Module 3

Basic Syntax and Semantics for Modal Propositional Logic

G. J. Mattey

January 5, 2016

Contents
1 Syntax of Modal Propositional Logic 2
1.1 Expressions of L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Rules of Formation for L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Informal Interpretations of L Operators 3

2.1 ‘’ as a Necessity Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 ‘♦’ as a Possibility Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 ‘J’ as a Strict Implication Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Inter-Definition of Modal Operators 4

3.1 Reduction to One Primitive Modal Operator . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Other Modal Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2.1 Intensional Disjunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2.2 Impossibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.3 Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.4 Strict Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

4 Basic Semantics for L 7

4.1 Truth-Value Assignments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2 Formal Semantics for L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2.1 Generalized Valuation-Functions . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2.2 The Accessibility Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.2.3 Semantical Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.3 Metalogical Properties and Relations in L . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3.1 Truth in a Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3.2 Entailment in a Class of Models . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.3.3 Equivalence in a Class of Models . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.3.4 Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.3.5 Tautologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5 An Alternative Semantics 17

6 Conclusion 18

1
 In this module, we will begin our examination of a language of Modal Propositional Logic, or L .
Modal Propositional Logic is an extension of non-modal Propositional Logic L0 .1 All the formulas
of L0 are formulas of L , and some formulas of L are not formulas of L0 . The first section of the
module will describe the syntax of L . The set of expressions of L0 will be expanded by the addition
of modal operators, and the set of formulas composed of those expressions will be expanded by the
addition of formulas containing modal operators. The brief second section shows how each of the
modal operators can be defined in terms of the others. The third and final section sets out the
basic elements of semantics of L . In later modules, the basic semantics will be refined in order to
produce various semantical systems of modal logic.

1 Syntax of Modal Propositional Logic

Historically, a number of different modal propositional languages have been constructed by taking
different modal operators as primitive.2 Most versions of modal logic syntax add to the expressions
of L0 two one-place operators: the box ‘’ and the diamond ‘♦.’3 We will treat these operators as
primitive and add to the set of primitive operators the two-place “strict implication” operator, ‘J.’
(There will also be a brief discussion below of other modal operators.) A formula which contains a
modal operator will be called a modal formula, and one which lacks a modal operator will be called
non-modal.

1.1 Expressions of L
We extend the expressions of L0 to obtain the vocabulary of L .

• A denumerably infinitely large set of propositional variables ‘p1 ,’ ‘p2 ,’ . . ..

• Two propositional constants ‘>’ and ‘⊥.’

• Two punctuation marks ‘(’ and ‘).’

• A set of five truth-functional operators, ‘∼,’ ‘∧,’ ‘∨,’ ‘⊃,’ and ‘≡.’

• A set of three modal operators, ‘,’ ‘♦,’ ‘J.’

Every expression of L0 is an expression of L , but not every expression of L is an expression of

L0 .

1.2 Rules of Formation for L

Corresponding to the three new modal operators are three new rules of formation, rules 8 through
10, which generate the set FmlL , the formulas of L .

1. All propositional variables are formulas of L .

1
Chellas counts Propositional Logic as a modal logic, noting that, “It does very little violence to the conception of
what constitutes a system of modal logic to count [Propositional Logic] as one, and it simplifies matters enormously to
do so” (Modal Logic: An Introduction, page 47.
2
An operator is primitive when it is specified in the original list of expressions of the vocabulary to which it belongs.
An operator is defined when it does not appear in the original list, but is instead specified through the use of a definition.
Examples will be given in the next section, beginning on page 3.
3
Hughes and Cresswell in A New Introduction to Modal Logic use ‘L’ and ‘M,’ respectively.

2
 2. ‘>’ is a formula of L .

3. ‘⊥’ is a formula of L .

4. If ϕ is a formula of L , then ∼ϕ is a formula of L .

5. If ϕ and ψ are formulas of L , then (ϕ ∧ ψ) is a formula of L .

6. If ϕ and ψ are formulas of L , then (ϕ ∨ ψ) is a formula of L .

7. If ϕ and ψ are formulas of L , then (ϕ ⊃ ψ) is a formula of L .

8. If ϕ and ψ are formulas of L , then (ϕ ≡ ψ) is a formula of L .

9. If ϕ is a formula of L , then ϕ is a formula of L .

10. If ϕ is a formula of L , then ♦ϕ is a formula of L .

11. If ϕ and ψ are formulas of L , then (ϕ J ψ) is a formula of L .

12. Nothing else is a formula of L .

We will stipulate that as an abbreviation, outermost parentheses may be omitted from any formula
which results from the final application of the rules of formation. (So, parentheses must remain in
any proper sub-formula of a formula.)
Since the formation rules for L include all those for L0 , every formula of L0 is a formula of
L . But the L rules of formation allow the generation of formulas of L that are not formulas of
L0 . Thus, L is an extension of L0 .

2 Informal Interpretations of L Operators

In the period before contemporary formal semantics was developed, modal operators were taken
to represent modal terms in natural languages such as English.4 This can be called their “in-
tended interpretation.” In themselves, the modal operators and formulas formed from them have
no meaning, but since the development of possible-worlds semantics, they have been given precise
meanings by formal semantical rules. Before introducing these rules, we will discuss the intended
interpretations of the modal operators.

2.1 ‘’ as a Necessity Operator

In the most common informal interpretation, the box ‘’ is supposed to represent necessary truth.5
For example, we might wish to represent the logically necessary truth6 of the truth-functionally
valid L0 formula ‘∼(p1 ∧ ∼p1 )’ by affixing a box to it:

∼(p1 ∧ ∼p1 ).
4
Or they were interpreted using matrices, as described in Module 1.
5
Analogous modal operators obeying exactly the same semantical rules have been used to represent knowledge, belief,
obligation, future times, and other notions. These informal interpretations will be discussed extensively in Module 6 and
following modules.
6
The necessary truth of “excluded middle” has been disputed by “intuitionist” logicians.

3
2.2 ‘♦’ as a Possibility Operator
The diamond generally is taken to represent possibility. A simple example is a case where it is
affixed to the left of a propositional variable such as ‘p1 .’ In the semantical system S (0), all propo-
sitional variables are true on at least one assignment, so we may wish to say that ‘p1 ’ is at least
possibly true, which can be expressed as follows:
♦p1 .

On a row of a truth-table on which ‘p1 ’ is false, its negation, ‘∼p1 ,’ is true. So we might say that
the negation is possibly true, in which case ‘p1 ’ is possibly false. We can state that ‘p1 ’ is possibly
true and possibly false in this way:
♦p1 ∧ ♦∼p1 .
In such a case, the formula is interpreted in English as representing a state of affairs that is contin-
gent.

2.3 ‘J’ as a Strict Implication Operator

The two-place fish-hook operator is intended to represent “strict implication.” This is a relation
between two formulas that holds when it is impossible for the first to be true and the second false.
When the truth-values are assigned as in the semantics for L0 , this amounts to the same thing as
saying that the first entails the second. In L0 , the formula ‘p1 ∧ p2 ’ entails ‘p1 ,’ {p1 ∧ p2 }  p1 . So we
may wish to write:
(p1 ∧ p2 ) J p1 .
In the next section, these informal notions will be given a rigorous treatment. But it can be
seen already that the modal operators can be used to express in L some of the most important
semantical properties and relations of L0 formulas.7 This was Lewis’s original goal in developing
modal logic.8

3 Inter-Definition of Modal Operators

The modal language L presented here contains a modest number of modal operators in its vo-
cabulary. Each of the operators in the vocabulary of L is treated as primitive or undefined. The
vocabulary could just as well have been stated using just one primitive modal operator. The other
operators could then be defined in terms of that one primitive operator. Despite the fact that ‘,’
‘♦,’ and ‘J’ are primitive in L , we will take a look at how they could be defined in terms of one
another. Then we will examine some other modal operators that can be defined.

3.1 Reduction to One Primitive Modal Operator

The definitions are motivated by the intended interpretations of the operators. Thus if we think of
the ‘’ as expressing necessity, the ‘♦’ as expressing possibility, and the ‘∼’ as expressing negation,
we might say that ‘ϕ’ means that ϕ is necessary. Given that what is necessary is not possibly not
As will be seen, they can also express semantical properties and relations of L formulas.
7

For example, “The strict implication, p J q, means ‘It is impossible that p be true and q false . . . .’ ” A Survey of
8

Symbolic Logic, pp. 332-333.

4
the case, it seems that ‘ϕ’ should be taken as expressing the same thing as ‘∼♦∼ϕ.’ It turns out that
the two formula-schemata are equivalent given the semantics to be developed later.
On the other hand, with the same intended interpretations, we might wish to say that ‘♦ϕ’
expresses that same thing as ‘∼∼ϕ.’ What is possible is not necessarily not the case. Again, these
two formula-schemata will proved to be equivalent in the semantical system we will develop.
The symbol for strict implication is intended to indicate the impossibility that the antecedent is
true and the consequent false. Thus, ϕ J ψ could be defined as ‘∼♦(ϕ ∧ ∼ψ)’ or as ‘(ϕ ⊃ ψ).’ All
three of these will prove to be equivalent in the semantical systems to be developed.
More generally, with ‘♦’ as primitive:
ϕ =df ∼♦∼ϕ.
ϕ J ψ =df ∼♦(ϕ ∧ ∼ψ).
With ‘’ as primitive:
♦ϕ =df ∼∼ϕ.
ϕ J ψ =df (ϕ ⊃ ψ).
We could also take the fish-hook as a primitive operator, defining the box and the diamond in
terms of it.
Exercise. Define the two primitive one-place operators in terms of the fish-hook [challenging].

3.2 Other Modal Operators

A number of other modal operators may be defined in L or taken as primitive.

3.2.1 Intensional Disjunction

In his earliest papers on modal logic, written in 1912, Lewis worked with an operator he called
“intensional” or “dilemmatic” disjunction.9 An intensional disjunction ϕ Y ψ is true just in case the
falsehood of one of the disjuncts logically (or “strictly”) implies the truth of the other.10 Another
way Lewis put it is that it is impossible for both disjuncts to be false.11 Lewis’s stock example of an
intensional disjunction was “Either Matilda does not love me or I am beloved.” We may symbolize
this formula in non-modal Predicate Logic using a one-place predicate ‘Lx,’ for ‘x loves me’ and the
letter ‘m’ for ‘Matilda.’ Then we have ‘∼Lm ∨ (∃x)Lx.’ The falsehood of ‘∼Lm’ implies the truth of
‘Lm,’ which in turn implies the truth of ‘(∃x)Lx.’ The falsehood of ‘(∃x)Lx’ implies the truth of ‘∼Lm.’
This account of intensional disjunction suggests the following definition in terms of the strict
implication operator, which Lewis in fact adopted:
ϕ Y ψ =df ∼ϕ J ψ.
We can also define ϕ Y ψ in terms of the other two primitive operators of the language L :
ϕ Y ψ =df ∼♦(∼ϕ ∧ ∼ψ).
ϕ Y ψ =df (ϕ ∨ ψ).
9
“Implication and the Algebra of Logic” and “A New Algebra of Implications.”
10
The symbol‘ Y’ is peculiar to this text. There is no standard symbol for intensional disjunction.
11
“The Calculus of Strict Implication,” 1914.

5
The reader might notice that the first definition parallels precisely the definition in non-modal
Propositional Logic of ‘∨’ in terms of ‘∼’ and ‘⊃.’ The definition of ‘⊃’ in terms of ‘∼’ and ‘∨’ is
paralleled by the following definition:
ϕ J ψ =df ∼ϕ Y ψ.

3.2.2 Impossibility
By 1914, Lewis employed a one-place primitive operator ‘∼,’ which was intended to express impos-
sibility.12 From this starting point, we can give the following definitions:
♦ϕ =df ∼∼ϕ.
ϕ =df ∼∼ϕ.
ϕ J ψ =df ∼(ϕ ∧ ∼ψ).
ϕ Y ψ =df ∼(∼ϕ ∧ ∼ψ).

3.2.3 Consistency
At the same time Lewis introduced the impossibility operator, he introduced a defined operator for
consistency: ‘◦.’ Two formulas ϕ and ψ are said to be consistent just in case it is possible that they
both be true. Consistency can be defined in terms of other operators as follows:
ϕ ◦ ψ =df ♦(ϕ ∧ ψ).
ϕ ◦ ψ =df ∼∼(ϕ ∧ ψ).
ϕ ◦ ψ =df ∼(ϕ J ∼ψ).
ϕ ◦ ψ =df ∼(ϕ Y ψ).
ϕ ◦ ψ =df ∼∼(ϕ ∧ ψ).

3.2.4 Strict Equivalence

The final modal operator introduced by Lewis was strict equivalence, symbolized here by ‘L.’13
We shall here only give the definition of strict equivalence in terms of the three primitive modal
operators of this text:
ϕ L ψ =df ∼♦(ϕ ∧ ∼ψ) ∧ ∼♦(ψ ∧ ∼ϕ).
ϕ L ψ =df (ϕ ⊃ ψ) ∧ (ψ ⊃ ϕ).
ϕ L ψ =df (ϕ J ψ) ∧ (ψ J ϕ).
Strict equivalence is the only one of the operators, primitive or defined, that cannot be made
primitive without some loss of expressive power. This is analogous to a result for L0 , according to
which the material biconditional ‘≡’ (even with ‘∼’) is not sufficient to define the other non-modal
operators.14
12
“The Matrix Algebra for Implications,” as well as in A Survey of Symbolic Logic, 1918. Note that the symbol in the
present type-face is slightly thicker than the symbol for negation.
13
Lewis’s symbol was ‘=.’ He used that symbol for both strict equivalence and definitional identity, which we would
now say is mixing the object-language with the meta-language.
14
See Geoffrey Hunter, Metalogic, 1996, p. 69.

6
Exercise. Given the informal interpretations of the operators in question, show how to define the
fish-hook in terms of the circle.

4 Basic Semantics for L

The semantics for Modal Propositional Logic is a generalization of the semantics for non-modal
Propositional Logic. The semantical rules for generating truth-tables of L0 assign a truth-value
to a non-atomic formula based on the truth-value(s) of its component(s). From the truth-value
assignment (TVA) made to the propositional variables, the truth-value of the formula containing
those variables results immediately from the application of the semantical rules. Thus if the TVA
assigns to the propositional variable p the value t, then the value of ∼p’ is f, etc.15 In the last
module, it was proved that the operators and their semantical rules of L0 are truth-functional. We
will now say that in such a case, the operators and the semantical rules governing them are directly
truth-functional.
For formulas whose main operator is modal, the situation is different: the standard semantical
rules for modal operators are not directly truth-functional. We cannot obtain the truth-value of ‘ϕ’
simply from a single truth-value for ‘ϕ’ resulting from its TVA. The semantical rules for formulas
whose main operator is a one-place modal operator are not directly truth-functional in the way the
one-place negation operator is. Nonetheless, the determination of the value of the modal formula
‘ϕ’ will turn out to be a function of (possibly) more than one truth-value assigned by the given
interpretation to ‘ϕ,’ and so semantics for modal propositional logic may be said to be indirectly
truth-functional in a way to be described fully below.

4.1 Truth-Value Assignments

We may depict in the standard truth-table format a partial TVA for an L0 formula. So, for example,
if a TVA is such that assigns t to ‘ϕ’ and f to ‘ψ,’ we have a table that looks like this.

ϕ ψ
t f

And by SR-∧ we get:

ϕ ψ ϕ∧ψ
t f f

where the value of the conjunction is f because the assignment to one of its conjuncts is f.
Note that this determination is entirely self-contained. No reference is made to any other TVA
but the one which assigns t to ‘ϕ’ and f to ‘ψ.’ To make another assignment would be to give
a different interpretation, since there is no distinction in the semantical system S (0) between an
interpretation and the truth-value assignments it makes. The truth and falsehood of a formula is
relative to a single given truth-value assignment, and hence to a single given interpretation.
A different interpretation with a different TVA gives a different result for the formula ‘ϕ ∧ ψ.’

ϕ ψ ϕ∧ψ
t t t
15
Here the letters ‘p,’ ‘q,’ and ‘r’ are used as metavariables standing for propositional variables of L .

7
Once again, no reference is made to any other truth-value assignment. It is as if each TVA represents
a “world” of its own.
Although the determination of the truth-values of formulas of L0 depends entirely on an inter-
pretation and its TVA, other semantic properties of the formula can be determined only by looking
at more than one interpretation of it. Whether a formula is truth-functionally valid, for example,
requires that we look at the values for the formula under different TVAs. Thus the formula ‘ϕ ∧ ψ’
is truth-functionally indeterminate, as we have already shown by producing a TVA under which it
is true and a different one in which it is false.16
Now let us consider the disjunction ‘ϕ ∨ ∼ϕ,’ which is truth-functionally valid. We can say this
because there are exactly two possible partial TVAs making an assignment to ‘ϕ,’ producing the
following two tables.17

ϕ ϕ ∨ ∼ϕ
t t

ϕ ϕ ∨ ∼ϕ
f t

Granting that the assignments made to any other propositional variables are irrelevant to the truth
or falsehood of ‘ϕ ∨ ∼ϕ,’ we can say that it is true on all interpretations (which just are truth-value
assignments), and so it is truth-functionally valid.
Truth-functional validity and the related notions are concepts which are not expressible in the
syntax of L0 itself, but only in the meta-language we use to talk about L0 . The formal meta-
linguistic notation for truth-functional truth,
S (0) ϕ ∨ ∼ϕ,
uses a symbol that is not included in the syntax of L0 . But we can express something analogous in
L:
(ϕ ∨ ∼ϕ).
One way to understand the modal operators is as simulating, in the modal object-language L ,
metalogical properties and relations. This is done by extending the object-language L0 to in-
clude modal formulas whose truth-values on an interpretation can depend on multiple truth-value
assignments. Such an extension allows the representation in the object language L of semantical
properties and relations that are more general than simple truth and falsehood.
Though there are three modal operators defined in the syntax of L , we shall limit our initial
discussion to the one-place modal operators ‘’ and ‘♦.’ From a given formula ϕ we can form the
necessity-formula ϕ in L . We will express in the formal semantics the informal meaning of the
necessity formula according to which ϕ is true just in case ϕ is necessarily true. A possibility-
formula ♦ϕ will be true in the formal semantics just in case it is possible that ϕ is true. (Again, it
must be stressed that these modal operators can be given other readings.)
From a formal standpoint, to interpret necessity- and possibility-formulas we need a way of
making reference to multiple truth-value assignments within a single interpretation. The semantics
for Modal Propositional Logic does not identify interpretations with TVAs in the manner of the
16
A formula is truth-functionally indeterminate just in case there is an interpretation that assigns it the value t and an
interpretation that assigns it the value f. All propositional variables are truth-functionally indeterminate.
17
Of course, this result is usually depicted in a single table, but two are used here to indicate the fact that it is not a
row on a truth-table by itself, but the matching of a truth-value to a formula, which gives the desired result.

8
semantics for non-modal Propositional Logic. In the semantics for L , a single interpretation must
be able to allow more than one distinct truth-value assignment to the propositional variables. This
means that the application of an L0 TVA to ϕ is not adequate for the semantics for L .
To accommodate multiple truth-value assignments, we require what are most commonly known
as “possible worlds.”18 We may think of possible worlds as “locations” with respect to which truth-
value assignments are made. Arbitrary worlds will be indicated by the meta-variable ‘w’ with or
without primes or a lower-case italic alphabetic subscript. Specific worlds will be indicated by ‘w’
with or without primes or positive integer subscripts.
Thus, to build on our previous example, we could say that a specific interpretation of ‘ϕ’ contains
two partial truth-value assignments. We can call the first partial TVA, which assigned ‘ϕ’ the value
t, an assignment at w, and the second partial TVA, which assigned ‘ϕ’ the value f, an assignment at
w0 .

ϕ ϕ ∨ ∼φ
t t

w0

ϕ ϕ ∨ ∼φ
f t

4.2 Formal Semantics for L

As with Propositional Logic, we will give a formal specification of the basic semantics for modal
logic. In later modules, we shall make good use of the formal semantics given here.

4.2.1 Generalized Valuation-Functions

Now we can extend the notion of a valuation-function to make it adequate for the formal semantics
of L . What is required is to define a new two-place function v.19 The assignment v maps a pair
hp, wi, consisting of an atomic formula p and a world w onto truth-values, i.e., the set {t,f}.20 In an
instance of the example just given, we can say with respect to v:
v(p1 , w) = t,
v(p1 , w0 )=f.
It is easy to see by SR-∼ and SR-∨, that for any world w,
v(p1 ∨ ∼p1 , w) = t.
It should be noted at this point that the assignments made to propositional variables are not sensi-
tive to values assigned to those same variable at other worlds. It will be seen below that valuations
of other formulas at a world will depend on the valuations of formulas at other worlds.
18
They have also been called assignments, cases, situations, states of affairs, indices, points, etc. Whatever their name,
they are nothing more than reference points for assignments of truth-values to formulas.
19
There should be no confusion in the use of the same symbol for the valuation of formulas of non-modal and modal
propositional logic.
20
Ordered n-tuples are expressed with angled brackets and n elements, separated by commas.

9
4.2.2 The Accessibility Relation
One more piece of machinery is required for the specific semantical rules for the determination of
the truth-values of necessity- and possibility-formulas at a world. Possible worlds in an interpreta-
tion are to be ordered under a relation of accessibility. The inclusion of this relation in the semantics
for modal logic is the hallmark of modern “possible worlds” semantics.
Suppose that a specific world, w0 , is accessible to a specific world, w. This can be depicted
graphically with an arrow from w to w0 .21

w −→ w0

The meta-variable R will indicate an arbitrary two-place relation of accessibility. Thus we can
state what was just depicted graphically in symbolic terms as: Rww0 . Every modal logic interpreta-
tion requires both a non-empty set of worlds and a relation of accessibility among the members of
the set of worlds. (Unless restrictions are placed on the accessibility relation, any world in a given
interpretation may or may not be accessible to any other.)
Together, the set of worlds and the accessibility relation constitute a frame. If we use ‘W’ to
indicate a set of worlds, a frame F can be represented as an ordered pair.
Frame
F=hW, Ri.
We add a valuation-function ‘v’ to a frame to get a model. The model formed in this way is said
to be based on the frame to which the valuation function is added. We will say that a world is in a
model M when it is a member of the set W, which itself is a member of M.
Model
M=hW, R, vi.
The relation of accessibility is a two-place relation, which can be expressed as a set of ordered
pairs hw, w0 i, the first member w of which bears the relation to the second member w0 . In the case of
accessibility, the first member of a pair will be a world and the second member a world accessible
to the first world. An example of a frame is:
h{{w, w0 , w00 }, {hw, w0 i, hw0 , w00 i, hw00 , wi}i.
This may be represented graphically as follows:

←−
w −→ w0 −→ w00

Insofar as we think of a possible world as representing a row of a truth-table, we can think of

each world in a frame as a row of a truth-table which does not have its values filled in. Suppose
we have a frame with two worlds, w and w0 , such that Rww0 . A partial representation of the frame,
covering only the propositional variables ‘p1 ’ and ‘p2 ,’ might look like the following.

w −→ w0
p1 p2 p1 p2
? ? ? ?
21
Hughes and Cresswell use the metaphor of being able to “see” w0 from w.

10
 An interpretation would give truth-values to p1 and p2 . One such interpretation, which assigns
t to p1 and f to p2 at w and t to both p1 and p2 at w0 , can be represented as follows.

w −→ w0
p1 p2 p1 p2
t f t t

The accessibility relation is the basis for determining the truth-value of modal formulas. Specif-
ically, a formula of the form ♦ϕ is true at a world if and only if it is true at some accessible world,
and a formula of the form ϕ is true at a world if and only if it is true at all accessible worlds.
In the example just given, the following values would be generated in an extended truth-table.

w −→ w0
p1 p2 p1 p2
t f t t

p1 ♦p2
t f

In the case of p1 , since there is only one world accessible to w, and p1 is true there, p1 is true at
w. There is no world accessible to w0 at which p2 is true, and so ♦p2 is false at w0 .
In some cases, accessibility plays no role in the evaluation of modal formulas. If, as with p1 ∨∼p1 ,
there is no way to give the formula the value f at any world, it will have the value t at all worlds
that could be accessible to a given world. So (p1 ∨ ∼p1 ) will have the value t at all worlds.
The examples given here have been simple illustrations of the basic semantics. We will now
turn to a formal specification of the semantics.

4.2.3 Semantical Rules

As with non-modal Propositional Logic, the base assignment v must be extended to handle formulas
which are not atomic. Thus, v will be extended to cover not just the non-modal operators, but also
to the modal operators.22 Thus, our notation ‘v’ will be extended further to cover the cases of modal
formulas.
Because valuation-functions have been generalized to include a second argument referring to
possible worlds, we will have to re-state the semantical rules for the truth-functional operators.
All the rules for those operators operate locally, so to speak, in that they apply only at a specific
possible world. To the generalization of the truth-functional rules we add the semantical rules (or
truth-definitions) for formulas governed by the modal operators.
SR-TVA v(p, w) = v(p, w);

SR-> v(>, w) = t;

SR-⊥ v(⊥, w) = f;

t

 if v(ϕ, w) = f;
SR-∼ v(∼ϕ, w) = 

f if v(ϕ, w) = t.



22
Kripke, “Semantical Analysis of Modal Logic I, Normal Propositional Calculi”, Zeitschrift für mathematische Logik und
Grundelagen der Mathematik, 9 (1963), pp. 67-96.

11
 
t

 if v(ϕ, w) = t and v(ψ) = t;
SR-∧ v(ϕ ∧ ψ, w) = 

f if v(ϕ, w) = f or v(ψ) = f.



t

 if v(ϕ, w) = t or v(ψ) = t;
SR-∨ v(ϕ ∨ ψ, w) = 

f if v(ϕ, w) = f and v(ψ) = f.



t

 if v(ϕ, w) = f or v(ψ) = t;
SR-⊃ v(ϕ ⊃ ψ, w) = 

f if v(ϕ, w) = t and v(ψ) = f.



t

 if v(ϕ, w) = v(ψ, w);
SR-≡ v(ϕ ≡ ψ, w) = 

f if v(ϕ, w) , v(ψ, w).



t

 if v(ϕ, w0 ) = t for all w0 such that Rww0 ;
SR- v(ϕ, w) = 

f

 if v(ϕ, w0 ) = f for some w0 such that Rww0 ;

t

 if v(ϕ, w0 ) = t for some w0 such that Rww0 ;
SR-♦ v(♦ϕ) = 

f

 if v(ϕ, w0 ) = f for no w0 such that Rww0 ;

These semantical rules and definitions of ‘frame’ and ‘model,’ when added to the extended
semantical rules for Propositional Logic, yield a semantical system for modal logic that we will for
the time being call the “basic” semantical system.
We will illustrate the use of the basic semantical system with an example. Consider a model
M where W = {w, w0 }, Rww, Rww0 , and v(ϕ, w) = t, v(ϕ, w0 ) = t. This is illustrated by the following
extended truth-table.

x
w −→ w0
ϕ ϕ
t t

On such an interpretation, it follows that v(A, w) = t. The reason is that worlds w and w0 are all
the worlds accessible to w, and ‘ϕ’ is true at both of them. The value of ‘ϕ’ at w0 is not as straight-
forwardly determined, because there are no worlds accessible to w0 . It is standard practice to say
that in this case, it is vacuously the case that ‘ϕ’ is true at w0 . (That is, SR- is understood as saying
that if there are any accessible worlds, then ϕ is true at such worlds. Since the antecedent of the
conditional is false in the present instance, the conditional itself is taken as true. The metalogical
conditional implicit in SR- is treated as a material conditional!) Therefore, v(ϕ, w0 ) = t.

x
w −→ w0
ϕ ϕ
t t

ϕ ϕ
t t

12
 Further, v(♦ϕ, w) = t, since there is an accessible world (w as well as w0 ) at which ‘ϕ’ is true.
However, v(♦ϕ, w0 ) = f. There is no world accessible to w0 at which ‘ϕ’ is true.

x
w −→ w0
ϕ ϕ
t t

♦ϕ ♦ϕ
t f

This shows the weakness of the basic semantics we have developed. There are interpretations on
which a formula is “necessarily true” but not “possibly true.” This is because the basic semantics
allows what Hughes and Cresswell call “dead-end” worlds, worlds to which no world is accessible.

4.3 Metalogical Properties and Relations in L

Given the truth-definitions for formulas of L , we can define semantical properties and relations of
L formulas analogous to those of L0 formulas. Because different restrictions on the accessibility
relation in frames will generate different systems of modal logic, we shall here give formal defi-
nitions which are fully general and can apply to any semantical system which will be considered
in what follows. For this reason, we will define the notions of entailment and validity relative to
a frame. We can then later define them for more specific modal systems. We shall postpone the
proof of modal forms Bivalence and Truth-Functionality, as well as discussion of a modal form of
Semantical Consistency, until Module 5, when we consider the semantical system S(K).

4.3.1 Truth in a Model

Formulas have truth-values relative to worlds and more generally to models, which themselves are
frames with the addition of a valuation function: M = hW, R, vi. The truth-value in a model of
any formula at a world is determined by the world, its relation to other worlds, and the valuation
function given by the model. Thus, if v ∈ M, we will say that ϕ is assigned the value true at w in W
in M by v in M if and only if v(ϕ, w) = t. This is symbolized as M w ϕ.
Truth in a Model
Given a model M, M w ϕ just in case v is in M, w is in W, which is in M, and v(ϕ, w) = t.
This definition can be applied to the semantical rules given above, to produce a definitionally
equivalent set of semantical rules SR*:

SR*-TVA M w p if and only if v(p, w) = t;

SR*-> M w > for all worlds w;

SR*-⊥ M 2w ⊥ for all worlds w;

SR*-∼ M w ∼ϕ if and only if M 2w ϕ;

SR*-∧ M w ϕ ∧ ψ if and only if M w ϕ and M w ψ;

SR*-∨ M w ϕ ∨ ψ if and only if M w ϕ or M w ψ;

13
SR*-⊃ M w ϕ ⊃ ψ if and only if M 2w ϕ or M w ψ;
SR*-≡ M w ϕ ≡ ψ if and only if M w ϕ if and only if M w ψ;
SR*- M w ϕ if and only if for every w0 in M, such that Rww0 , M w ϕ;
SR*-♦ M w ♦ϕ if and only if for some w0 in M, such that Rww0 , M w ϕ;

For example, by the definition of truth in a model, M w ∼ϕ if and only if v(∼ϕ, w) = t. By SR-∼,
v(∼ϕ, w) = t if and only if v(ϕ, w) = f. By definition of truth in a model again, v(ϕ, w) = f if and only
if M 2 ϕ. So M w ∼ϕ if and only if M 2 ϕ.
Exercise: Show the equivalence of the other SR* rules to the SR rules.

4.3.2 Entailment in a Class of Models

Modal semantics requires that we expand the definition of entailment to accomodate the inclusion
in the semantics of possible worlds, the accessibility relation, and the two-place valuation-function.
In the modal semantics, a formula only has a truth-value at a world. Thus we will say that relative
to a given class C of models, the relation of entailment holds between a set Γ = {γ1 , γ2 , . . . , γn } and
a formula ϕ when all models based on the frame which make all the formulas γi of the set true at
a world also make the formula ϕ true at that world. The strict definition of entailment in a class of
models C is as follows
Entailment in a Class of Models
Given a class of models C , {γ1 , γ2 , . . . , γn } C ϕ just in case for any model M in C and any w in
M, if M w γ1 , M w γ2 , . . . , M w γn , then M w ϕ.

For example, consider the frame used earlier, which determines a class C of models which differ
only with respect to v. F = {h{w1 , w2 }, {hw1 , w1 i, hw1 , w2 i}. Suppose we want to determine whether
{ϕ} C ♦ϕ. That is, we want to determine whether there are any models M based on F such
that at some world in W, ‘ϕ’ is true while ‘♦ϕ’ is false. In fact, there is such a model, the one we
considered above. It contains a world, w0 , such that v(ϕ, w0 ) = t and v(♦ϕ, w0 ) = f.
w −→ w0
ϕ ψ ϕ ψ
t f t t

ϕ ♦ϕ
t f

So the entailment fails in the given class of models.

Now we will consider a entailment which holds for all classes of models: {(ϕ ∧ ψ)} entails ϕ.
Suppose for an arbitrary frame F, an arbitrary world w in W in F, and an arbitrary model M based
on F, that v((ϕ ∧ ψ), w) = t. Then by SR-, at all accessible worlds w0 , v(ϕ ∧ ψ, w0 ) = t. By SR-∧,
v(ϕ, w0 ) = t. Then again by SR-, v(ϕ, w) = t.
We may illustrate this reasoning by way of truth-tables. To do so, we need a further notational
device. When we place an asterisk over an accessibility arrow, it indicates that we are reasoning
about all accessible worlds. (When an asterisk is placed below the arrow, it indicates that we are
reasoning about at least one accessible world.) The first step is to assume that (ϕ ∧ ψ) is true at an
arbitrary world w and infer that ϕ ∧ ψ is true at all accessible worlds w0 .

14
 ∗
w −→ w0
(ϕ ∧ ψ) ϕ∧ψ
t t

Next we infer that the truth-value of ϕ ∧ ψ at w0 is t.

∗
w −→ w0
(ϕ ∧ ψ) ϕ∧ψ
t t
ϕ
t

Finally, we infer that ϕ is true atw.

∗
w −→ w0
(ϕ ∧ ψ) ϕ∧ψ
t t
ϕ
t
ϕ
t

4.3.3 Equivalence in a Class of Models

We can use the definition of entailment in a class of models to form a definition of equivalence in a
class of models. Two formulas of L ϕ and ψ are equivalent in a class of models C just in case the
set consisting of the one entails in the frame the other, and vice-versa.
Equivalence in a Class of Models
ϕ is equivalent to ψ in a class of models C if and only if for all models M in C and all worlds w
in M, v(ϕ, w) = v(ψ, w).
The following meta-theorem can then be proved.
ϕ is equivalent to ψ in a class of models C just in case, in C, {ϕ} C ψ and {ψ} C ϕ.
Exercise. Prove the meta-theorem just stated.
We will here illustrate with extended truth-tables the equivalence of ‘∼ϕ’ and ‘♦∼ϕ.’
∗
w −→ w0
∼ϕ
t
∼ϕ
t
ϕ
f
♦ϕ
f
∼♦ϕ
t

15
 The following table reflects the fact that if a possibility-formula ♦ϕ is false at a world, then ϕ is
false at all accessible worlds, by SR-♦.
∗
w −→ w0
∼♦ϕ
t
♦ϕ
f
ϕ
f
∼ϕ
t
∼ϕ
t

4.3.4 Validity
The limiting case of entailment in a class of models is validity in a class of models. We have seen
in the case of Propositional Logic that validity can be treated as entailment by the empty set of
formulas, and here it is understood as relativized to a class of models: ∅ C ϕ. A formula ϕ is valid
in a class of models if and only if it is true at all worlds in all models in that class.
Validity in Class of Models
C  ϕ iff for all M in C and all w ∈ W, M w ϕ.
To return to a previous example, ‘(ϕ ∨ ∼ϕ)’ is valid in any class of models whatsoever. No
matter how ‘ϕ’ is interpreted at any given world, as t or f at that world, the formula ‘ϕ ∨ ∼ϕ’ is true
at every world that could be in a frame. Therefore, ‘ϕ ∨ ∼ϕ’ is true at any world accessible to a
given world w in any frame, in which case ‘(ϕ ∨ ∼ϕ)’ is true at w. So in any model and any world
based on any frame, ‘(ϕ ∨ ∼ϕ)’ is true, which was to be proved. We can generalize this reasoning
to conclude that the necessitation of every S(0)-valid formula of L0 is valid in every frame in the
basic semantical system.

∗
w −→ w0
ϕ ∨ ∼ϕ
t
(ϕ ∨ ∼ϕ)
t

We can extend the notion of validity in a class of models to that of a class F of frames. Thus
we write ‘F ` ϕ’ to indicate that ϕ is true at all worlds in all models based on a frame in F .
Validity in Class of Models
F  ϕ iff for all M based on any frame F in F and all w ∈ W, M w ϕ.
When entailment, equivalence, or validity holds for all classes of frames, we shall drop the ref-
erence to the frames. Later, we will develop a number of semantical systems by placing restrictions

16
on the accessibility relation. Then we will introduce various classes of frames, each of which con-
tains the restricted accessibility relation as a member. Accordingly, we will speak of entailment,
equivalence, and validity in all frames of a certain class.

4.3.5 Tautologies
A special case of formulas valid in all classes of frames is the tautology, a formula that is valid in
S (0), in the sense that it is true on all S (0) valuations. More formally, if ϕ is a formula of L0 , then
ϕ is a tautology if and only if v(ϕ) = t for all valuation-functions v. When we expand the language
of Propositional Logic to L , we can generate tautology-like formulas, such as p1 ⊃ p1 . To prove
that such a formula is valid in a frame, we would show that either v(p1 , w) = f or v(p1 ) = t. By
SR, this is equivalent to saying that either p1 is true at all accessible worlds or p1 is false at some
accessible world. To show this requires proving bivalence, that at each world a formula receives
the value t or f. This will be done in the next module.

5 An Alternative Semantics
The basic semantical system developed above follows one approach of Kripke, which was adopted
in the canonical text by Hughes and Cresswell. For another purpose, Kripke devised an alterna-
tive presentation of the semantics. In its simplest guise, his idea was that rather than work with
valuation-functions which map propositional-variable/world pairs to truth-values, we can use a
function which maps propositional variables to sets of worlds, i.e., those at which the variable is
true. 23 This idea was developed in detail by Brian Chellas and later by Aldo Antonelli.24 The
approach trades on the bivalence and truth-functionality of assignments of truth-values to atomic
formulas, that is, that every atomic formula has one and only one of the values t and f.
Here is an example to motivate the alternative semantics. Consider in non-modal Propositional
Logic two propositional variables p1 and p2 . To each of these, there are only two partial assignments
of truth-values. Let us label as ‘vt ’ those assignments that give the variable the value t and as ‘vf ’
those which give it the value f. Then there are four possibilities: vt (p1 ) = t, vf (p1 ) = f, vt (p2 ) = t,
and vf (p2 ) = f. This is the array of possibilities encoded in a four-row truth-table, with each row
receiving a unique label.

row p1 p2 ϕ
r1 t t ?
r2 t f ?
r3 f t ?
r4 f f ?

Now consider a function V which maps each propositional variable into a set of rows of the
truth-table, such that if the value of p is t at row ri , then ri ∈ V(p). In the present case, we would
have:
V(p1 ) = {r1 , r2 }
V(p2 ) = {r1 , r3 }
23
“Semantical Analysis of Modal Logic I,” pp. 93-94.
24
Brian F. Chellas, Modal Logic: An Introduction, Cambridge University Press, 1980, G. Aldo Antonelli, “Classical
Correspondence Theory for Basic Modal Logic,” 2014, http://www.aldo-antonelli.org/papers/CCTML.pdf.

17
 The rows r1 and r2 are the rows at which p1 is true and those at r3 and r4 are those at which
p1 is false. Given that each row of the truth-table is a “model” in the sense of being a truth-value
assignment, we could then write M r1 p1 and M r2 p1 , and M r1 p2 and M r3 p2 . Moreover,
we would have it that M 2r3 p1 and M 2r4 p1 , and M 2r2 p2 and M 2r4 p2 . Then we could give a
semantical rule for negation, such that M r3 ∼p1 , M r4 ∼p1 , M r2 ∼p2 , and M r4 ∼p2 . Similar
rules could be given for the other non-modal operators. Finally, one might give a semantical rule
for the ‘’:
M ri ϕ if and only if, for all r j , M r j ϕ, i.e., r j ∈ V(ϕ).
If we take rows on a truth-table to constitute “worlds,’ then we could say that V(p) constitutes a
set of worlds at which p “holds” or “is true.” A set W of “worlds,” would be a set of rows on a
truth-table.
This approach is equivalent to taking valuation-functions, rather than rows on truth-tables, as
the “worlds.” Then a world would be a function from propositional variables to sets of valuations.
Understanding worlds in terms of valuation-functions was the basis of Kripke’s original semantics
for modal logic. A world for Kripke was “a ‘complete assignment’, that is, a function assigning
a truth-value to every atomic sub-formula of a formula.”25 As Kripke noted, this kind of set-up
generates a semantical system for the Lewis system S5 (a fact which we are not now in a position
to demonstrate). What is important here is that the range of “worlds” is limited: “there can be no
two worlds in which the same truth-value is assigned to each atomic formula.” 26 There can be no
“duplicate” rows on any truth-table assigning values to all the propositional variables.
To gain the flexibility we need to generate a number of different semantical systems, we must
de-couple “worlds” from rows on complete truth-tables, so that there could be more than one
“world,” all of whose members are the same. Each propositional variable p is assigned a set V(p)
of worlds. More technically put, V maps the set of all propositional variables into the power-set
P(W) of worlds, that is into the set of all subsets of W. For example, let W = {w1 , w2 , w3 }. Then
V(p) might be {w1 , w3 }, which is another way of saying that p is true at worlds w1 and w3 . A frame
is defined as before, as consisting of a set of worlds and a binary relation on the worlds. We can
then stipulate the following semantical rule:
M w p if and only if w ∈ V(p).
Semantical rules for formulas which are not propositional variables are as with SR* above.
The definitions for entailment, equivalence, and validity in classes of models and frames are as
with the Kripke-style semantics.27 It is apparent that the two approaches of Kripke and Chellas are
equivalent. That is, it does not matter whether we define M w p as v(p, w) = t or as w ∈ V(p), and
the semantical rules for all non-atomic sentences work in exactly the same way on both approaches.
For example, as noted above, v(∼ϕ, w) = t if and only if M w ∼ϕ.
25
“Semantical Analysis of Modal Logic I,” p. 69.
26
“Semantical Analysis of Modal Logic I,” p. 69.
27
The rule as stated here is from Chellas, who informally describes a “valuation” as “an assignment of truth values
(truth and falsity) to propositionally atomic sentences.” He includes as atomic modal sentences ϕ and ♦ϕ in order
to account for the validity of tautologies such as p1 ⊃ p1 . Antonelli gives the formal account described above of
valuations for non-modal propositionally atomic sentences and defines tautologies thereby. He then counts formulas
such as that just mentioned as valid because they are substitution instances of non-modal tautologies.

18
6 Conclusion
It may be useful to reflect on the nature of the accessibility relation. One way to think of it is as
representing those situations that matter with respect to the modality in question. For example, we
might think of something as necessary when it is inevitable in the sense that there is no alternative
that matters to its being the case. The alternatives that matter might be thought of as being “ac-
cessible worlds.” Truth of the necessity-formula ϕ would require truth of the embedded formula
ϕ in all those worlds that matter. A possibility-formula would indicate that its embedded formula
is true in at least one of the worlds that matters. We shall in what follows have much more to say
about how accessibility captures our intuitions about the various kinds of modalities.
We are now finally in a position to see why the semantics for modal propositional logic is in-
directly truth-functional. Although the truth of a necessity formula (or any other modal formula)
is not directly a function of the assignments made to its components, it is a function of the as-
signments made to its components at the accessible possible worlds. And those assignments are
ultimately based on truth-value assignments to the propositional variables from which the formula
is composed. In Module 5, we will give a definition of modal Truth-Functionality in connection
with specific semantical system and prove inductively that the system conforms to that definition.

19