Williamsonian modal epistemology, possibility-based

Abstract

Williamsonian modal epistemology (WME) is characterized by two commitments: realism about modality, and anti-exceptionalism about our modal knowledge. Williamson’s own counterfactual-based modal epistemology is the best known implementation of WME, but not the only option that is available. I sketch and defend an alternative implementation which takes our knowledge of metaphysical modality to arise, not from knowledge of counterfactuals, but from our knowledge of ordinary possibility statements of the form ‘x can F’. I defend this view against a criticism indicated in Williamson’s own work, and argue that it is better connected to the semantics of modal language.

Keywords:

• Modality
• modal epistemology
• modal semantics
• counterfactuals

Notes

1 Anti-exceptionalism about philosophical knowledge is, of course, one of the main tenets of Williamson 2007; Williamson uses the term ‘exceptionalism’ to describe his opponents, see e.g. Williamson (2007, 3).

2 I adapt these labels from Hale (2003). Note, however, that Hale only distinguishes between possibility-based and necessity-based modal epistemologies. Williamson’s modal epistemology would be necessity-based by Hale’s standards (if it is asymmetric at all) because the natural, and probably the epistemically primary, equivalences in Williamson’s picture link necessity to counterfactuals. Moreover, I would like to stress that the possibility-based epistemology that Hale (2003) discusses is quite different from the one that I will sketch in Section 3, precisely because my own version is broadly Williamsonian.

3 In addition, as Krödel (2012) points out, our credence in a necessity claim will normally be higher than that in the corresponding counterfactual on the right-hand side of an instance of (V) or (V*), so our knowledge of the former can hardly be explained in terms of knowledge of the latter. I agree; but as Krödel also points out, this objection does not apply to (Q).

4 Here is why. If counterpossibles are sometimes false, and if A is a necessary truth, then and should be prime examples for false counterpossibles: if there is anything that the impossibility does not counterfactually imply, it’s its own contradiction (A) or indeed any contradiction (). So both (V) and (V*) require that all counterpossibles be true. And so does (Q); otherwise replacing p with on its right-hand side would yield a counterexample to the universal quantification despite the necessary truth of A.

5 Some linguists claim that we can use tense to express a modal meaning; cf. Iatridou (2000) on ‘fake tense’; see also Section 6.

6 Strohminger speaks about ‘abilities’ where I speak of can statements; Nanay is concerned with ‘action properties’, a notion that is close to Gibsonian affordances. Both switch between explicit can statements and dispositional idioms such as ‘edible’, ‘reachable’, ‘breakable’, etc. I take such idioms to be equivalent, probably even synonymous, with corresponding can statements (see Vetter 2014), so I am happy to include them as well. See Nanay (2012, 431f) for some differentiations between his own view, which I have here labelled ‘Gibsonian’, and Gibson’s affordance theory of perceptual content.

7 Williamson (forthcoming-b) can be read as claiming that imagination is a route to knowledge about can statements – but there he appears to think of can statements along the lines of a version of the conditional analysis of ability ascriptions, see Williamson (forthcoming-b, 6, 9). I treat can statements as a kind of restricted possibility statement.

8 Pace Spencer forthcoming, who argues against that entailment. My response to his arguments would take us too far afield for present purposes.

9 Of course, we may conclude that dinosaurs do not exist if we are presentists or temporaryists. But the relevant claim then would be that it is always that case that what exists is what presently exists (in some sense of that frustrating term), so we must still go beyond the temporal restriction of ordinary existence claims.

10 I am assuming here that we can quantify unrestrictedly (see Williamson 2003). If not, the claims in the text could be reformulated accordingly.

11 Am I using counterfactuals in extending beyond ordinary can statements here? If so, do I not require the same resources as Williamson’s epistemology, in addition to those of my own? – I am not entirely convinced that the envisaged extension must use counterfactuals. But note that even if it does, it uses them only to get from one objective possibility to another objective possibility. Counterpossibles, or the Williamsonian equivalences, do not feature in the picture at all; and it is the modality expressed by ‘can’, not that expressed by the counterfactual conditional, which settles the nature of the modality at issue.

12 Deontic modals are harder to classify, but I will focus on the distinction between circumstantial and epistemic modals, since I agree with Williamson that there is less danger of confusing our target type of modality, circumstantial modality, with deontic modality; see Williamson ms, 2.

13 Thanks to Tim Williamson for prompting this clarification and several others.

14 Such as: ‘These cities share one feature: Napoleon might have visited them’.

15 ‘Can’ has hardly any epistemic uses in its positive, present-tensed form. But its negation may be used epistemically, as in ‘This cannot be true!’; and of course, ‘could’ has an established epistemic use. Other languages, such as German, have equivalent expressions which, while predominantly circumstantial, are more easily recruited for epistemic uses.

16 Polysemy is, in short, non-accidental ambiguity – the kind of ambiguity exhibited by ‘since’ (temporal or causal relation) but not by ‘bank’. Emanuel Viebahn and I provide an extended argument for the polysemy of modals in Viebahn and Vetter forthcoming.

17 See also Edgington 2007, 2011.

18 A more detailed argument can be found in Vetter (forthcoming).

19 I have elsewhere argued that a dispositionalist account for the truth-conditions of counterfactuals can be defended; see Vetter (forthcoming). This is compatible with saying that we do not syntactically or semantically construe counterfactuals in this way.

20 The markers may be somewhat surprising: Coates found that only epistemic, and no circumstantial, modals had (i) inanimate sentence subjects, (ii) existential sentence subjects, (iii) stative verbs, (iv) progressive aspect, and (v) quasi-modals in their scope. As a corpus linguist, Coates makes no claim that these markers hold for all possible expressions of epistemic and circumstantial modality respectively, and philosophers may find it straightforward to produce counterexamples to some of them (e.g. ‘the machine can crush oranges’ as against (i)).

21 Add ‘have to’ after ‘would’ to make the reading even clearer; I claim that the addition makes no difference to the meaning of (17). Contrast a context where it is A’s character that is being discussed: A is extremely cautious and thorough. To illustrate, someone utters (17). That context, I submit, would elicit a circumstantial reading.

22 There are very tricky issues about contextualism vs. relativism here, which I am trying to avoid by imagining the context of utterance and the context of evaluation to coincide. I think that this is fair in the present context because we are not so much interested in agreement and disagreement, but rather in one speaker/thinker coming to a conclusion about whether or not she takes a certain counterfactual to be true.

23 I am not sure that Williamson would disagree with any of these remarks. The role that he sketches for the imagination in Williamson (forthcoming-b), for instance, while tailored to knowledge of objective possibilities, seems to work just as well for epistemic possibilities. Williamson (forthcoming-b, 10) uses an indicative conditional – which is standardly thought to be epistemic – in tandem with a counterfactual one – which is, of course, standardly thought to be circumstantial – as two analogous applications of knowing by imagining, noting only that there are ‘subtle cognitive differences in the cognitive process’.

24 I have done so, with only a slightly less conservative point of view but without commitment to the Barcan formulas, in Vetter (2015, ch.7.5).

25 For very helpful comments and discussion, I would like to thank Simona Aimar, Max Bohnet, Mathias Böhm, Luca Castaldo, Steffen Koch, Thomas Krödel, Christian Nimtz, Sonia Roca Royes, Margot Strohminger, Kilu von Prince, and Jonas Werner as well as the audience of a workshop on modality and causation at UCL on 19 December 2015 and my research colloquium at Humboldt-University, Berlin. Special thanks to Mathias Böhm for a very helpful overview of the linguistic literature on conditionals, and of course to Tim Williamson for his extremely helpful comments on an earlier version of this paper.