Abstract
We consider a naturallanguage sentence that cannot be formally represented in a firstorder language for epistemic twodimensional semantics. We also prove this claim in the “Appendix” section. It turns out, however, that the most natural ways to repair the expressive inadequacy of the firstorder language render moot the original philosophical motivation of formalizing a priori knowability as necessity along the diagonal.
This is a preview of subscription content,
to check access.Similar content being viewed by others
Notes
For more on this, and different characterizations of twodimensionality, see Humberstone (2004).
Perhaps the main point can be generalized for other cases as well. What is important, as we shall make explicit, is a semantics considering necessity and actuality interpreted by the corresponding modal operators, as well as apriority taken as truth on the diagonal.
Where \({\mathcal {M}}\) denotes a Kripke model and \(w*\) is the distinguished element of a set W, this can be done by rewriting \({\mathcal {M}},w\,\vDash\,p\) as \({\mathcal {M}},\langle w*,w\rangle \,\vDash\, p\).
We purposefully avoid single quotation marks in order to not clutter the presentation. The context should make usemention distinctions clear.
The accessibility relations \({\mathcal {R}}_{\Box }\) and \({\mathcal {R}}_{\mathcal {D}}\) are just the ones in Fritz (2013, p. 1758), except that he also adds an accessibility relation for \({\mathcal {A}}\) formulas, which would be useful for us if we were investigating properties of the logic corresponding to how the frames are defined.
Here we follow Holliday and Perry (2014) in their quantified twodimensional semantics. Even though their semantics is defined to deal with the Hintikka–Kripke problem in the context of an epistemic logic, their rigidity condition is useful for our purposes. More on this below.
About notation: Davies and Humberstone (1980, p. 4) write \({\mathcal {M}}\,\vDash\, ^v_w\varphi\), appending the superscripts and subscripts to the right side of the turnstile. We simply prefer having the world variables on the left side, which, in its present form, is intended as a twodimensional version of \({\mathcal {M}},\langle v,w\rangle \,\vDash\, \varphi\).
This is also the case in Kocurek (2016).
Local validity and general validity coincide for basic onedimensional modal logics without actuality operators since there is no use for a distinguished element in the models. Similarly, there is no use for a distinguished point on the diagonal of a 2D frame since the actuality operator is not fixed to any particular point but to the first coordinate of every pair of possible worlds.
Some terminological remarks are in order. It is usual to say ‘realworld’ instead of ‘local’ validity, in accordance with the terminology used in Crossley and Humberstone (1977). We find it odd, however, to call realworld validity truth at the distinguished pair in every model, otherwise we have no reason to prefer a different nomenclature.
See p. 1611.
We have adapted, of course, the semantic clause of \(\mathcal {F}\) for pairs of worlds.
Davies (2004, p. 89) makes it clear that they did not intend to formalize anything like an epistemic logic, although the resulting system does give rise to a priori truths. Moreover, in response to Evans’ criticisms to the fixedly operator, he also considered the possibility of adding a primitive operator \(\mathcal {D}\) for diagonal necessity (cf. p. 92).
Axiom \(\mathcal {F}\)6, for example, of \({\mathbf{S5}}\mathcal {AF}\), reads \(\Box \varphi \leftrightarrow \mathcal {FA}\varphi\) for \(\mathcal {A}\)free \(\varphi\). See Davies and Humberstone (1980, p. 4).
See, for example, Chalmers and Rabern (2014, p. 212), although in the same paper he recognizes difficulties for the semantics of the apriority operator caused by the nesting problem.
Proof: Suppose \({\mathcal {M}}\,\vDash\, x=y\) for an assignment V. By the unrestricted semantics of identity terms, \(V(x)=V(y)\). Let \(\langle w,w\rangle\) be any pair of possible worlds in W such that \(\langle w*,w*\rangle \mathcal {R}_{\mathcal {D}}\langle w,w\rangle\). Then, given \(V(x)=V(y)\), it follows that \({\mathcal {M}}^w_w\,\vDash\, x=y\), whence \({\mathcal {M}}\,\vDash\, \mathcal {D}x=y\), therefore \({\mathcal {M}}\,\vDash\, x=y\supset \mathcal {D}x=y\). An analogous result is available for Davis and Humberstone’s \({\mathbf{S5}}\mathcal {AF}\), the only difference being that it involves the compound operator \(\mathcal {FA}\) rather than \(\mathcal {D}\).
We thereby assimilate proper names to individual constants in \(\mathcal {L}\). This follows closely the presentation in Holliday and Perry (2014, §4.5).
Cf. Chalmers (2004, pp. 158–162; 2014, p. 212).
The same strategy was employed in tense logics by Kamp (1971) for the “now” operator.
In fact, in \({\mathbf{S5}}\mathcal {A}\) there is no need at all of evaluating formulas with respect to a pair of worlds.
This is similar to how the sentence appears in Lampert (manuscript). The only difference being that (3c) was formulated in terms of deep possibility, the dual of deep necessity. Admittedly, not much of a naturallanguage statement, since the notion of deep necessity seems to be even more philosophically loaded in comparison to apriority. Yet, in a language intended to formalize it, this is exactly the kind of sentence expected to be expressible.
Notice that a rendering similar to (4) is available for Davies and Humberstone’s \({\mathbf{S5}}\mathcal {AF}\), the only difference being that \(\mathcal {FA}\) occupies the place of \(\mathcal {D}\).
Hodes (1984, p. 25, Theorem 15) proved that a sentence resembling (1), but restricted to a single predicate letter, is not representable in S5, although similar inexpressibility results were previously conjectured by Hazen (1976). Wehmeier (2001) offers an elegant simplification of Hodes’ argument. More recently, Kocurek (2016) presents a thorough investigation of several inexpressibility results using bisimulations.
See Lampert (manuscript).
This is also acknowledged by Humberstone. See his (1982, fn. 16).
This contrasts with a purely rhetorical use of ‘actually’, as pointed out by Crossley and Humberstone (1977, p. 11).
This interpretation comes from Boolos (1984).
We should point out that Bricker’s formalization depends on his assumption that \(y\prec xx\) may hold at a world at which the objects assigned to the plural variables do not exist (see p. 387). We run along with Bricker on this, but only to illustrate how plural quantifiers do not solve the problem for the twodimensional case even on such controversial grounds.
The move from de dicto to de re knowledge ascriptions has been defended by Soames (2005) on the basis of the following exportation principle—understood relative to an assignment function and pair of possible worlds:
 (E):

For any name n and predicate F, if ‘a knows/believes that n is F’ is true, then ‘a knows/believes that [x / n] is F’ is true.
Soames takes the above principle to be “intuitively compelling,” (p. 261) which seems to be the only support offered in favour of (E). Yet, some twodimensionalists—Chalmers, in particular—deny that (E) has any plausibility whatsoever. See Chalmers (2011, pp. 630–633) for counterexamples to (E) based on knowledge attributions. A more recent defense of de re a priori knowledge can nevertheless be found in Dorr (2011).
Ditto for the corresponding 2D Barcan formula.
For example, see Humberstone (1982, p. 13).
It would be the same if we were quantifying over sets rather than plurally.
See Uzquiano (2011, p. 225).
See also Williamson (2013, p. 248) for similar remarks.
The proof is similar to the necessity of identity. For suppose that \({\mathcal {M}}\,\vDash\, y\prec xx\) for an assignment V. Where \(\langle w,w\rangle\) is any pair of possible worlds in W such that \(\langle w*,w*\rangle \mathcal {R}_{\mathcal {D}}\langle w,w\rangle\), since \(V(y)\in V(xx)\), it follows that \({\mathcal {M}}^w_w\,\vDash\, y\prec xx\), whence \({\mathcal {M}}\,\vDash\, \mathcal {D}y\prec xx\). The other direction is obvious, and the argument for \((\mathcal {D}\nprec )\) is analogous.
Another possible move is to use full quantification over sets, rather than pluralities. This is the strategy adopted by Meyer (2013) in order to eliminate the actuality operator in quantified modal logic. However, the same issues arise concerning what is validated by the models, regardless of whether we assume sets or pluralities in the semantics.
Again, this inference rule is valid for any sentence unless @ is in the language.
The models are constructed resembling the ones in Wehmeier (2001), where he proves a similar result for a firstorder modal language.
References
Boolos, G. (1984). To be is to be a value of a variable (or to be some values of some variables). Journal of Philosophy, 81, 430–49.
Bricker, P. (1989). Quantified modal logic and the plural de re. Midwest Studies in Philosophy, XIV, 372–394.
Chalmers, D. (1996). The conscious mind: In search of a fundamental theory. Oxford: Oxford University Press.
Chalmers, D. (2004). Epistemic twodimensional semantics. Philosophical Studies, 118, 153–226.
Chalmers, D. (2011). Propositions and attitude ascriptions: A Fregean account. Nous, 45(4), 595–639.
Chalmers, D., & Jackson, F. (2001). Conceptual analysis and reductive explanation. Philosophical Review, 110, 315–361.
Chalmers, D., & Rabern, B. (2014). Twodimensional semantics and the nesting problem. Analysis, 74(2), 210–224.
Cresswell, M. (1990). Entities and indices. Dordrecht: Kluwer Academic Publishers.
Crossley, J. N., & Humberstone, I. L. (1977). The logic of “actually”. Reports on Mathematical Logic, 8, 11–29.
Davies, M. (2004). Reference, contingency, and the twodimensional framework. Philosophical Studies, 118, 83–131.
Davies, M., & Humberstone, I. L. (1980). Two notions of necessity. Philosophical Studies, 38(1), 1–30.
Dorr, C. (2011). De re a priori knowledge. Mind, 120, 939–91.
Evans, G. (1979). Reference and contingency. The Monist, 62, 161–189.
Forbes, G. (1985). The metaphysics of modality. Oxford: Clarendon Press.
Forbes, G. (1989). Languages of possibility. Oxford: Basil Blackwell.
Fritz, P. (2013). A logic for epistemic twodimensional semantics. Synthese, 190, 1753–1770.
Fritz, P. (2014). What is the correct logic of necessity, actuality, and apriority? The Review of Symbolic Logic, 7(3), 385–414.
Hanson, W. (2006). Actuality, necessity, and logical truth. Philosophical Studies, 130, 436–459.
Hanson, W. (2014). Logical truth in modal languages: A reply to Nelson and Zalta. Philosophical Studies, 167, 327–339.
Hazen, A. (1976). Expressive completeness in modal language. Journal of Philosophical Logic, 5, 25–46.
Hodes, H. (1984). Some theorems on the expressive limitations of modal languages. Journal of Philosophical Logic, 13, 13–26.
Holliday, W., & Perry, J. (2014). Roles, rigidity, and quantification in epistemic logic. In A. Baltag & S. Smets (Eds.), Trends in logic, outstanding contributions: Johan van Benthem on logic information dynamics (pp. 591–629). Berlin: Springer.
Humberstone, L. (1982). Scope and subjunctivity. Philosophia, 12(1–2), 99–126.
Humberstone, L. (2004). Twodimensional adventures. Philosophical Studies, 118, 17–65.
Jackson, F. (1998). From metaphysics to ethics: A defence of conceptual analysis. Oxford: Oxford University Press.
Kamp, J. A. W. (1971). Formal properties of “now”. Theoria, 37, 227–273.
Kaplan, D. (1989). Demonstratives. In J. Almog, J. Perry & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press.
Kocurek, A. W. (2016). On the expressive power of firstorder modal logic with twodimensional operators. Synthese. doi:10.1007/s1122901611723.
Kripke, S. (1963). Semantical considerations on modal logic. Acta Philosophica Fennica, 16, 83–94.
Kripke, S. (1980). Naming and necessity. Cambridge, MA: Harvard University Press.
Lampert, F. (manuscript). Actuality, tableaux, and twodimensional modal logic.
Lewis, D. (1970). How to define theoretical terms. Journal of Philosophy, 67, 427–446.
Linnebo, Ø. (2016). Plurals and modals. Canadian Journal of Philosophy, 46(4–5), 654–676.
Meyer, U. (2013). Counterpart theory and the actuality operator. Mind, 122(485), 27–42.
Montague, R. (1968). Pragmatics. In R. Klibansky (Ed.), Contemporary philosophy: A survey (pp. 102–122). Florence: La Nuova Italia Editrice. (Reprinted in R. H. Thomason (Ed.), Formal philosophy: Selected papers of Richard Montague. New Haven, CT: Yale University Press, 1974, pp. 95–118).
Nelson, M., & Zalta, E. (2012). A defence of contingent logical truths. Philosophical Studies, 157, 153–162.
Restall, G. (2012). A cutfree sequent system for twodimensional modal logic, and why it matters. Annals of Pure and Applied Logic, 163, 1611–1623.
Soames, S. (2005). Reference and description: The case against twodimensionalism. Princeton: Princeton University Press.
Stalnaker, R. (1978). Assertion. In P. Cole (Ed.), Syntax and Semantics: Pragmatics (Vol. 9, pp. 315–339). New York: Academic Press.
Urquhart, A. (2015). First degree formulas in quantified S5. Australasian Journal of Logic, 12(5), 204–210.
Uzquiano, G. (2011). Plural quantification and modality. Proceedings of the Aristotelian Society, 111(2), 219–250.
van Benthem, J. (2010). Frame correspondences in modal predicate logic. In Proofs, categories and computations: Essays in honor of Grigori Mints (pp. 1–14). London: College Publications.
Vlach, F. (1973). ‘Now’ and ‘Then’: A formal study in the logic of tense anaphora. Ph.D. thesis, UCLA, Los Angeles.
Wehmeier, K. (2001). World travelling and mood swings. In B. Löwe, T. Räsch, & W. Malkzorn (Eds.), Foundations of the formal sciences II. Trends in Logic (pp. 257–260). Dordrecht: Kluwer.
Wehmeier, K. (2004). In the mood. Journal of Philosophical Logic, 33, 607–630.
Wehmeier, K. (2005). Modality, mood, and descriptions. In R. Kahle (Ed.), Intensionality: An interdisciplinary discussion (pp. 187–216)., Lecture notes in logic Wellesley: AK Peters.
Williamson, T. (2003). Everything. Philosophical Perspectives, 17, 415–65.
Williamson, T. (2010). Necessitism, contingentism, and plural quantification. Mind, 119(475), 657–748.
Williamson, T. (2013). Modal logic as metaphysics. Oxford: Oxford University Press.
Zalta, E. (1988). Logical and analytic truths that are not necessary. Journal of Philosophy, 85, 57–74.
Acknowledgements
I am deeply grateful to ISen Chen, Alex Kocurek, Hanti Lin, Harrison SmithJaoudi, Shawn Standefer, and Lloyd Humberstone for comments on different versions of this paper. Thanks also to Adam Sennet, Greg Ray, Ted Shear, G. J. Mattey, Rachel Boddy, Tyrus Fisher, Rohan French, and Greg Restall for many helpful conversations and suggestions. This paper has been presented at the Philosophy Department of the University of California, Davis; the Davis Logic, Language, Epistemology, and Mathematics Working Group; and the Logic Seminar at the University of Melbourne, Australia. I want to express my gratitude to the audiences for many insightful questions and suggestions. Finally, I am thankful to Andrew Parisi for encouraging me to write this paper, as well as to an anonymous referee for comments and advice.
Author information
Authors and Affiliations
Corresponding author
Appendix: Proof of the expressive incompleteness of \(\mathcal {L}\)
Appendix: Proof of the expressive incompleteness of \(\mathcal {L}\)
Since bisimulation implies model equivalence, we show that there is a bisimulation between two constant domain 2Dcentered models differing with respect to sentence (3). But first some notation is in order. Let \(\vec {c}\) denote a sequence of objects or terms of \(\mathcal {L}\), where its length is denoted by \(\Vert \vec {c}\Vert\). The nthmember of \(\vec {c}\) is denoted by \(\vec {c}_n\). Furthermore, let \(\varphi \vec {x}\) be a formula whose free variables are all in \(\vec {x}\), and whenever \(\Vert \vec {c}\Vert =\Vert \vec {x}\Vert\), let \(\varphi \vec {c}\) be the result of appropriately substituting constants in \(\vec {c}\) for variables in \(\vec {x}\). Also, we use \(\langle v,w\rangle\) for members of \(W_1\) and \(\langle y,z\rangle\) for members of \(W_2\), as well as c for members of \({\mathscr {D}}_1\) and d for members of \(\mathscr {D}_2\). We omit any mention of \({\mathcal {M}}\) in the evaluation clauses by writing \(^v_w\,\vDash_1\varphi\) whenever \(\varphi\) holds at \(\langle v,w\rangle\) in \({\mathcal {M}}_1\)—similarly for \({\mathcal {M}}_2\).
Next we define the general notion of a worldobject bisimulation for constant domain 2Dcentered models:
Definition 5.1
(WorldObject Bisimulation) Let \(\mathcal {N}_1=\langle W _1,\langle w*,w*\rangle ,\mathcal {R}_{\Box 1},\mathcal {R}_{\mathcal {D}1},\mathscr {D}_1,V_1\rangle\) and \(\mathcal {N}_2=\langle W _2,\langle v*,v*\rangle ,\mathcal {R}_{\Box 2},\mathcal {R}_{\mathcal {D}2},\mathscr {D}_2,V_2\rangle\) be two constant domain 2Dcentered models. A worldobject bisimulation between \(\mathcal {N}_1\) and \(\mathcal {N}_2\) is a nonempty relation \(\cong \subseteq (W_1\times \mathscr {D}_1)\times (W_2\times \mathscr {D}_2)\) such that \(\langle w*,w*\rangle \vec {c}\cong \langle w*,w*\rangle \vec {d}\), satisfying the following conditions:

1.
\((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\Rightarrow (^v_w\,\vDash_1 P^n_i\vec {c}\, \Leftrightarrow \, ^y_z\,\vDash_2 P^n_i\vec {d})\), for every predicate symbol \(P^n_i\) (the atomic condition);

2.
\((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\Rightarrow \forall n,m\le \Vert \vec {c}\Vert (^v_w\,\vDash_1 \vec {c}_n=\vec {c}_m\Leftrightarrow \, ^y_z\,\vDash_2\vec {d}_n=\vec {d}_m)\) (the identity condition);

3.
\(((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\wedge \langle y,z\rangle \mathcal {R}_{\Box 2}\langle y,z'\rangle )\Rightarrow (\exists \langle v,w'\rangle \in W_1:\langle v,w\rangle \mathcal {R}_{\Box 1}\langle v,w'\rangle \wedge (\langle v,w'\rangle \vec {c}\cong \langle y,z'\rangle \vec {d}))\) (the \(\mathcal {R}_{\Box }\)forth condition);

4.
\(((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\wedge \langle v,w\rangle \mathcal {R}_{\Box 1}\langle v,w'\rangle )\Rightarrow (\exists \langle y,z'\rangle \in W_2:\langle y,z\rangle \mathcal {R}_{\Box 2}\langle y,z'\rangle \wedge (\langle v,w'\rangle \vec {c}\cong \langle y,z'\rangle \vec {d}))\) (the \(\mathcal {R}_{\Box }\)back condition);

5.
\((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\Rightarrow (\langle v,v\rangle \vec {c}\cong \langle y,y\rangle \vec {d})\) (the actuality condition);

6.
\(((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\wedge \langle y,z\rangle \mathcal {R}_{\mathcal {D}2}\langle y',y'\rangle )\Rightarrow (\exists \langle v',v'\rangle \in W_1:\langle v,w\rangle \mathcal {R}_{\mathcal {D}1}\langle v',v'\rangle \wedge (\langle v',v'\rangle \vec {c}\cong \langle y',y'\rangle \vec {d}))\) (the \(\mathcal {R}_{\mathcal {D}}\)forth condition);

7.
\(((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\wedge \langle v,w\rangle \mathcal {R}_{\mathcal {D}1}\langle v',v'\rangle )\Rightarrow (\exists \langle y',y'\rangle \in W_2:\langle y,z\rangle \mathcal {R}_{\mathcal {D}2}\langle y,y'\rangle \wedge (\langle v',v'\rangle \vec {c}\cong \langle y',y'\rangle \vec {d}))\) (the \(\mathcal {R}_{\mathcal {D}}\)back condition);

8.
\(((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\wedge d\in \mathscr {D}_2)\Rightarrow (\exists c\in \mathscr {D}_1:\langle v,w\rangle \vec {c},c\cong \langle y,z\rangle \vec {d},d)\) (the quantifierforth condition);

9.
\(((\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d})\wedge c\in \mathscr {D}_1)\Rightarrow (\exists d\in \mathscr {D}_2:\langle v,w\rangle \vec {c},c\cong \langle y,z\rangle \vec {d},d)\) (the quantifierback condition);
Conditions 1, 3, and 4 are standard for both atomic and modal formulas, although they appear here under a twodimensional framework. Similarly, 8 and 9 are usual for constant domain quantifiers.^{Footnote 47} One can find conditions 2 and 5 for (unrestricted) identity and actuality in Kocurek (2016), which also presents conditions for the fixedly operator. Our conditions 6 and 7 for \(\mathcal {R}_{\mathcal {D}}\) are defined in order to handle the \(\mathcal {D}\) operator.
Lemma 5.1
(Invariance) For any constant domain 2Dcentered models \(\mathcal {N}_1\) and \(\mathcal {N}_2\), if \(\cong\) is a worldobject bisimulation between \(\mathcal {N}_1\) and \(\mathcal {N}_2\), \(\langle v,w\rangle \in W_1\), \(\langle y,z\rangle \in W_2\), and \(\varphi \vec {x}\) is a formula,
Proof
By induction on \(\varphi \vec {x}\). The atomic cases including identity hold by construction given conditions 1 and 2, while the truthfunctional cases are straightforward.
Let \(\varphi \vec {x}\) be \(\Box \varphi \vec {x}\). Suppose that \(\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d}\) and \(^v_w\,\vDash_1\Box \varphi \vec {c}\). If \(\langle y,z\rangle \mathcal {R}_{\Box 2}\langle y,z'\rangle\), then \(\exists \langle v,w'\rangle \in W_1\) such that \(\langle v,w\rangle \mathcal {R}_{\Box 1}\langle v,w'\rangle\) and \(\langle v,w'\rangle \vec {c}\cong \langle y,z'\rangle \vec {d}\), by condition 3. Thus, \(^v_{w'}\,\vDash_1\varphi \vec {c}\). By the induction hypothesis, \(^y_{z'}\,\vDash_2\varphi \vec {d}\). Therefore, \(^y_z\,\vDash_2\Box \varphi \vec {d}\). The other direction is analogous given condition 6.
Let \(\varphi \vec {x}\) be \(\mathcal {A}\varphi \vec {x}\). Suppose that \(\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d}\) and \(^v_w\,\vDash _1\mathcal {A}\varphi \vec {c}\). Thus, \(^v_v\,\vDash_1\varphi \vec {c}\), by the semantics of \(\mathcal {A}\). By condition 5, \(\langle v,v\rangle \vec {c}\cong \langle y,y\rangle \vec {d}\), whence \(^y_y\,\vDash_2\varphi \vec {d}\), by the induction hypothesis. Therefore, \(^y_z\,\vDash_2\mathcal {A}\varphi \vec {d}\), by the semantics of \(\mathcal {A}\). The other direction is analogous.
Let \(\varphi \vec {x}\) be \(\mathcal {D}\varphi \vec {x}\). Suppose that \(\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d}\) and \(^v_w\,\vDash_1\mathcal {D}\varphi \vec {c}\). If \(\langle y,z\rangle \mathcal {R}_{\mathcal {D}2}\langle y',y'\rangle\), then \(\exists \langle v',v'\rangle \in W_1\) such that \(\langle v,w\rangle \mathcal {R}_{\mathcal {D}1}\langle v',v'\rangle\) and \(\langle v',v'\rangle \vec {c}\cong \langle y',y'\rangle \vec {d}\), by condition 6. Thus, \(^{v'}_{v'}\,\vDash_1\varphi \vec {c}\). By the induction hypothesis, \(^{y'}_{y'}\,\vDash_2\varphi \vec {d}\). Therefore, \(^y_z\,\vDash_2\mathcal {D}\varphi \vec {d}\). The other direction is analogous given condition 7.
Let \(\varphi \vec {x}\) be \(\exists y\varphi \vec {x},y\). Suppose that \(\langle v,w\rangle \vec {c}\cong \langle y,z\rangle \vec {d}\) and \(^v_w\,\vDash_1\exists y\varphi \vec {c},y\). Thus, \(\exists c\in \mathscr {D}_1\) such that \(^v_w\,\vDash_1\varphi \vec {c},c\). By condition 9, \(\exists d\in \mathscr {D}_2\) such that \(\langle v,w\rangle \vec {c},c\cong \langle y,z\rangle \vec {d},d\). By the induction hypothesis, \(^y_z\,\vDash_2\varphi \vec {d},d\), whence \(^y_z\,\vDash_2\exists y\varphi \vec {d},y\). The other direction is analogous given condition 8. \(\square\)
We define our two models as follows.^{Footnote 48} Let \({\mathcal {M}}_1=\langle W _1,\langle w*,w*\rangle ,\mathcal {R}_{\Box 1},\mathcal {R}_{\mathcal {D}1},\mathscr {D}_1,V_1\rangle\) be such that \(W_1\) is the set of all pairs of subsets of \(\mathbb {N}\) that are both infinite and coinfinite containing at least one odd number, \(\langle w*,w*\rangle\) is the pair \(\langle 2\mathbb {N}+1,2\mathbb {N}+1\rangle\) of sets of odd numbers, \(\mathscr {D}_1\) is \(\mathbb {N}\), for each \(\langle i,j\rangle \in W_1\) let \(V_1(R,\langle i,j\rangle )=V_1(S,\langle i,j\rangle )=i\cup j\), and let the accessibility relations be defined just as in Definition 1.2. The extension of every other predicate symbol is empty. On the other hand, our second model, \({\mathcal {M}}_2=\langle W _2,\langle w*,w*\rangle ,\mathcal {R}_{\Box 2},\mathcal {R}_{\mathcal {D}2},\mathscr {D}_2,V_2\rangle\), is just like \({\mathcal {M}}_1\) except for \(W_2=W_1\cup \{\langle 2\mathbb {N},2\mathbb {N}\rangle \}\), where \(2\mathbb {N}\) is the set of even numbers, and \(V_2(R,\langle 2\mathbb {N},2\mathbb {N}\rangle )=V_2(S,\langle 2\mathbb {N},2\mathbb {N}\rangle )=2\mathbb {N}\cup 2\mathbb {N}\). No other predicate has an extension defined on \(\langle 2\mathbb {N},2\mathbb {N}\rangle\).
In order to prove that \({\mathcal {M}}_1\) and \(\mathcal {M}_2\) are bisimilar, let \(\langle v,w\rangle \in W_1\) and \(\langle v',w'\rangle \in W_2\) be corresponding pairs of worlds in the models \({\mathcal {M}}_1\) and \(\mathcal {M}_2\). For any \(\langle v,w\rangle \in W_1\) and \(\langle v',w'\rangle \in W_2\{\langle 2\mathbb {N},2\mathbb {N}\rangle \}\), say that the mapping \(\rho\) from \(W_1\) onto \(W_2\{\langle 2\mathbb {N},2\mathbb {N}\rangle \}\) is an isomorphism between \({\mathcal {M}}_1\) and the submodel of \({\mathcal {M}}_2\) defined on \(W_2\{\langle 2\mathbb {N},2\mathbb {N}\rangle \}\) such that \(\rho (c)=d\), in which case for \(\langle v,w\rangle \in W_1\) and \(\langle v',w'\rangle \in W_2\) other than \(\langle 2\mathbb {N},2\mathbb {N}\rangle\), and for any nonempty predicate symbol \(P^n_i\), we have
Moreover, we define the following relations between \({\mathcal {M}}_1\) and \(\mathcal {M}_2\). For any \(\langle v,w\rangle \in W_1\) and \(\langle v',w'\rangle \in W_2\) other than \(\langle 2\mathbb {N},2\mathbb {N}\rangle\), set
For the actuality condition, let
In the case of the extra pair of worlds \(\langle 2\mathbb {N},2\mathbb {N}\rangle \in W_2\), we define
and
Finally, set
Lemma 5.2
\(\cong\) is a bisimulation between \({\mathcal {M}}_1\) and \({\mathcal {M}}_2\).
Proof
Conditions 1, 2, and 5 hold by construction. Since \(\langle 2\mathbb {N},2\mathbb {N}\rangle\) is not \(\mathcal {R}_{\Box }\)accessible with respect to any pair of worlds, conditions 3 and 4 are easily seen to be met as well.
For conditions 6 and 7, the only cases we need to consider involve \(\langle 2\mathbb {N},2\mathbb {N}\rangle\). Suppose \(\langle v,w\rangle \vec {c}\cong \langle v',w'\rangle \vec {d}\) and that \(\langle v',w'\rangle \mathcal {R}_{\mathcal {D}2}\langle 2\mathbb {N},2\mathbb {N}\rangle\). Since \(\vec {c}\) contains a single element, we have only two cases. If \(\vec {c}\) is even, choose a pair \(\langle v,w\rangle \mathcal {R}_{\mathcal {D}1}\langle y,y\rangle\) such that \(\vec {c}\in V_1(S,\langle y,y\rangle )\). By the definition of \({\mathcal {M}}_1\), we know that there will be such a pair, in which case we have \(\langle y,y\rangle \vec {c}\cong \langle 2\mathbb {N},2\mathbb {N}\rangle \vec {d}\), by construction. If \(\vec {c}\) is odd, choose a pair \(\langle v,w\rangle \mathcal {R}_{\mathcal {D}1}\langle z,z\rangle\) such that \(\vec {c}\notin V_1(S,\langle z,z\rangle )\), whence \(\vec {d}\notin V_2(S,\langle 2\mathbb {N},2\mathbb {N}\rangle )\). Therefore, \(\langle z,z\rangle \vec {c}\cong \langle 2\mathbb {N},2\mathbb {N}\rangle \vec {d}\). Condition 7 is analogous.
With respect to conditions 8 and 9, again, we only need to check the cases involving the extra pair \(\langle 2\mathbb {N},2\mathbb {N}\rangle\). Suppose that \(\langle v,w\rangle \vec {c}\cong \langle 2\mathbb {N},2\mathbb {N}\rangle \vec {d}\) and that \(c\in \mathscr {D}_1\). We only have cases involving the predicates S and R, since all the other predicates are empty. If \(c\in V_1(S,\langle v,v \rangle )\), then \(\langle v,w\rangle \vec {c},c\cong \langle 2\mathbb {N},2\mathbb {N}\rangle \vec {d},d\), by construction, since \(V_2(S,\langle 2\mathbb {N},2\mathbb {N}\rangle )\) is not empty. If \(c\notin V_1(S,\langle v,v \rangle )\), then choose any \(d\in V_2(S,\langle w*,w*\rangle )\), in which case \(d\notin V_2(S,\langle 2\mathbb {N},2\mathbb {N}\rangle )\), and then we have \(\langle v,w\rangle \vec {c},c\cong \langle 2\mathbb {N},2\mathbb {N}\rangle \vec {d},d\). The argument for R is very similar. Condition 8 is analogous. \(\square\)
Theorem 5.1
There is no sentence \(\varphi\) of \(\mathcal {L}\) such that for every constant domain 2Dcentered model \({\mathcal {M}}=\langle W ,\langle w*,w*\rangle ,\mathcal {R}_{\Box },\mathcal {R}_{\mathcal {D}},\mathscr {D},V\rangle\), \({\mathcal {M}}\,\vDash\, \varphi\) if and only if there is a pair of possible worlds \(\langle w,w\rangle \in W\), such that for two predicate symbols R and S, \(V(R,\langle w*,w*\rangle )\cap V(S,\langle w,w\rangle )=\varnothing\).
Proof
By construction of the models, in \({\mathcal {M}}_2\) there is a pair \(\langle 2\mathbb {N},2\mathbb {N}\rangle \in W_2\) such that \(V_2(R,\langle w*,w*\rangle )\cap V_2(S,\langle 2\mathbb {N},2{\mathbb {N}}\rangle )=\varnothing\), but in \({\mathcal {M}}_1\) there is no pair of worlds \(\langle w,w\rangle \in W_1\) such that \(V(R,\langle w*,w*\rangle )\cap V(S,\langle w,w\rangle )=\varnothing\), since \(V_1(R,\langle w*,w*\rangle )\) contains only odd numbers, and for any \(\langle w,w\rangle \in W_1\), \(V_1(S,\langle w,w\rangle )\) contains at least one odd number for each coordinate of the pair \(\langle w,w\rangle\). \(\square\)
Rights and permissions
About this article
Cite this article
Lampert, F. Actuality and the a priori. Philos Stud 175, 809–830 (2018). https://doi.org/10.1007/s1109801708945
Published:
Issue Date:
DOI: https://doi.org/10.1007/s1109801708945