Kripke's Semantical Considerations on Modal Logic: Possible Worlds Semantics
Saul Kripke's 1963 account of possible-worlds semantics for quantified modal logic, showing how varying domains affect reasoning about necessity and possibility.
PhilosophyAll chapters Reading tools Notes & comments
Reasoning about necessity and possibility › A World With a Different Guest ListFull chapter · Sections in this chapter1 / 9 ↓
A World With a Different Guest List
Imagine two complete scenarios for a committee meeting. In the first, Alice is the sole member present, and she has signed the minutes. In the second, Alice and Bob are both present, and Alice has signed but Bob has not. Suppose both scenarios count as possible from the standpoint of the first. Then each person present in the first scenario has signed in every relevant scenario: Alice is that person. Yet it is false that everyone present signs in every relevant scenario. Bob is the counterexample in the second one.
It takes some care to express the difference. One reading fixes the people in the actual scenario and asks whether each has a property in all relevant alternatives. The other visits each alternative first, then asks whether all the people present there have the property. The two procedures can choose different people. This is a variation on one of the counterexamples in Saul Kripke's “Semantical Considerations on Modal Logic”, originally published in 1963. He uses it to show why putting a quantifier before or after “necessarily” changes what a sentence requires.
The paper begins with a framework for modal sentences and concentrates on extending it to quantifiers. Its motivating questions are precise. Which other situations matter to truth at a given situation? Which individuals do quantifiers range over in each one? How can a variable keep referring to an individual when the evaluation moves to another situation? Kripke proposes specific answers and shows what follows from them.
Evaluating a Claim at a World
In the opening discussion of model structures, Kripke calls the situations worlds. A normal model structure is a triple (G, K, R). K is the set of worlds. G is a particular world in K where validity will be tested. R is a relation between worlds. If w R v, then v is accessible from w, in the sense of representing a possibility relative to w. The relation in the paper's basic normal structures is reflexive. A world counts as one of its own alternatives.
A model adds an assignment of truth-values to atomic propositions at each world. The claim that Alice signed is represented here by an atomic proposition. Its value in one world may be true and in another false. To evaluate a conjunction at a world, we look at both of its components at that world. To evaluate a negation, we look at the proposition it negates. Necessity introduces a different operation: we look at a whole set of other evaluation points, determined by the accessibility relation from where we began.
□A holds at w if A holds at every world accessible from w. ◇A holds at w if A holds at some world accessible from w. The possibility clause follows by treating ◇A as ¬□¬A.
It matters that accessibility is relative to a starting point. Suppose three plans are on a table. From the first plan only the first and second are allowed alternatives. From the second, the third can also be considered. If a claim is true in the first two plans and false in the third, it can be necessary at the first and fail to be necessary at the second. It depends on which alternatives count relative to where the evaluation is taking place. A nested modal claim can make us change that starting point.
These conditions define a mathematical structure. One can use the same machinery without deciding that there is a single correct metaphysical interpretation of the worlds. A model gives a way to test which modal inferences follow from assumptions about the relation. Questions about which actual alternatives deserve the name “possible” require an interpretation of that model.
How a Relation Governs a Logical Law
Start with □A implies A. If each world is accessible from itself, necessity of A at w requires A at w. This is the role of reflexivity. A modern textbook may label the axiom or basic reflexive system T. The 1963 paper calls the basic model structures M, and discusses stronger systems obtained by adding conditions. The letter K in Kripke's triple is the set of worlds, regardless of what names a later textbook gives to modal calculi.
Transitivity supplies a second familiar case. If v is accessible from w and u from v, transitivity makes u accessible from w. Suppose □A holds at w. Then A is true throughout the alternatives to w. Every alternative to v is among those points, so □A holds at v. Since the same reasoning applies to each v accessible from w, □□A holds at w. This is how the axiom □A implies □□A is validated on transitive structures. Reflexivity and transitivity characterize the S4 structures in Kripke's discussion.
We can test the role of transitivity with the three plans. Keep every world's access to itself. Give the first access to the second and the second access to the third. Leave out the first world's access to the third. A proposition that is true at the first and second and false at the third is necessary at the first. It is not necessary at the second, so it fails to be necessarily necessary at the first. The countermodel spells out the exact missing link.
Kripke also distinguishes a symmetric case, associated with the Brouwersche system, and an equivalence relation for S5. In an equivalence relation, worlds fall into classes in which each member is accessible from every member of the class. A world in one class need not have access to a world in another. The set of accessible worlds stays the same as one moves within a class. That restricted statement explains why possibility can be stable across alternatives in S5, even though worlds may differ in what is actually true at each of them.
The paper states a completeness result relating logical derivability in the named calculi to validity at the distinguished world across the corresponding model structures. This is a connection between two tests. Soundness says the deductive rules license only valid formulas, while completeness says they suffice to derive every valid formula. Establishing that a formula is valid takes more than finding one model in which it is true. Finding one model of the right kind where it fails, however, is enough to disprove its validity.
Quantifying Over Individuals
The paper's main work starts when we add expressions such as “everyone” and “someone.” In its section on quantificational model structures, each world w receives a domain D(w), a set of individuals that exist in that world. A universal quantifier at w checks each individual in D(w). An existential quantifier asks whether at least one such individual satisfies the formula. These are local clauses. If the evaluation has already moved to a different world, the quantifier uses the domain of that world.
What happens when a variable was already assigned an individual before evaluating a modal operator? Its assignment can stay fixed across evaluation at another world. If that individual does not belong to the new world's domain, a question remains about the truth of a predicate applied to it. Kripke discusses both the possibility of leaving such claims without a truth-value and the convention he adopts for the treatment in this paper, which evaluates free-variable assignments from the union of the domains. A predicate gets an extension at each world, and an assignment from that union can be checked even when the assigned individual is absent from the local domain. The objects that count for a quantifier at that world are still those in its own domain.
One can see the point with a club's mailing list. In an alternative where someone is no longer a member, the universal claim about everyone in the club leaves that person outside its scope. A claim about whether that particular person has already signed a document can still be evaluated. Different predicates and choices of their extensions will make different claims true under such assignments. Kripke gives reasons for allowing the broader extensions and explains what would have to change if atomic predicates were uniformly made false of individuals outside the current domain.
The two conventions concern different aspects of evaluation. The domain tells us whom we can range over with a quantifier at the current world. A variable assigned earlier gives us an individual about whom a claim can be considered at another world. Combining the two operations requires both pieces of information. Moving a quantifier across a modal operator can change which domain is used before we ever reach the predicate.
Why the Barcan Formulas Can Fail
In the two countermodels, Kripke uses two worlds accessible from each other and themselves. This makes accessibility an S5 equivalence relation. The failures are therefore possible even with a strong propositional modal logic. The examples turn on the change in domain. In the first, G has domain containing only a, while H includes both a and b. Let P hold of a in both worlds and fail of b in H. Then ∀x □P(x) holds at G. Its one relevant individual, a, satisfies P at every accessible world. But □∀x P(x) fails at G, since H is accessible and b in H fails to satisfy P. That invalidates the Barcan formula ∀x □P(x) → □∀x P(x) for these models.
For the converse, swap the sizes of the domains. G contains a and b, while H contains only a. Let P hold of everyone in each world's local domain, and stipulate that P is false of b at H. Then □∀x P(x) is true at G because every accessible world satisfies the local universal claim. The claim ∀x □P(x) is false at G. Its quantifier ranges over b as well as a, and b fails the modal test at H. This invalidates the converse Barcan formula in the same framework.
The two examples use absence differently. In the first, the new individual b is available at the very world where the universal claim fails. Nothing turns on how P would be evaluated of b in G, where b is absent. In the second, the failure depends on evaluating P at a world where b is absent. Kripke points this out and considers what changes under alternative conventions about truth-values in that case. A logic with fixed domains would remove both changes of population. These countermodels would then be unavailable, since every quantifier would range over the same set for these comparisons.
These examples also show why a countermodel is more informative than saying a formula sounds doubtful. Each world, its domain, and the values or extensions needed for the calculation are specified. To see if the example works, one can ask whom the outermost operator sends us to check, then apply the next operator there. For the two formulas the order of that procedure differs. The resulting truth-values track the difference precisely.
Why Necessitation Needs Care
The paper tests a tempting derivation of the converse Barcan formula. Start with ordinary universal instantiation. If everything has a property, an individually specified thing has the property. The derivation tries to put necessity around such a claim and then moves toward a statement beginning with a quantifier over individuals. The issue is how to read the individual variable before and after putting the box in front of the formula.
Kripke uses the generality interpretation of theorems with free variables. When one of these is asserted as a theorem, read it as asserting its universal closure, with the free variables bound by universal quantifiers. So putting a box before the theorem in the legitimate application of necessitation places it outside that closure. In the attempted derivation, the claimed result is itself a theorem with a free variable. Its universal closure places its quantifier outside the box. The two operations have silently traded places. That is exactly the exchange the examples give us reason to question.
Notice how the formal and semantic parts of the paper meet. One can look at a proof sequence and see a familiar rule with a familiar formula. That formula's interpretation as a whole determines what the rule can be applied to. Once the relevant closure is made explicit, one can see why the apparent deduction does not establish an implication that already failed in a particular model. No change to the counterexample is needed to account for the mistake.
An Existence Predicate and a Test
The paper's discussion of existence goes one step further. Kripke allows a predicate E(x) whose extension at a world is just the domain of that world. It is true of a given individual precisely where that individual belongs to the domain. If one quantifies over everyone at a world, everyone in that domain satisfies E. At every relevant world that much is true. Nevertheless a person in the domain where we began can fail to belong to the domain of an accessible world. So the claim that everyone in the current domain necessarily exists can fail.
The example makes the second Barcan countermodel especially easy to remember. In G, both a and b exist. In H only a exists. When we ask whether everyone exists at H, the answer is yes because H's quantifier considers only a. When we carry the question whether b exists from G to H, the answer is no. In a model with that domain change it is coherent to say both things. It would take an additional assumption about individuals and worlds to make existence persist along access to alternatives.
What the Semantics Commits Us To
The structure can be used with different interpretations, but the truth of a particular formula depends on the interpretation one actually supplies. Kripke illustrates the point near the end of the paper by considering provability interpretations. In one, worlds include a consistent extension of Peano arithmetic together with a model of that extension. An alternative is accessible if its theory extends the original one. More information can then narrow which models remain under consideration. Reflexivity and transitivity come from how extension works. The proposal lets Kripke connect a necessity claim about an arithmetical sentence to what the associated extension proves.
This is another example of the distinction between a verified interpretation and a label. Calling alternatives “worlds” does not establish a particular principle about them. The relation has to be given, the valuations have to be specified, and the inference must hold on the resulting structures. In the arithmetic interpretation, taking a further extension need not be reversible. In the committee case, access can be symmetric even with different populations. The apparatus is flexible enough for both, and explicit enough that we can see what would need to change if we wanted a different result.
Reading Kripke
The primary text in the UNAM-hosted reprint begins on reprint p. 63, corresponding to p. 83 in the original Acta Philosophica Fennica volume. Look at reprint pp. 64–65 for the model structures and necessity clause, pp. 65–67 for quantificational structures, and pp. 67–69 for the Barcan countermodels and the use of necessitation. Work through the first countermodel at the designated world, then repeat with the two domains swapped and the predicate extensions changed as Kripke specifies. Following the two orders of evaluation is the quickest way to see what the definitions add.
Join the discussionAsk a question or share what you learned.
Keep exploring
Reference
Citation details
Cite or share this article.

Comments
Add to the conversation
A question, a correction, or an example from your own work.
No comments yet. Be the first to share your thoughts!