Evidentiality


The paper distinguishes between weak and strong epistemic possibility and argues that the notion of strong epistemic possibility is the key to solving some of the most vexing puzzles about the semantics of epistemic modality.

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


  1. 1.

    Ornstein, N. (2014). The Atlantic. Ornstein's article is tellingly titled, "The U.S. Needs a Constitutional Right To Vote".

  2. 2.

    One worry some authors have raised about basic disagreement dialogues like this is that it is unclear whether Sherlock's denial ("That's false") targets the embedded proposition or the modal claim. While it is true that standard presentations oversimplify the disagreement data (as we will later on see), this particular worry is unfounded. As MacFarlane (2011) observes, we might clarify the target of disagreement simply by asking:

    Do you mean that it's false that Moriarty is in London or false that he might be in London?

    Like many other authors, including MacFarlane, I judge that both claims are false: it is false that Moriarty is in London, and false (modulo Yalcin and Knobe's findings, discussed below) that he might be in London.

  3. 3.

    Another view on the market is von Fintel and Gillies' (2011) cloudy contextualism, where utterances of bare epistemic modal sentences like (5) are sui generis speech acts that ‘put in play’ a cloud of propositions. A speaker is warranted in asserting "might φ" just in case the speaker is warranted in asserting at least one of the propositions that her utterance puts in play, while an assessor is warranted in accepting/rejecting the speaker's claim if the assessor is warranted in accepting/rejecting the strongest proposition that the speaker ‘reasonably has an opinion about’ (329). Suppose that Watson's utterance of (4) ("Moriarty might be in Beijing") puts in play (a) the proposition that it is compatible with what Watson knows that Moriarty is in Beijing, and (b) the proposition that it is compatible with what a group including Watson and Sherlock knows that Moriarty is in Beijing. The cloudy contextualist will then say that Watson is warranted in asserting (4) on the (true) solipsistic or speaker-oriented interpretation, while Sherlock is warranted in rejecting (4) on the (false) group interpretation.

    The are several problems with cloudy contextualism. One is that since an assessor is only warranted in rejecting "the strongest proposition that the speaker reasonably has an opinion about", cloudy contextualism sheds no light on cases like Hacking's Salvage Ship: the mate of the ship does not reasonably have an opinion about what we know (as historically-removed observers)—yet disagreement judgments remain robust. Another is that it is unclear why charity does not lead interpreters to prefer the true, speaker-oriented interpretations of speakers' modal claims when the group interpretation is false.

  4. 4.

    Yalcin and Knobe on eavesdropping cases: "Paradigmatically, [eavesdropping cases] are cases wherein some agent A utters a sentence of the form might p, and another agent B, better informed in relevant respects than A and eavesdropping on A from outside A's discourse context, vocalizes a truth value judgment about what A has said—typically to the effect that what A said is false" (Yalcin and Knobe 2010, 2).

  5. 5.

    …though see Andrew Moon, "Knowing Without Evidence" (forthcoming). Moon argues against the evidence thesis: S knows that φ at t only if S believes that φ on the basis of evidence at t. Significantly, Moon's counterexamples would amount to cases where the indirectness signal on must (more on which, below) would make it impossible for S to felicitiously assert, at t, "It must be that φ''.

  6. 6.

    Note that the claim here is not that might-claims evolve from explicit reasoning from (implicit or explicit) premises. That is not the only (or even the main) kind of inferential reasoning. The evidence that φ can be indirect in virtue of supporting φ through implicit and perhaps unconscious reasoning processes that take place against tacit, background or default assumptions.

  7. 7.

    While it is an interesting question whether this is the right way to think of strong deontic possibility, it is not a question that I will take up here.

  8. 8.

    As Hacking (1967, 148) observes, disagreement and retraction data can make it seem that,

    Whenever it turns out to be false that p, we say, of an earlier era, that in those times it may have seemed possible that p, but it was not really possible at all. (1967, 148)


    Whenever it turns out to be false that p, we say that while there was apparent evidence that p – evidence that seemed to support p – this ‘evidence’ did not really support p (was not really evidence that p) at all.

    However, both these conjectures are too strong, as Hacking's fair lottery case (discussed below) reveals. The latter observation does, however, push us toward a more objective conception of the kind of evidence that satisfies the positive evidence requirement.

  9. 9.

    The mate lacks (undefeated) evidence not because his claim to evidence is undermined, by his miscalculations, but because of the opposing evidence that the wreck is further south. Undermining evidence implies only that the mate's calculations are not evidence that the wreck is in the bay—not that there is no evidence that the wreck is in the bay (that (10a) is false). Without opposing evidence, the problem with the mate's assertion would be that it is unwarranted—not that it is false.

  10. 10.

    Dorr and Hawthorne's proposal has certain affinities to a widely-held view in the linguistics literature that epistemic modals are ambiguous between subjective and objective readings (Lyons 1977). Kratzer (1981) claims that the difference between these readings derives from different conversational backgrounds. The subjective reading calls for a ‘subjective stereotypical ordering source’ while the objective reading calls for an ‘objective stereotypical ordering source’. While a subjective stereotypical ordering source can include “superstitions… [and other propositions that] couldn't be defended on objective grounds”, an objective stereotypical ordering source will consist of “established facts" (57–58).

    Interestingly, much of the data used to support subjective and objective readings better supports our own weak and strong possibility readings. Consider the following Hungarian data from Kiefer (1984)

    1. (a)

      Petor irnat leveiet.

      ‘Peter may be writing a letter’

    2. (b)

      Peter a szobaban doigozhat.

      ‘Peter may be writing a letter’

    According to Kiefer, (a) conveys that the prejacent proposition “is compatible with what the speaker knows”:

    [T]he state-of-affairs described by the [prejacent] proposition Peter is writing a letter is possible, it is not excluded by the given epistemic background. But other possibilities, too, exist. Peter may very well do something else… [T]he speaker does not have any special reasons to believe that the proposition… is more likely to be true than any other proposition compatible with the epistemic background… (62)

    The interpretation of (b) “is quite different…”:

    The meaning can be paraphrased in the following manner. The speaker has certain reasons to believe that the most likely thing which Peter may be doing just now is letter writing… (63)

    At the level of data, Kiefer's distinction between (a) and (b) corresponds to our distinction between weak and strong epistemic possibility statements. Indeed, Kiefer describes (a) and (b) in precisely these terms. He claims that (a) is a 'weak' epistemic possibility statement and that (b) is a 'strong' epistemic possibility statement. There, however, the similarities end; following Kratzer and Lyons, Kiefer analyzes weak readings as subjective readings and strong readings as objective readings. This is an incongruous view of the data. At any rate, it fits poorly with Kiefer's informal characterizations of weak and strong readings. The salient difference between (a) and (b) is not a difference in the relevant kind of evidence, but a difference in the strength of the epistemic relation between the prejacent and the epistemic background. According to Kiefer, (b) is true just in case the proposition Peter is writing a letter is supported by the speaker's evidence (“The speaker has certain reasons to believe that the most likely thing which Peter may be doing just now is letter writing”), while (a) is true just in case the same proposition is compatible with the speaker's evidence (“that Peter writing a letter is compatible with what the speaker knows”). Kiefer does not mention any subjective body of evidence on which the truth of (a) might depend; indeed, he explicitly says that the speaker has no evidence (that does not derive from the absence of contrary evidence) about Peter's letter-writing:

    [T]he speaker does not have any special reasons to believe that the proposition… is more likely to be true than any other proposition compatible with the epistemic background. (Kiefer 1984, 62)

    (Though see Strevens 2008 for the view that the absence of evidence for a proposition p is subjective as opposed to objective evidence that not-p.).

  11. 11.

    It is the agent-independent, objective character of the evidential requirement that reconciles modal production with modal uptake. At the level of production, we need to explain why speakers are warranted in making epistemic possibility claims despite their limited epistemic access to evidence other than their own. The important fact about the present view is that both weak and strong epistemic possibility statements concern the speaker's own evidence; the speaker is not taking a stand about the evidence available to others (e.g., to the wider epistemic community). What reconciles the speaker-oriented nature of the modal claim with uptake is that speakers are not absolutely authoritative about what evidence they have. The mate, for example, can be mistaken about whether he has evidence for that the wreck is in the bay, and the expert about whether she has evidence that Tony is dead.

    There is more that needs to be said (and more that will be said, in Chapter Two) about the notion of objective evidence informally characterized here by reference to pretheoretical judgments about evidential and epistemic modal claims. There is also reason to be especially interested in the epistemological project of spelling this notion of evidence out. Although epistemologists often talk of justification and warrant, these are not the most common ordinary language terms of epistemic appraisal. In everyday circumstances, speakers talk of what might, may or must be the case. Given this, evidence relations tied to epistemic modality ought to be of special epistemological interest.

  12. 12.

    Another problem with Dorr and Hawthorne's account concerns the way that the ‘salient’ non-epistemic facts are determined. If we focus on the empirical facts that are salient to the speaker, we get mistaken predictions: we want to predict that the fixed lottery player speaks falsely, even though he hasn't considered the possibility that the lottery is rigged. However, if we focus on the empirical facts salient in the assessor's context, Dorr and Hawthorne's proposal collapses into a version of relativism (not Dorr and Hawthorne's intention). Relativism also makes mistaken predictions—but in the fair lottery case. The relativist predicts that (12) (above, main text) is false in the mouth of the fair lottery player (since we, the assessors, know that he did not win).

    One relativist, John MacFarlane (2011), argues that Hacking is simply wrong about the data. According to MacFarlane, Hacking's original sentence (a) below does not have a true epistemic reading; when the subjunctive conditional is replaced with the indicative (which typically gives rise to an epistemic reading), the sentence is false. This seems to me to be wishful thinking: to my ear, (a)–(d) have true epistemic possibility readings. In any case, MacFarlane's reply does not engage with the retrospective judgment the player spoke truly in asserting (12).

    1. (a)

      It was possible that he could have won.

    2. (b)

      It was perfectly possible that he had the winning ticket.

    3. (c)

      It was possible for the ticket to have won.

    4. (d)

      He might have won.

  13. 13.

    Given that the lottery is fair, buying a ticket increases the player's odds of winning. In this sense, it is evidence that the player will win. If the lottery is large, then it is not a lot of evidence—but it is some.


  1. Brayshaw TC (1965) Native poplars of southern Alberta and their hybrids. Publication 1109. Canada Department of Forests, Ottawa

  2. Coleman T (2004) The Nelson touch: the life and legend of Horatio Nelson. Oxford University Press, Oxford

    Google Scholar 

  3. Cossins A (2013) The legacy of the Makin case 120 years on. Syd Law Rev 35:731

    Google Scholar 

  4. Court of Appeals of North Carolina (2002) STATE of North Carolina v. Harry REED, Jr. No COA01-1371. 15 Oct 2002

  5. DeRose K (1991) Epistemic possibilities. Philos Rev 100(4):581–605

    Article  Google Scholar 

  6. Dorr C, Hawthorne J (2013) Embedding epistemic modals. Mind 122(488):867–913

    Article  Google Scholar 

  7. Hacking, I (1967) Possibility. Philos Rev 76:143–168

    Article  Google Scholar 

  8. Harper K, Shane J, Jones J (1985) Aspen ecology and management in the Western United States. US DOA General Technical Report RM-119

  9. Kiefer F (1984) Focus and modality. In: Abraham W, de Mey S (eds) Papers from the 6th Groninger grammar talks on ‘topic, focus and configurationality’, vols 1, 24, pp 55–81. 6–8 Apr 1984. GAGL

  10. Kratzer A (1981) The notional category of modality. In: Eikmeyer HJ, Rieser H (eds) Words, worlds, and contexts. New approaches in word semantics. de Gruyter, Berlin, pp 38–74

    Google Scholar 

  11. Kratzer A (1991) Modality. In: von Stechow A, Wunderlich D (eds) Semantics: an international handbook of contemporary research. de Gruyter, Berlin, pp 639–650

    Google Scholar 

  12. Lasersohn P (2005) Context dependence, disagreement, and predicates of personal taste. Linguist Philos 28:643–686

    Article  Google Scholar 

  13. Lasersohn P (2008) Quantification and perspective in relativist semantics. Philos Perspect 22(1):305–337

    Article  Google Scholar 

  14. Lyons J (1977) Semantics, 2 vols. Cambridge University Press, Cambridge

    Book  Google Scholar 

  15. MacFarlane J (2011) Epistemic modals are assessment-sensitive. In: Egan A, Weatherson B (eds) Epistemic modality. Oxford University Press, Oxford, pp 144–178

    Chapter  Google Scholar 

  16. Mackey R (2014) War crimes experts call Syrian defector’s photographs ‘Direct Evidence’ of Assad’s ‘Killing Machine’. New York Times Blog, 20 Jan 2014

  17. Ornstein N (2014) The U.S. Needs a Constitutional Right to Vote. THEATLANTIC.COM.

  18. Richard M (2004) Contextualism and relativism. Philos Stud 119(1-2):215–242

    Article  Google Scholar 

  19. Ronald WG, Lenz LM, Cumming WA (1973) Biosystematics of the genus Populus LI Distribution and morphology of native Manitoba species and variants. Canad J Bot 51(12):2431–2442

    Article  Google Scholar 

  20. Stephenson T (2007) A Parallel account of epistemic modals and predicates of personal taste. Proceedings of Sinn und Bedeutung 11:583–597

  21. Switek B (2012) Did dinosaurs swim? SMITHSONIAN.COM.

  22. von Fintel K, Gillies S (2008) CIA leaks. Philos Rev 117(1):77–98

    Article  Google Scholar 

  23. von Fintel K, Gillies A (2011) Might made right. In: Egan A, Weatherson B (eds) Epistemic modality. Oxford University Press, Oxford, pp 108–130

    Chapter  Google Scholar 

  24. von Fintel K, Iatridou S (2008) How to say ought in foreign: the composition of weak necessity modals. In: Guéron J, Lecarme J (eds) Time and modality. Springer, Berlin, pp. 115–141

    Chapter  Google Scholar 

  25. von Hacquard V (2009) Modality. In: von Heusinger K, Maienborn C, Portner P (eds) Semantics, vol 2. de Gruyte, Berlin, pp 1484–1514

    Google Scholar 

  26. von Wright GH (1963) Norm and action. A logical inquiry. Routledge and Kegan Paul, London

    Google Scholar 

  27. Westmoreland RR (1998) Information and intonation in natural language modality. PhD thesis, Indiana University.

  28. Yalcin S (2011) Nonfactualism about epistemic modality. In: Egan A, Weatherson B (eds) Epistemic modality. Oxford University Press, Oxford, pp 295–332

    Chapter  Google Scholar 

  29. Yalcin S, Knobe J (2010) Fat Tony might be dead: an experimental note on epistemic modals. Unpublished manuscript.

Download references

Author information



Corresponding author

Correspondence to Katrina Przyjemski.

Additional information

Katrina Przyjemski tragically died just before she was about to enter the job market in the Fall of 2014. She had been writing a thesis on epistemic modals under my supervision at New York University; and the present paper is her writing sample. The editors and I thought it well worthy of publication and it bears testimony to the power and promise of this brilliant young philosopher—Kit Fine.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Przyjemski, K. Strong Epistemic Possibility and Evidentiality. Topoi 36, 183–195 (2017).

Download citation


  • Modality
  • Evidence
  • Disagreement
  • Epistemic modality
  • Epistemic possibility
  • Epistemic necessity