Philosophy 134

Module 2
Non-Modal Propositional Logic
G. J. Mattey

January 4, 2016

Contents
1 Syntax of Propositional Logic 3
1.1 Expressions of L0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Rules of Formation for L0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Semantics for Propositional Logic 5

2.1 Truth-Table Semantics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Verum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Falsum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.4 Conjunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.5 Disjunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.6 Material Conditional . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.7 Material Biconditional . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Formal Semantics for L0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Semantical Rules for L0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Semantical Properties and Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3.1 Bivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.2 Truth-Functionality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.3 Entailment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.4 Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.5 Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.6 Satisfiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Natural Deduction in Propositional Logic 17

3.1 Reiteration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Negation Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 Conjunction Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Disjunction Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5 Material Conditional Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.6 Material Biconditional Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.7 The Verum Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.8 Falsum Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

1
 3.9 Derived Rules Using Falsum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.10 Derivations of Schemata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.11 Derivational Properties and Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.11.1 Derivability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.11.2 Derivational Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.11.3 Theoremhood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.12 Excluded Middle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.13 Non-Contradiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.14 Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

4 Soundness and Completeness 28

4.1 Soundness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2 Completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

5 Axiomatic Formulation of Propositional Logic 33

5.1 The Axiom System A(0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.2 A(0) and S(0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.3 A(0) and D(0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2
 The overall goal of this text is to introduce the reader to some of the main systems of modal
logic. It is customary to preface the treatment of modal logics with preliminary accounts of non-
modal logics, and this in two ways. First, there is a treatment of Propositional Logic, which is
followed by a treatment of systems created by adding modal symbolism to the symbolism of Propo-
sitional Logic.1 Second, there is a treatment of Predicate Logic, which is followed by systems adding
modal symbolism to it.2 This text will adhere to that custom, though it is quite feasible to introduce
modal Predicate Logic from the start.
The logic of propositions was first investigated by the ancient Stoics, who recognized five rules
of inference.3 It was investigated extensively in the nineteenth century by such logicians as Augus-
tus De Morgan, George Boole, and Ernst Schröder, who explored the relations between logic and
algebra.4 The first presentation in the modern form was by Gottlob Frege5 , and its first in the cur-
rent notation was by Bertrand Russell and Alfred North Whitehead.6 The present-day truth-table
semantics was developed by Ludwig Wittgenstein7 and Emil Post.8 Non-axiomatic, or “natural de-
duction” systems were introduced by Gerhard Gentzen,9 and given the form used here by Frederick
Fitch.10

1 Syntax of Propositional Logic

Propositional Logic is based on a formal language L0 which is entirely artificial or conventional. All
formal languages have a syntax, which determines what symbols belong to the language and how
those symbols can be combined to form grammatically correct strings of symbols. Here we will
give one formulation of L0 , with the recognition that there are many variant ways of specifying its
symbolism.

1.1 Expressions of L0
The formal language L0 consists of a set of expressions (its vocabulary) and a set of rules of formation
which generate a set of sentences of L0 .11 The expressions of L0 consist of the following:
Expressions of L0

• A denumerably infinitely large set of propositional variables: p0 , p1 , . . ..

• A propositional constant: ‘>’ (verum).

• A propositional constant: ‘⊥’ (falsum).

1
What is here called “propositional logic” is elsewhere called “sentence logic,” “sentential logic,” “propositional cal-
culus,” “boolean logic,” “truth-functional logic,” and “the two-valued logic.”
2
What is here called “predicate logic” is elsewhere called “predicate calculus,” “quantificational logic,” “first-order
logic,” among other things.
3
See William and Martha Kneale, The Development of Logic, 1962, Chapter III, Section 5.
4
C. I. Lewis gives a lengthy description of these developments in his Survey of Symbolic Logic, 1918, Chapters I-III.
5
Begriffsschrift, 1879
6
Principia Mathematica, 1910
7
Tractatus Logico-Philosophicus, 1921
8
“Introduction to a General Theory of Elementary Propositions,” The American Journal of Mathematics, Vll. XLIII
(1921), pp. 163-185.
9
“Untersuchungen über das logische Schliessen,” Mathematische Zeitschrift, Vol. 39 (1934), pp. 176-210.
10
Symbolic Logic: An Introduction, 1952.
11
The strings of symbols generated by the formation rules are often called “well-formed formulas” or “wffs.”

3
 • A set of two punctuation marks: ‘(’ and ‘).’

• A set of five propositional operators: ‘∼,’ ‘∧,’ ‘∨,’ ‘⊃,’ and ‘≡.’

1.2 Rules of Formation for L0

In specifying our rules of formation for formulas of L0 , we use the meta-linguistic variables (or
meta-variables) ‘ϕ’, ‘ψ’, ‘γ,’ ‘θ,’ χ,’ with or without primes or positive integer subscripts, to refer
to formulas.12 Thus we will be writing things like, “Suppose γ1 , . . . , γn are sentences of L0 .” We
will refer to sets of formulas of L0 using upper-case Greek letters, such as ‘Γ.’ Further, we will
distinguish those sentences of L0 which have no structure as atomic sentences from those compound
sentences which have structure. The former are specified in the first two formation rules, the latter
in the next five.
Formation Rules of L0

1. All propositional variables are formulas of L0 .

2. ‘>’ is a formula of L0 .

3. ‘⊥’ is a formula of L0 .

4. If ϕ is a formula of L0 , then ∼ϕ is a formula of L0 .13

5. If ϕ and ψ are formulas of L0 , then (ϕ ∧ ψ) is a formula of L0 .

6. If ϕ and ψ are formulas of L0 , then (ϕ ∨ ψ) is a formula of L0 .

7. If ϕ and ψ are formulas of L0 , then (ϕ ⊃ ψ) is a formula of L0 .

8. If ϕ and ψ are formulas of L0 , then (ϕ ≡ ψ) is a formula of L0 .

9. Nothing else is a sentence of L0 .

A convention we shall employ is that outermost parentheses may be omitted from any formula
which is the end-result of the application of the formation rules. Internal parentheses are always
retained.14
The definition of a formula of L0 is recursive. Recursive definitions of formulas of formal lan-
guages begin with a listing of basic formulas. For L0 , the basic formulas are atomic formulas. The
second element of a recursive definition of a formula is a set of rules for generating more complex
formulas from simpler ones. So every compound formula is made up of atomic formulas, to which
the rules are applied initially. The results of the intitial application of the rules may then be used
as the basis for the formulation of more complex formulas by the application of formation rules to
them. The last element of the formation rules is a “closure” clause which states that no string of
symbols that is not generated by the rules is a formula.
The recursive character of the definition of a formula is of crucial importance for our under-
standing of the language. In much of what follows, we will be developing some of the meta-logic
12
These meta-variables will also be used to refer to sentences of the modal languages to follow.
13
The operators are properly part of the object language, L0 , but they are also used as names of themselves in meta-
linguistic expressions.
14
Many logic texts employ other conventions which eliminate the display of some or all internal parentheses

4
of L0 and more sophisticated languages. That is, we will be proving that the language and its
formulas have certain properties. The metalogical proofs that we give will generally rely on the
recursive character of the definition of a formula. One proof technique, mathematical induction is
very powerful, and it derives its power from the structure of the definition of a formula. We shall
soon have occasion to make use of it.

2 Semantics for Propositional Logic

Sentences of L0 have always been informally intended to abbreviate formulas of natural language,
or perhaps to stand in for the propositions which those formulas express. The operators are in-
tended to stand for grammatical particles of natural language. Thus ‘∼’ is taken to stand for the
word ‘not’ or the phrase ‘it is not the case that,” and the expressions ‘∧’ is taken to stand for the
word ‘and’ or the word ‘but.’ This relation of symbols to natural language is only a first step in
developing a system of logic. In such a system, we are primarily interested in how formulas with
certain structures are related to other formulas with their structures. More specifially, we are pri-
marily interested in the question of which formulas follow from or are logical consequences of other
formulas.
We could, as do those who produce axiomatic or natural-deduction systems, proceed directly to
identify certain formulas and/or patterns of inference as generating correct logical consequences.
Another approach, which will be used here, is first to develop in a more formal way the meanings
of the formulas of L0 , by generating a semantical system for interpreting them rigorously. This
system will give rise to a semantical version of the notion of logical consequence. Then we develop
a system based on rules of inference and an axiom system, both of which generate the relations of
logical consequence which are in a certain way equivalent.
From a formal standpoint, formulas of L0 are interpreted as designating exactly one of two
truth-values, ‘t’ and ‘f.’ Informally, we think of ‘t’ as standing for truth or being true, and ‘f’ as
standing for falsehood or being false. But in fact, the only thing our semantics requires is that
they stand for two distinct objects, perhaps even the letters ‘t’ and ‘f’ themselves! In computer
applications, ‘t’ and ‘f’ might stand for the state of a transistor being in the “on” position or in the
“off” position, respectively.15 At times in what follows, we will take ‘t’ to refer to truth and ‘f’ to
falsehood.
The propositional variables of L0 are completely open as far as which truth-value they have.
Any propositional variable may be interpreted as having the value t or the value f. The sentential
constant ‘>’ is always interpreted as having the value t, and ‘⊥’ (“falsum”) as having the value f.
The assignment of truth-values for compound formulas is based on an informal understanding of
the meanings of the respective operators. The truth-values of compound formulas (i.e., formulas
formed using formation rules 3 through 7) are uniquely determined by the truth-values of their
component parts. For this reason, the semantics is called truth-functional. A function may be
understood here as a rule which yields no more than one output for every input. In Section 2.3.2
we will prove that the semantical rules for L0 yield truth-functions.

2.1 Truth-Table Semantics

The most commonly-taught form of semantics for L0 is based on the truth-table. We will first
introduce the interpretations of the sentential operators using truth-table semantics. Later, a more
15
Hughes and Cresswell in A New Introduction to Modal Logic use the values 1 and 0 rather than t and f, respectively.

5
abstract version of the semantics will be given.
A row of a truth-table always has an ‘f’ under ‘⊥.’ It has at least one of the values t or f, but not
both, under each propositional variable pi displayed in it. In addition, it makes has at least one of
t or f, but not both, under compound formulas formed from the displayed propositional variables.
The specific truth-values are based on the following schemata.

2.1.1 Verum
The verum sentential constant has the same value whenever it appears in a truth-table.

⊥
t

If verum appears in a formula of L0 , it is always assigned t, no matter where in the formula it

appears.

2.1.2 Falsum
The falsum sentential constant has the same value whenever it appears in a truth-table.

⊥
f

If falsum appears in a formula of L0 , it is always assigned f, no matter where in the formula it

appears.

2.1.3 Negation
The tilde, ‘∼’ is intended to represent negation.16 The orthodox view of negation is that it effects a
reversal of the truth-value of the negated formulas.17 So if ϕ has the value t, then ∼ϕ has the value
f. And if ϕ has the value f, then ∼ϕ has the value t. The semantical behavior of negation is seen
graphically in the following truth-table.

ϕ ∼ϕ
t f
f t

2.1.4 Conjunction
The inverted wedge ‘∧’ is intended to symbolize conjunction.18 It is generally held that a conjunc-
tion is true just in case both its conjuncts are true. Thus, if either conjunct is false, the conjunction
is false as well.
16
Negation is often indicated by ‘¬’ and sometimes by ‘–.’ Lewis used ‘–’ in his original papers on modal logic, inter-
preting ‘∼’ as “it is impossible that.”
17
Negation is treated differently in the semantics for intuitionistic and relevant (or “relevance”) logics, but this matter
is beyond the scope of the current text.
18
Other conjunction symbols are ‘&’ and ‘.’

6
 ϕ ψ ϕ∧ψ
t t t
t f f
f t f
f f f

2.1.5 Disjunction
The wedge ‘∨’ is intended to symbolize inclusive disjunction.19 The standard view about inclusive
disjunction is that such a formula is true just in case at least one of its disjuncts is true.20 So, if
either disjunct is true, the disjunction is true as well.

ϕ ψ ϕ∨ψ
t t t
t f t
f t t
f f f

2.1.6 Material Conditional

The horseshoe ‘⊃’ is intended to represent an “if . . . then ” formula or conditional.21 The
conditional the horseshoe represents is generally known as the “material” conditional. Sometimes
it is said to represent “material implication.”22
As far as truth-values are concerned, an uncontroversial property of any conditional is that it is
false when its antecedent is true and its consequent is false. This gives us one row of a truth-table.

ϕ ψ ϕ⊃ψ
t f f

In all other cases, the conditional is said to be true. This yields a full truth-table.

ϕ ψ ϕ⊃ψ
t t t
t f f
f t t
f f t

Most logicians hold that the material conditional is the weakest form of a conditional. That is,
the material conditional fulfills the most important (if not the only) necessary condition for any
conditional, which is expressed on the second line of the truth-table. It might be held that the
material conditional does not fulfill a further necessary condition for a conditional, i.e., that there
be some connection between the content of the antecedent and that of the consequent. The truth
of the antecedent of a material conditional may not be a “condition” of the truth of the consequent.
Even if one accepts that ϕ ⊃ ψ expresses a conditional, it seems wrong to say that it expresses
any kind of logical implication. This was the objection of C.I. Lewis that led him to invent modern
19
An exclusive disjunction is one in which the disjunction is false if both disjuncts are true.
20
Some relevant logics recognize an “intensional” form of disjunction which is not truth-functional.
21
Many texts use an arrow ‘→’ instead of the horseshoe.
22
This usage goes back to Russell and Whitehead in Principia Mathematica.

7
modal logic.23 In most cases, a true material conditional is true (or formally, has the value t) in a
way that does not determine whether there is a relation of implication.
Consider two propositional variables, ‘p0 ’ and ‘p1 ,’ and the following row of a truth-table for a
material conditional formed from them:

p0 p1 p0 ⊃ p1
t t t

On this interpretation, ‘p0 ⊃ p1 ’ has the value t. Should we therefore say that ‘p0 ’ implies ‘p1 ’? There
is a reason to think that we should not. The fact that ‘p1 ’ is assigned t has nothing to do with any
fact of logic. ‘p0 ’ could just as easily be interpreted as having the value f given that ‘p0 ’ has the value
t, as is seen from the following partial truth-table:

p0 p1 p0 ⊃ p1
t f f

On this row of the truth-table, the material conditional has the value f, which might lead one
to conclude that ‘p0 ’ does not imply ‘p1 .’ This is what makes the conditional “material.” It is the
material facts about the truth-values of its components which determine whether it is true or false.
So ‘p0 ’ “materially implies” ‘p1 ’ in some cases and not in other cases, depending on what factually
is the case.
The upshot seems to be that the truth-value of a material conditional on a row of a truth-table
is not enough to establish any kind of logical implication, which ought to be independent of any
facts that could be either true or false. Perhaps what is needed to express logical implication is that
the material conditional be true on all rows of the truth-table. (With this stricter condition, the
propositional variable ‘p0 ’ does not imply the propositional variable ‘p1 .’) For any two sentences ϕ
and ψ, if ϕ ⊃ ψ is true on all rows of a truth-table, there is no row on which ϕ has the value t and ψ
has the value f.
It is worth exploring this suggestion a little further. It is easy to establish by truth-tables that
there is no row on the truth-table in which ‘p1 ’ is true and ‘p1 ⊃ p0 ’ is false. So we might wish to
say that ‘p0 ’ strictly implies ‘p1 ⊃ p0 .’ Some people understand this to mean that a true formula is
implied by every formula. That is, if ‘p0 ’ is true, ‘p0 ’ is implied by ‘p1 ’ because ‘p1 ⊃ p0 ’ is true by
the truth-table.

p0 p1 p1 ⊃ p0
t t

If ‘p1 ⊃ p0 ’ is taken to express implication, then it would be the case that a true formula ‘p0 ’ is
implied by any formula. But in fact, the only kind of “implication” we have with ‘p1 ⊃ p0 ’ is
“material implication,” as we have seen. Since ‘p0 ’ and ‘p1 ’ are propositional variables, there is a
row on a truth-table on which ‘p0 ’ is true and ‘p1 ’ is false. It is a trivial fact of the semantics that
a true formula is “materially implied” by any formula, and this is all that is contained in the claim
that a true formula is implied by every formula.
A further consideration in this connection is the fact that ‘p0 ⊃ p1 ’ is equivalent in the semantics
to ‘∼p0 ∨ p1 .’ As Lewis pointed out, this loose connection by inclusive disjunction seems hardly
strong enough to warrant the title “implication.”
23
See E. M. Curley, “The Development of Lewis’ Theory of Strict Implication,” Notre Dame Journal of Formal Logic Vol.
XVI, No. 4, pp. 517-530.

8
2.1.7 Material Biconditional
The material biconditional, expressed by the triple-bar ‘≡,’ has the same truth-table as the conjunc-
tion of two material conditionals.24 Such a biconditional is true just in case its components have
the same truth-value.

ϕ ψ ϕ≡ψ
t t t
t f f
f t f
f f t

Whether the triple-bar should be understood as a genuine biconditional or as expressing co-implication

hinges on whether the material conditional should be understood as a genuine conditional or as
expressing implication, so it will not be discussed further here.

2.2 Formal Semantics for L0

The truth-table method for determining the truth-values of formulas of L0 has been presented in an
informal way. In what follows, the semantics for L0 will be presented through the use of functional
definitions and rules. This way of approaching the semantics of L0 (and richer languages) will be
referred to here as formal semantics.
The core notion in the formal semantics for Propositional Logic is that of a truth-value assign-
ment (TVA) v, which is a complete function from the set of propositional variables of L0 to the two
values t and f. That is, each propositional variable is assigned by a TVA one and only one of the two
values. (In providing specfic truth-tables, we must always work with finite fragments of truth-value
assignments, in the sense that the representation of the assignments leaves out the values of propo-
sitional variables of L0 not found in the formulas under evaluation. An alternative to assigning
truth-values to all propositional variables is to relativize truth-value assignments to a given set of
propositional variables that are “in play,” what Richmond Thomason calls a “morphology.”25 )

2.2.1 Semantical Rules for L0

An extension (or “lifting”) of a given TVA v, v, is defined recursively by way of a set of semantical
rules for the interpretation of ⊥ and for compound formulas.26

SR-TVA v(pn ) = v(pn );

SR-> v(>) = t;

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

24
Sometimes the material biconditional is symbolized as ‘↔.’
25
Symbolic Logic: An Introduction, pp. 101ff.
26
This technique is used in Saul Kripke’s early paper on the semantics for modal logic, “Semantical Analysis of Modal
Logic I, Normal Propositional Calculi”, Zeitschrift für mathematische Logik und Grundelagen der Mathematik, 9 (1963),
pp. 67-96.

9
 
t

 if v(ϕ) = t and v(ψ) = t;
SR-∧ v(ϕ ∧ ψ) = 

f if v(ϕ) = f or v(ψ) = f.



t

 if v(ϕ) = t or v(ψ) = t;
SR-∨ v(ϕ ∨ ψ) = 

f if v(ϕ) = f and v(ψ) = f.



t

 if v(ϕ) = f or v(ψ) = t;
SR-⊃ v(ϕ ⊃ ψ) = 

f if v(ϕ) = t and v(ψ) = f.



t

 if v(ϕ) = v(ψ);
SR-≡ v(ϕ ≡ ψ) = 

f if v(ϕ) , v(ψ).



Note that the “if” clauses should be understood as metalogical biconditionals. The semantical rules
show how to generate a truth-value for a more complex formula given the truth-values of their less
complex components, but there will be times when we wish to move from the truth value of a more
complex formula to the possible truth-values of the components. For example, if v(ϕ ⊃ ψ) = t, then
it must be the case that either v(ϕ) = f or v(ψ) = t, by SR-⊃. Note further that while ‘v’ will generally
be used to represent an arbitrary extension of a TVA v, there will be cases where subscripts will be
added to distinguish different TVAs.
The definition of the assignment function taken together with the semantical rules for its exten-
sion make up a semantical system for L0 . We will call the present system “S(0)” (for “semantical
system for language L0 ”).

2.3 Semantical Properties and Relations

Having given formal definitions for the truth or falsehood of formulas of L0 , we may now turn
to some of the important semantical properties and relations of L0 formulas. These properties
and relations follow directly from the semantical rules which make up the semantical system S(0).
We will engage in metalogical reasoning in proving that the properties and relations hold. The
results of our reasoning will be metatheorems. We will also be using this kind of reasoning for other
purposes.
In most logic texts, metalogical reasoning is done informally. We will do some informal se-
mantical reasoning in this text, but in many cases, our reasoning will be of a more formal sort.
Specifically, we will employ metalogical derivations to reason to our conclusions. These derivations
will be in predicate logic, and they will appeal to the semantical rules as well as to the rules of
inference of predicate logic. In our metalogical reasoning, we will be using metalogical symbols
which correspond to English expressions.
Metalogical Symbols

Not ¬
And
V
Or
W
Only if →
If and only if ↔
For all Π
There is Σ

10
 We will also use generally accepted rules of inference and annotations for them. The style of
the proofs will for the most part parallel the way truth-tables might be used to argue for the same
results.

2.3.1 Bivalence
The first metalogical property we will investigate is bivalence: every formula of L0 has either the
value t or f.
BV: v(ϕ) = t v(ϕ) = f
W

The semantical rules stipulate that all the atomic formulas have either the value t or the value f.
So it remains to be shown that all compound formulas have one of these two values. The argument
is that if all of the components of a compound formula have a truth-value, the compound formula
itself has a truth value. This can be seen by considering all the operators. If ϕ has the value t, then
∼ϕ has the value f, and if ϕ has the value f, then ∼ϕ has the value t. The cases of the two-place
operators are the same. The semantical rules guarantee that if the formulas joined by the operators
have truth-values, then so does the compound formula. So, since all the atomic formulas have
truth-values, any compound formula generated from them has a truth-value, and any compound
formulas generated from the resulting compound formulas have truth-values, etc. This argument
is an informal application of mathematical induction.
We will now undertake part of a formal proof by mathematical induction. This proof will
introduce the reader to the technique by giving an easy result. A proof by mathematical induction
has three parts. The first is the basis step, where the property is shown to hold for some set of
objects, which in the proofs to follow will be atomic formulas. Next, an inductive hypothesis is
made. It is assumed that the property in question holds for all objects with less than a specified
level of complexity n. In the proofs to follow, n will be the number of operators in a formula. Finally,
in an inductive step it is shown that the property holds for objects with complexity n. With this, the
proof is concluded, as we have shown that the property holds for objects of any complexity. In our
metalogical proofs we will routinely omit noting that the reasoning applies to all interpretations
because the choice of I in the proof is arbitrary.
Proof of Bivalence by induction on the number n of connectives in ϕ.
Basis Step. n = 0. Case 1. ϕ is a atomic formula. Since v is a complete function, every atomic
formula has at least one of the two truth-values, so by SR-TVA, v(ϕ) = t v(ϕ) = f. Case 2. ϕ is >.
W
Since by SR->, v(>) = t, it follows that v(>) = t v(>) = f, in which case v(ϕ) = t v(ϕ) = f. Case 3.
W W
ϕ is ⊥. Since by SR-⊥, v(⊥) = f, it follows that v(⊥) = t v(⊥) = f, in which case v(ϕ) = t v(ϕ) = f.
W W

Inductive Hypothesis. Suppose that for all formulas ϕ with fewer than n operators, v(ϕ) =
v(ϕ) = f.
W
t
Induction Step. Let ϕ contain n operators. The goal is to show that v(ϕ) = t v(ϕ) = f. We must
W
consider five cases, which reflect what ϕ might be.
Case 1. ϕ is ∼ψ.

11
 1 v(ψ) = t v(ψ) = f Inductive hypothesis
W

2 v(ψ) = t Assumption
3 v(∼ψ) = f 1 SR-∼
4 v(∼ψ) = t v(∼ψ) = f 3 I
W W

5 v(ψ) = f Assumption
6 v(∼ψ) = t 5 SR-∼
7 v(∼ψ) = t v(∼ψ) = f 6 I
W W

8 v(∼ψ) = t v(∼ψ) = f 1 2-4 5-7 E

W W

Case 2. ϕ is ψ ∧ γ.

1 v(ψ) = t v(ψ) = f Inductive hypothesis

W

2 v(θ) = t v(θ) = f Inductive hypothesis

W

3 v(ψ) = t Assumption
4 v(θ) = t Assumption
5 v(ψ ∧ θ) = t 3 4 SR-∧
6 v(ψ ∧ θ) = t v(ψ ∧ θ) = f 5 I
W W

7 v(θ) = f Assumption
8 v(ψ ∧ θ) = f 7 SR-∧
9 v(ψ ∧ θ) = t v(ψ ∧ θ) = f 8 I
W W

10 v(ψ ∧ θ) = t v(ψ ∧ θ) = f 2 4-6 7-9 E

W W

11 v(ψ) = f Assumption
12 v(ψ ∧ θ) = f 11 SR-∧
13 v(ψ ∧ θ) = t v(ψ ∧ θ) = f 12 I
W W

14 v(ψ ∧ θ) = t v(ψ ∧ θ) = f 1 3-10 11-13 E

W W

Case 3. ϕ is ψ ∨ θ.
Case 4. ϕ is ψ ⊃ θ.
Case 5. ϕ is ψ ≡ θ.
The last three cases may be completed as an exercise.
So we have shown that given the assumption that formulas with fewer than n operators have
at least one truth-value, any formula with n operators, no matter how it is formed, has at least one
truth-value. And since all formulas with 0 operators have at least one truth-value, we can conclude
that no matter how complex a formula, it must have at least one truth-value.

12
2.3.2 Truth-Functionality
It was noted earlier that the operators as interpreted in S(0) are truth-functional. If one begins
with the truth-value(s) of one or two formulas, the formula resulting from the application of the
formation rule for a given operator will have only a single truth-value. In semantical system S(0),
for any formula ϕ, and any TVA v, it is not the case that ϕ has both the value t and the value f.
TF: ¬(v(ϕ) = t v(ϕ) = f).
V

We will say that because no formula has more than one truth-value, S(0) is truth-functional. A
partial proof TF will be given. The last three cases can be completed as an exercise.
Proof of Truth-Functionality by mathematical induction on the number of connectives n of ϕ.
Basis Step. n = 0. Case 1. ϕ is a atomic formula. Since v is a function, no atomic formula has
more than one truth-value on a TVA v, so by SR-TVA, ¬(v(ϕ) = t v(ϕ) = f). Case 2. ϕ is >. Since by
V
SR->, for any assignment v, v(⊥) = t and never f, it follows that ¬(v(>) = t v(>) = f). Case 3. ϕ is
V
⊥. Since by SR-⊥, for any assignment v, v(⊥) = f and never t, it follows that ¬(v(⊥) = t v(⊥) = f).
V
Inductive Hypothesis. Suppose that for all formulas ϕ with fewer than n operators, ¬(v(ϕ) =
t v(ϕ) = f).
V
Induction Step. Let ϕ contain n operators. We may distinguish five cases.
Case 1. ϕ is ∼ψ. (In the proofs of this and the next case, we will use a derived metalogical
rule which we will call “Negated Conjunction.” From a negation, one may refer the negation of the
conjunction of the negated statement with any other statement: from ¬A one may infer ¬(A B),
V
where A and B are statements in the metalanguage.)

1 ¬(v(ψ) = t v(ψ) = f) Inductive hypothesis

V

2 ¬v(ψ) = t ¬v(ψ) = f 1 De Morgan’s

W

3 v(ψ) = t v(ψ) = f BV
W

4 v(ψ) = t Assumption
5 ¬¬v(ψ) = t 4 Double Negation
6 ¬v(ψ) = f 2 5 Disjunctive Syllogism
7 ¬v(∼ψ) = t 6 SR-∼
8 ¬(v(∼ψ) = t v(∼ψ) = f) 7 Negated Conjunction
V

9 v(ψ) = f Assumption
10 ¬¬v(ψ) = f 9 Double Negation
11 ¬v(ψ) = t 2 10 Disjunctive Syllogism
12 ¬v(∼ψ) = f 11 SR-∼
13 ¬(v(∼ψ) = t ¬v(∼ψ) = f) 12 Negated Conjunction
V

14 ¬(v(∼ψ) = t ¬v(∼ψ) = f) 3 4-8 9-13 E

V W

Exercise. Prove the result by assuming the negation of the conclusion.

13
Case 2. ϕ is ψ ∧ γ.

1 ¬(v(ψ) = t v(ψ) = f) Inductive hypothesis

V

2 ¬(v(γ) = t v(γ) = f) Inductive hypothesis

V

3 v(ψ) = t v(ψ) = f BV
W

4 v(γ) = t v(γ) = f BV
W

5 ¬v(ψ) = t ¬v(ψ) = f 1 De Morgan’s

W

6 ¬v(γ) = t ¬v(γ) = f 2 De Morgan’s

W

7 v(ψ) = t Assumption
8 v(γ) = t Assumption
9 ¬¬v(γ) = t 8 Double Negation
10 ¬v(γ) = f 6 9 Disjunctive Syllogism
11 ¬¬v(ψ) = t 7 Double Negation
12 ¬v(ψ) = f 5 11 Disjunctive Syllogism
13 ¬v(ψ ∧ γ) = f 10 12 SR-∧
14 ¬(v(ψ ∧ γ) = t v(ψ ∧ γ) = f) 13 Negated Conjunction
V

15 v(γ) = f Assumption
16 ¬¬v(γ) = f 15 Double Negation
17 ¬v(γ) = t 6 16 Disjunctive Syllogism
18 ¬v(ψ ∧ γ) = t 17 SR-∧
19 ¬(v(ψ ∧ γ) = t v(ψ ∧ γ) = f) 18 Negated Conjunction
V

20 ¬(v(ψ ∧ γ) = t v(ψ ∧ γ) = f) 4 8-14 15-19 E

V W

21 v(ψ) = f Assumption
22 ¬¬v(ψ) = f 21 Double Negation
23 ¬v(ψ) = t 5 22 Disjunctive Syllogism
24 ¬v(ψ ∧ γ) = t 23 SR-∧
25 ¬(v(ψ ∧ γ) = t v(ψ ∧ γ) = f) 24 Negated Conjunction
V

26 ¬(v(ψ ∧ γ) = t v(ψ ∧ γ) = f) 3 7-20 21-25 E

V W

Case 3. ϕ is ψ ∨ γ.
Case 4. ϕ is ψ ⊃ γ.
Case 5. ϕ is ψ ≡ γ.27

14
 So we have shown that given the assumption that formulas with fewer than n operators have
no more than one truth-value, any formula with n operators, no matter how it is formed, has no
more than one truth-value. And since all formulas with 0 operators have no more than one truth-
value, we can conclude that no matter how complex a formula, it must have no more than one
truth-value. Combining Truth-Functionality with Bivalence, we get the result that all formulas of
L0 have exactly one truth-value when interpreted through the semantical machinery of S(0).
It was noted in Section 2 that the operators as interpreted in S(0) are truth-functional. The
proof of Truth-Functionality establishes this point. If one begins with a formula or two formulas
having a single truth-value, the formula resulting from their combination by way of a connective
will have only a single truth-value.

2.3.3 Entailment
The relation of entailment holds between a set of formulas and a single formula.28 The set of
formulas Γ = {γ1 , · · · , γn } entails a formula ϕ if and only if on all assignments v on which v(γ1 )=t,
and . . . and v(γn )=t, v(ϕ) = t. We symbolize entailment using the double-turnstyle ‘’ between
{γ1 , · · · , γn } and ϕ: {γ1 . . . γn }  ϕ, or Γ  ϕ.29
Entailment. Γ  ϕ if and only if for all assignments v, if γ1 ∈ Γ and . . . and γn ∈ Γ, then if
v(γ1 )=t, and . . . and v(γn )=t, then v(ϕ) =t
One typically proves that an entailment holds by assuming for an arbitrary assignment v that
the value given by v to each γi is t and then using the semantical rules or their consequences to
show that the value of ϕ is t as well. Suppose, for example, that v(ϕ) = t and v(ϕ ⊃ ψ) = t. By
Truth-Functionality, it is not the case that v(ϕ) = t and v(ϕ) = f. Therefore, it is not the case that
v(ϕ) = f. By SR-⊃, v(ϕ) = f or v(ϕ) = t. Then by Disjunctive Syllogism, v(ϕ) = t.

Semantical proof that: {ϕ, ϕ ⊃ ψ}  ψ

1 v(ϕ) = t Assumption
2 v(ϕ ⊃ ψ) = t Assumption
3 ¬(v(ϕ) = t v(ϕ) = f) TF
V

4 ¬v(ϕ) = t ¬v(ϕ) = f 3 De Morgan’s

W

5 ¬¬v(ϕ) = t 1 Double Negation

6 ¬v(ϕ) = f 4 5 Disjunctive Syllogism
7 v(ϕ ⊃ ψ) = t → (v(ϕ) = f v(ψ) = t) SR-⊃
W

8 v(ϕ) = f v(ψ) = t 2 7 Modus Ponens

W

9 v(ψ) = t 6 8 Disjunctive Syllogism

27
The reader might notice that the use of semantical rules in these derivations requires their statement as bicondition-
als. For example, in the proof of Case 3, we appeal in step 13 to the contrapositive of the claim that if ψ ∧ γ is false, then
ψ is false.
28
This technical notion must be distinguished from more vague, intuitive notions of entailment, which may be stronger
or weaker than entailment as defined here.
29
The entailment relation is defined here for any semantical system, and hence for semantical system S(0). Semantical
systems introduced in later modules will carry with them their own entailment relations.

15
2.3.4 Equivalence
Two formulas ϕ and ψ are equivalent just in case they have the same truth-value on all assign-
ments.30
Equivalence. ϕ is equivalent to ψ if and only if for all assignments v, v(ϕ) = v(ψ).
The relation of equivalence holds for formulas ϕ and ψ just in case {ϕ}  ψ and {ψ}  ϕ. Suppose
that {ϕ}  ψ and {ψ}  ϕ. Then any assignment v such that v(ϕ) = t is such that v(ψ) = t. Further,
any assignment v such that v(ψ) = t is such that v(ϕ) = t. So on any assignment v, v(ϕ) = t if and
only if v(ψ) = t. By Truth-Functionality, it follows that on any assignment v, v(ϕ) = f if and only if
v(ψ) = f. So given that {ϕ}  ψ and {ψ}  ϕ, on any assignment, ϕ and ψ have the same truth-value,
in which case they are equivalent.
Exercise. Prove the converse.

2.3.5 Validity
The limiting case of entailment, Γ  ϕ, is that in which Γ contains no formulas at all. The set of
formulas is then ∅, the empty set. If ∅  ϕ, then on all assignments v, v(ϕ) = t. In that case, we say
that ϕ is valid (here, in S(0)).31 If ϕ is S(0)-valid, we write ‘ ϕ.’
Validity.  ϕ if and only if on all assignments v, v(ϕ) = t.
For example, ‘ϕ ∨ ∼ϕ’ is valid in S(0). Consider an arbitrary assignment v. By Bivalence, v(ϕ) =
v(ϕ) = f. Now suppose v(ϕ) = t. It follows from SR-∨ that v(ϕ∨∼ϕ) = t. And suppose that v(ϕ) = f.
W
t
In that case, by SR-∼ it follows that v(∼ϕ) = t. Therefore, by SR-∨, v(ϕ ∨ ∼ϕ) = t. So in either case
we have the result that ϕ ∨ ∼ϕ has the value t , and since the choice of v was arbitrary, the result
holds for all assignments.

Semantical proof that:  ϕ ∨ ∼ϕ

1 v(ϕ) = t v(ϕ) = f BV
W

2 v(ϕ) = t Assumption
3 v(ϕ) = t → v(ϕ ∨ ∼ϕ) = t SR-∨
4 v(ϕ ∨ ∼ϕ) = t 23→E
5 v(ϕ) = f Assumption
6 v(ϕ) = f → v(∼ϕ) = t SR-∼
7 v(∼ϕ) = t 56→E
8 v(∼ϕ) = t → v(ϕ ∨ ∼ϕ) = t SR-∨
9 v(ϕ ∨ ∼ϕ) = t 78→E
10 v(ϕ ∨ ∼ϕ) = t 1 2-4 5-9 E
W

30
There is no generally accepted metalogical symbol to denote semantic equivalence.
Such a formula of L0 is often referred to as a “tautology.” It is also called a “logical truth,” though that notion seems
31

to be more generic than that of a tautology, as formulas of predicate logic and modal logic are also called logical truths.

16
 The relation of entailment and the property of validity, if they hold, can be established using
truth-tables. In effect, the formal semantical reasoning we have been using here can be mapped
onto the construction of a truth-table.

ϕ ∼ϕ ϕ ∨ ∼ϕ
t t
f t t

Steps 1 and 2 of the metalogical deduction correspond to the placing of ‘t’ and ‘f’ respectively
in the two rows under ‘ϕ.’ Step 4 corresponds to the placing of the ‘t’ under ‘ϕ ∨ ∼ϕ’ in the first
row. Step 7 corresponds to the placing of ‘t’ under ‘∼ϕ’ in the second row. Step 9 corresponds to the
placing of ‘t’ under ‘ϕ ∨ ∼ϕ’ in the second row. (There is no need to calculate a value for ∼ϕ in the
first row.) As was noted above, in general, truth-table reasoning and formal semantical reasoning
proceed in parallel.

2.3.6 Satisfiability
A formula is satisfied by an assignment v just in case v(ϕ) = t, and a set of formulas is satisfied by v
just in case all its members are satisfied by v. We may define a more general notion of satisfiability,
which applies to a set of formulas Γ. Specifically, Γ is satisfiable if and only if all its members are
jointly satisfied by at least one assignment. A formula which is not satisfiable is unsatisfiable. A set
of formulas Γ is unsatisfiable, then, if and only if there is no assignment v on which all the formulas
of Γ value t assigned by v, which by Bivalence is equivalent to saying that on all assignments, at
least one member of Γ has the value f.
Satisfiability. Γ is satisfiable if and only if there is an assignment v such that for all γi in Γ,
v(γ1 )=t, and . . . and v(γn )=t; Γ is unsatisfiable if and only if there is no assignment v such that for
all γi in Γ, v(γ1 )=t, and . . . and v(γn )=t.
For example, the set of formulas {p0 , p0 ⊃ p1 , p1 } is satisfiable, while the set {p0 , p0 ⊃ p1 , ∼p1 } is
unsatisfiable.
There is a close relation between unsatisfiability and entailment. Let Γ ∪ ∼ϕ be the union of the
set of formulas Γ and the formula ∼ϕ.32 The following metatheorem may now be proved.
Γ  ϕ if and only if Γ ∪ ∼ϕ is unsatisfiable.
Γ  ϕ if and only if for any assignment v, and any γi in Γ, if v(γi ) = t then v(ϕ) = t. By Truth-
Functionality, this holds if and only if, if v(γi ) = t then ¬(v(ϕ) = f). By SR-∼, this holds if and only
if, if v(γi ) = t, then ¬v(∼ϕ) = t. Finally, this result holds if and only if there is no assignment v such
that all the members γi of Γ have the value t and ∼ϕ has the value t, i.e. Γ ∪ ∼ϕ is unsatisfiable.

3 Natural Deduction in Propositional Logic

As was noted in Module 1, Frederick Fitch devised a set of “natural deduction” rules that show
which formulas “follow from” which sets of formulas. These rules are purely syntactical, in the
sense that they do not rely (at least explicitly) on any interpretation of the formulas to which they
are applied. The end-result of the use of the rules is a derivation of a conclusion from zero or more
premises. We will begin by using a fairly restrictive set Introduction and Elimination rules which
32
The union of two sets is the set consisting of all members of both sets.

17
produce a derivation system we call ‘D(0)’ (for “sentential derivations”). The many systems of
natural deduction do not contain rules for verum or falsum, which are not symbols in all languages
for which natural deduction systems are constructed. For each operator, as well as verum and
falsum, there will be one rule for its “introduction” and one for its “elimination.”33 We will allow
the use of two additional, derived, rules for falsum.
The first step in any Fitch-style proof is setting down one or more assumptions. A vertical line
known as the scope line is drawn next to the assumption(s) and a horizontal line is drawn below
the final assumption.

γ1 Assumption
..
.
γn Assumption

Certain rules allow the assumption to be discharged. When an assumption is discharged, the
scope line is ended. An assumption that has not, at a given point in the derivation, been discharged
is called an undischarged assumption. A formula ϕ to the right of a scope line is said to be in the
scope of the assumption (which may be ϕ itself) made at that scope line. A formula in the scope
of an assumption and not in the scope of any further assumption to its right is said to be in its
immediate scope.

3.1 Reiteration
The rule of Reiteration (R) allows any formula already arrived at in a derivation to be repeated at
any step below it in its immediate scope or in the scope of an assumption made in its immediate
scope. We shall require that the rest of the rules be used entirely within a single scope line. Thus
Reiteration may have to be applied once or more in order to allow a rule to be used. Although this
restriction is of no real value in sentential logic, it is very helpful for derivations in modal logic,
which make use of special modal scope lines.

Reiteration

ϕ Already Derived
..
.
ϕ R
ψ Assumption
..
.
ϕ R
33
Gerhard Gentzen, who originally formulated introduction/elimination rules, found philosophical significance in the
symmetry between introduction and elimination rules.

18
This schematic representation of the rule incorporates both kinds of permissible reiterations. One
need not make use of both kinds in all derivations.

3.2 Negation Rules

The rules for introducing and eliminating negation operators (∼ I and ∼ E, respectively) are quite
similar to each other. In each case, an assumption is made and a pair of formulas ψ and ∼ψ is
derived (in either order) within the immediate scope of that assumption. At that point, the scope
line ends and a negation sign is either added to or removed from the assumption (depending on
whether the introduction or elimination rule is being used).

Negation Introduction

ϕ Assumption
..
.
ψ
..
.
∼ψ
∼ϕ ∼I

Negation Elimination

∼ϕ Assumption
..
.
ψ
..
.
∼ψ
ϕ ∼E

3.3 Conjunction Rules

The conjunction rules are perhaps the most straightforward of all. A conjunction can be introduced
by conjoining two formulas (in either order) in the same scope line (by ∧ I).

19
 Conjunction Introduction

ϕ Already Derived
..
.
ψ Already Derived
..
.
ϕ∧ψ ∧I
..
.
ψ∧ϕ ∧I

A conjunction can be eliminated (using ∧ E) by writing down one of its conjuncts in the same
scope line. (Note that one need not write down both conjuncts; the two possible uses of the rule
are compressed in a single schematic representation.)

Conjunction Elimination

ϕ∧ψ Already Derived

..
.
ϕ ∧E
..
.
ψ ∧E

3.4 Disjunction Rules

The rule for introducing a disjunction (∨ I) is very straightforward. One may prepend or append
any formula as a disjunct to any given formula. (As with conjunction, only one application of the
rule need be made at any one time.)

Disjunction Introduction

ϕ Already Derived
..
.
ϕ∨ψ ∨I
..
.
ψ∨ϕ ∨I

The rule for eliminating a disjunction, also known as “simple dilemma” and as “constructive
dilemma” (∨ E), is the most complicated of all the rules. If a disjunction occurs in a derivation,
and one derives in the same scope line a single formula by assuming each of the disjuncts, one may
discharge the two assumptions and write down the derived formula.

20
 Disjunction Elimination

ϕ∨ψ Already Derived

ϕ Assumption
..
.
γ
ψ Assumption
..
.
γ
γ ∨E

3.5 Material Conditional Rules

The rule for introducing a material conditional (⊃ I) is moderately complex. One makes an assump-
tion, derives a formula from that assumption in the same scope line, discharges the assumption,
and writes down a material conditional with the assumption as the antecedent and the derived
formula as the consequent.

Material Conditional Introduction

ϕ Assumption
..
.
ψ
ϕ⊃ψ ⊃I

The elimination rule for the material conditional (⊃ E), widely known as “modus ponens,” is
quite simple. If one has derived a conditional and has also derived its antecedent within the same
scope line, then one may write down the consequent of the conditional.

Material Conditional Elimination

ϕ Already Derived
..
.
ϕ⊃ψ
..
.
ψ ⊃E

It does not matter whether the conditional or its antecedent occurs first.

21
3.6 Material Biconditional Rules
Because the material biconditional is semantically equivalent to the conjunction of two material
conditionals, the operator is introduced (using ≡ I) by using twice the same procedure that in-
troduces conditionals. The formula on one side is derived in the scope of the assumption of the
formula on the other side, and vice-versa.

Material Biconditional Introduction

ϕ Assumption
..
.
ψ
ψ Assumption
..
.
ϕ
ϕ≡ψ ≡I

The elimination rule for the material biconditional (≡ E) likewise reflects the operator’s relation
to the conditional. If a biconditional is present and either formula making it up is also present, then
the other formula can be written down.

Material Biconditional Elimination

ϕ Already Derived
..
.
ϕ≡ψ
..
.
ψ ≡E

ψ Already Derived
..
.
ϕ≡ψ
..
.
ϕ ≡E

It does not matter whether the biconditional or its component formula occurs first.

3.7 The Verum Rule

There is only one rule for verum: it may be introduced at any point in any derivation.

22
 Verum Introduction
..
.
> >I

3.8 Falsum Rules

The introduction rule for falsum (⊥ I) allows one to write down ‘⊥’ when both ϕ and ∼ϕ (in either
order) occur on the same scope line.

Falsum Introduction

ϕ Already Derived
..
.
∼ϕ Already Derived
⊥ ⊥I

Falsum is “eliminated” (using ⊥ E) in the sense that when it occurs, any formula whatsoever can be
written down in the current scope line.

Falsum Elimination

⊥ Already Derived
ϕ ⊥E

3.9 Derived Rules Using Falsum

Here we introducte two derived rules which are very useful in Predicate Logic derivations. Derived
rules are rules that could always be dispensed with in favor of the primitive rules that have just
been stated.

Negation Introduction-⊥

ϕ Assumption
..
.
⊥
∼ϕ ∼ I-⊥

Negation Elimination-⊥

∼ϕ Assumption
..
.
⊥
ϕ ∼ E-⊥

23
 We can show that the derived rules hold by providing a template for generating the same results
without their use. The illustration will be made for ∼ I-⊥ only. Suppose that one can derive ⊥ from
assumption ϕ. We show that one can subsequently discharge the assumption and write ∼ϕ.

ϕ Assumption
..
.
⊥ Derived from ϕ
ψ ⊥E
∼ψ ⊥E
∼ϕ ∼I

Exercise: Give a template for dispensing with ∼ E-⊥.

3.10 Derivations of Schemata

The derivation rules are set up to allow derivations using formulas of L0 . We will have occasion
to use derivations whose steps contain instead metalogical schemata of L0 formulas. For example,
the expression ‘ϕ ∧ ψ’ is not a formula of L0 , but it shows schematically the form of L0 formu-
las. Derivations of schemata will be given using the same annotations as for derivations using L0
formulas.

3.11 Derivational Properties and Relations

As with the semantical system S(0), the derivational system D(0) has various properties that are the
subject of metatheorems. The properties corresponding to Bivalence and Truth-Functionalality can-
not be understood until we have defined the relations and properties corresponding to entailment
(derivability) and validity (theoremhood). So we shall turn to them next.

3.11.1 Derivability
The derivability relation, {γ1 , · · · , γn } `D(0) ϕ, holds between a set of formulas {γ1 , · · · , γn } and a
formula ϕ when each member of the set is an assumption not in the scope of any other assumption,
and ϕ is a step in the immediate scope of those assumption(s). The derivability relation is the
relation of logical consequence in the derivational system D(0); hence the subscript.

3.11.2 Derivational Equivalence

Two formulas ϕ and ψ of L0 are derivationally equivalent in D(0) and only if {ϕ} `D(0) ψ and
{ψ} `D(0) ϕ. Note that derivational equivalence cannot be defined in the same manner as was
semantical equivalence (having the same truth-value on all interpretations), since the formulas are
not interpreted by the derivational system.

24
3.11.3 Theoremhood
A formula ϕ derived from no undischarged assumptions is said to be a theorem of D(0). We signify
the theoremhood of ϕ by writing ‘`D(0) ϕ.’ We will here give derivations that prove the theoremhood
of the three schemata for the axiomatic system SA described below in 5.1.

Proof that `D(0) ϕ ⊃ (ψ ⊃ ϕ)

1 ϕ Assumption
2 ψ Assumption
3 ϕ 1R
4 ψ⊃ϕ 2-3 ⊃ I
5 ϕ ⊃ (ψ ⊃ ϕ) 1-4 ⊃ I

Proof that `D(0) (ϕ ⊃ (ψ ⊃ γ)) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))

1 ϕ ⊃ (ψ ⊃ γ) Assumption
2 ϕ⊃ψ Assumption
3 ϕ ⊃ (ψ ⊃ γ) 1R
4 ϕ Assumption
5 ϕ ⊃ (ψ ⊃ γ) 3R
6 ψ⊃γ 45⊃E
7 ϕ⊃ψ 2R
8 ψ 47⊃E
9 γ 68⊃E
10 ϕ⊃γ 4-9 ⊃ I
11 (ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ) 2-10 ⊃ I
12 (ϕ ⊃ (ψ ⊃ γ)) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ)) 1-11 ⊃ I

25
 Proof that `D(0) (∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ B)

1 ∼ψ ⊃ ∼ϕ Assumption
2 ∼ψ ⊃ ϕ Assumption
3 ∼ψ ⊃ ∼ϕ 1R
4 ∼ψ 1 Assumption
5 ∼ψ ⊃ ϕ 2R
6 ∼ψ ⊃ ∼ϕ 3R
7 ∼ϕ 45⊃E
8 ϕ 46⊃E
9 ψ 3-8 ∼ E
10 (∼ψ ⊃ ϕ) ⊃ ψ 2-9 ⊃ I
11 (∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ) 1-10 ⊃ I

3.12 Excluded Middle

Corresponding to the semantical property of Bivalence described in Section 2.3.1 above is a formula-
schema known as the Excluded Middle, ϕ ∨ ∼ϕ is a theorem of D(0). The derivational system does
not have a way of expressing falsehood directly. Intuitively, we take every step of a derivation to
be an assertion of truth. Falsehood can, however, be expressed indirectly. To assert ∼ϕ is, from
the point of view of S(0), tantamount to asserting that ϕ is false. So the semantical property of
formulas, that each one is either true or false, is expressed in L0 formulas of the form ϕ ∨ ∼ϕ. It
may be derived as follows.

1 ∼(ϕ ∨ ∼ϕ) Assumption

2 ϕ Assumption
3 ϕ ∨ ∼ϕ 2∨I
4 ∼(ϕ ∨ ∼ϕ) 1R
5 ∼ϕ 2-4 ∼ I
6 ϕ ∨ ∼ϕ 5∨I
7 ϕ ∨ ∼ϕ 1-6 ∼ E

3.13 Non-Contradiction
The counterpart to Truth-Functionality described in Section 2.3.2 above is the formula-schema
known as Non-Contradiction, ∼(ϕ ∧ ∼ϕ), which is a theorem-schema of D(0). Once again, reference
to falsehood is made through negation. From the standpoint of S(0), formulas of this sort assert
that it is not the case that ϕ and ∼ϕ are both true, in which case ϕ is not both true and false.

26
 1 ϕ ∧ ∼ϕ Assumption
2 ϕ 1∧E
3 ∼ϕ 1∧E
4 ∼(ϕ ∧ ∼ϕ) 1-3 ∼ I
The reader will not fail to notice how easy these derivations are compared with the semantical
derivations made in the previous section.

3.14 Consistency
Corresponding to satisfiability described in Section 2.3.6 above is the notion of consistency. A set of
formulas Γ of L0 is consistent, if and only if it is not the case that there is a formula ϕ of L0 such
that Γ `D(0) ϕ and Γ `D(0) ∼ϕ.34 Any set of formulas which is not consistent is inconsistent. We can
prove the following metatheorem:
Γ `D(0) ϕ if and only if Γ ∪ ∼ϕ is inconsistent.
Suppose Γ `D(0) ϕ. Then there is a derivation of ϕ from the members of Γ as assumptions.
Adding an assumption does not affect the proof, so Γ ∪ ∼ϕ `D(0) ϕ. By the rule of Reiteration,
Γ ∪ ∼ϕ `D(0) ∼ϕ. Therefore, Γ ∪ ∼ϕ is inconsistent.
This reasoning can be illustrated by the following derivation-schemata. First, suppose that
Γ `D(0) ϕ.

γ1 Assumption
...
γn Assumption
...
ϕ By Hypothesis
We add ∼ϕ as a further assumption.

γ1 Assumption
...
γn Assumption
∼ϕ Assumption
...
ϕ By Hypothesis
We derive ∼ϕ from Γ ∪ ∼ϕ.
34
Equivalently, if and only if there is a formula ϕ such that Γ 0D(0) ϕ.

27
 γ1 Assumption
...
γn Assumption
∼ϕ Assumption
...
∼ϕ Reiteration
Now suppose Γ ∪ ∼ϕ is inconsistent. Suppose further that the members of Γ are the outermost
assumptions in a derivation. Assume ∼ϕ and reiterate all the members of Γ. Because Γ ∪ ∼ϕ is
inconsistent, we may then derive a pair of formulas ψ and ∼ψ. The assumption of ∼ϕ may be
discharged and ϕ written down as the last step in the derivation, in which case Γ `D(0) ϕ.

γ1 Assumption
...
γn Assumption
∼ϕ Assumption
γ1 Reiteration
...
γn Reiteration
...
ψ From inconsistency of Γ ∪ ∼ϕ
...
∼ψ From inconsistency of Γ ∪ ∼ϕ
ϕ ∼E
The reader will note that there is a structural similarity between this result and the correspond-
ing result for semantical inconsistency.

4 Soundness and Completeness

Thus far, our metalogical results have been confined to either the semantical system S(0) or the
derivational system D(0). There are further metatheorems which concern the relation between
the two systems. The first is the soundness of D(0). A derivational system is sound relative to a
semantical system just in case every relation of derivability is a relation of entailment. With respect
to our present systems, if {γ1 , . . . , γn } `D(0) ϕ, then {γ1 , · · · , γn }  ϕ. A derivational system is complete
relative to a semantical system just in case all entailments are derivable. Applied to Propositional
Logic, this means that if {γ1 , · · · , γn }  ϕ, then {γ1 , . . . , γn } `D(0) ϕ. Because theoremhood and
validity are degenerate cases of derivability and semantic entailment, respectively, we can define

28
more restricted (or “weaker”) senses of soundness and completeness. If D(0) is sound relative to
S(0), then `D(0) ϕ, then  ϕ. If D(0) is complete relative to S(0), then if  ϕ, then `D(0) ϕ.

4.1 Soundness
We may think of the relation of entailment as one which excludes the possibility of a set of formulas
being true and its consequence being false. If a set of derivation rules is such that all derivations
are entailments, then the derivation rules do not allow the inference of a false consequence from
true premises. This accords with an intuitive sense in which the system of derivation rules is sound.
The soundness of D(0) relative to S(0) will not be proved here, but we can give an indication of
how the proof goes. The main work of the proof is to establish a lemma: at any point in a derivation,
if all the undischarged assumptions {γ1 , . . . , γn } up to that point are true on an interpretation, then
the last formula ϕ is true on that interpretation as well.35 The lemma applies to the limiting
case where ϕ is in the immediate scope of the outermost assumptions {γ1 , . . . , γn }, in which case
{γ1 , . . . , γn } `D(0) ϕ.
The lemma is proved by mathematical induction on the length of the derivation. Suppose n
steps of a derivation have been constructed. The proof shows that if the formulas needed for the
application of the rule at the nth step are true given that the undischarged assumptions are true,
then the result of the application of the rule is true as well. As an illustration, consider Conjunction
Elimination. Suppose that {γ1 , . . . , γn } are all the undischarged assumptions, and that ϕ ∧ ψ is a
formula that lies in their scope. By the inductive hypothesis, if {γ1 , . . . , γn } are all true, then ϕ ∧ ψ
is true. But we have shown already that if this is the case, then ϕ is true and ψ is true. So, if
{γ1 , . . . , γn } are all true, so are ϕ and ψ. Reasoning for the rules involving assumptions is more
complicated and will not be described here.
Another way to look at soundness is by noticing the parallel between the derivational rules
and the semantical reasoning used to establish entailment. Here we propose without proof that
each of the rules of D(0) can be mapped onto a corresponding argument that uses only semantical
rules. For example, the rule of Conjunction Elimination corresponds to the rule that if the value of
a conjunction is t, then the values of each of its conjuncts are t.

Conjunction Elimination

ϕ∧ψ Already Derived

...
ϕ ∧E

Semantical proof that: ϕ ∧ ψ  S I ϕ

v(ϕ ∧ ψ) = t Assumption
...
v(ϕ ∧ ψ) = t → v(ϕ) = t SR-∧
v(ϕ) = t Modus Ponens
35
A lemma is is a principal result used in proving a metatheorem.

29
 We can think of the derivational rule as mirroring the use of semantical rules, but in a “cleaner”
or “less-cluttered” way. The derivational system dispenses with references to truth-values because
it implicitly takes all the truth-values to be t. Falsehood of a formula ϕ is displayed by the using its
negation, ∼ϕ. Here is another example to illustrate the point.

Negation Introduction

ϕ Assumption
...
ψ
...
∼ψ
∼ϕ ∼I

Semantical proof of the Soundness of ∼ Introduction

v(ϕ) = t Assumption
...
v(ψ) = t Semantical Reasoning
...
v(∼ ψ) = t Semantical Reasoning
v(ψ) = f SR-∼
v(ψ) = t v(ψ) = f I
V V

¬(v(ψ) = t v(ψ) = f) TF
V

¬v(ϕ) = t ¬I
v(ϕ) = f BV
v(∼ϕ) = t SR-∼

We will exploit the parallelism between semantical rules and derivational rules when we con-
struct derivational rules for systems of modal logic.

4.2 Completeness
If we think of the relation of entailment in S(0) as exhaustive of the logical consequences holding
among formulas of L0 , then the derivational system is, intuitively, complete if it can reproduce all
those consequences. Proofs of completeness are a good deal more difficult than proofs of sound-
ness.
There are many strategies for completeness proofs. The most intuitively satisfying proofs are
constructive, showing how for any entailment a derivation can be constructed. These proofs are
complicated in detail, and will be discussed further below.

30
 Another proof-strategy is non-constructive. In such proofs, all that is shown is that for every
entailment there must be a derivation, but what that derivation looks like is completely unspecified.
This type of proof was first advanced by Leon Henkin, and so it has come to be known as the
“Henkin-style” proof-strategy for completeness.36 One advantage of the Henkin approach is that it
is easily adapted to modal logics.
Here we will only sketch the strategy for proving completeness in the manner of Henkin. The
key move is to show how to begin with a consistent set of formulas Γ and then build a “maximal”
consistent set of formulas Γ∗ , of which the original set is a subset.37 “Lindenbaum’s Lemma,” which
states that every consistent set is a subset of a maximal consistent set is then proved, using the
specific characteristics of D(0) derivations, which determine what is and what is not a member of
a maximal consistent set.
It is next shown that there is an assignment according to which a given formula is a member of
Γ if and only if that formula has the value t on that assignment. This is done by using a special
∗

recipe for constructing an assignment which guarantees that all and only the members of Γ∗ have
the value t, and hence that Γ∗ is satisfiable. The recipe is easy: assign to all the atomic formulas in
Γ∗ the value t and to all the atomic formulas not in Γ∗ the value f. It is then proved by mathematical
induction that on such an assignment, all and only members of Γ∗ have the value t. This proof
also relies on the specific features of D(0) derivations, which determine what is and what is not a
member of Γ∗ . Finally, if Γ∗ is such that there is an assignment on which all its formulas have the
value t, then all its subsets have the value t as well, by simple reasoning in predicate meta-logic.
We sketch the proof-strategy in the form of a metalogical derivation.

Sketch of Proof that: If Γ is consistent, then Γ is satisfiable.

1 Γ is consistent Assumption
2 Γ ⊆ a maximal consistent set Γ∗ 1 “Lindenbaum’s Lemma”
3 (Σv)(Πϕ)(ϕ ∈ Γ∗ ↔ v(ϕ) = t) By Mathematical Induction
4 Γ∗ is satisfiable 3 Definition of ‘satisfiable’
5 Γ is satisfiable 2 4 Definition of ‘satisfiable’
6 Γ is consistent → Γ is satisfiable 1-5 → I

Now we may apply this strategy to a set Γ ∪ ∼ϕ, which we assume to be consistent. By Linden-
baum’s Lemma, it is a subset of a maximal consistent set Γ∗ . We apply the assignment that makes
all the formulas in Γ∗ have the value t, in which case all the formulas in Γ ∪ ∼ϕ have the value t;
that is, Γ ∪ ∼ϕ is satisfiable.
36
“The Completeness of the First-Order Functional Calculus,” Journal of Symbolic Logic Vol. 14 (1949), pp. 42-50.
37
We use ‘⊆’ to denote the subset relation: Θ ⊆ Γ if and only if every member of Θ is a member of Γ.

31
 Sketch of Proof that: If Γ  ϕ, then Γ `D(0) ϕ.

1 Γϕ Assumption
2 Γ  ϕ → Γ ∪ ∼ϕ is unsatisfiable Proved in 2.3.6
3 Γ ∪ ∼ϕ is unsatisfiable 12→E
4 Γ ∪ ∼ϕ is consistent → Γ ∪ ∼ϕ is satisfiable Instance of above result
5 Γ ∪ ∼ϕ is inconsistent 3 4 Modus Tollens
6 Γ ∪ ∼ϕ is inconsistent → Γ `D(0) ϕ Proved in 3.14
7 Γ `D(0) ϕ 56→E

This general strategy can be adapted to any pair of systems for which inconsistency and un-
satisfiability can be defined in the way they have been defined here. This will include the modal
logics that will be studied in later modules. The means of implementing the strategy in the proof of
Lindenbaum’s Lemma and of the equivalence between the consistency and satisfiability of maximal
consistent sets will depend on the specific character of the systems studied.
The un-intuitive character of this mode of proof is apparent. A more intuitively satisfying ap-
proach would be to exploit the striking parallelism between semantical reasoning and derivations,
as was done with soundness. However, there is a difficulty in so doing. It is not the case that
any semantical derivation can be converted into a derivation in D(0) by making the appropriate
conversion from formulas having the value f to the negation of those formulas. Here is an earlier
example of semantical reasoning, followed by a corresponding derivation.

Semantical proof that:  ϕ ∨ ∼ϕ

1 v(ϕ) = t v(ϕ) = f BV
W

2 v(ϕ) = t Assumption
3 v(ϕ) = t → v(ϕ ∨ ∼ϕ) = t SR-∨
4 v(ϕ ∨ ∼ϕ) = t 23→I
5 v(ϕ) = f Assumption
6 v(ϕ) = f → v(∼ϕ) = t SR-∼
7 v(∼ϕ) = t 56→E
8 v(∼ϕ) = t → v(ϕ ∨ ∼ϕ) = t SR-∨
9 v(ϕ ∨ ∼ϕ) = t 78→I
10 v(ϕ ∨ ∼ϕ) = t 1 2-4 5-9 E
W

There is nothing in the derivational system that corresponds to the assertion of the rule of
Bivalence. The closest we can come would be the formula ϕ ∨ ∼ϕ, which is the very thing we are
trying to derive. The derivation of that schema in D(0) proceeds as follows.

32
 Proof that: `D(0) ϕ ∨ ∼ϕ

1 ∼(ϕ ∨ ∼ϕ) Assumption

2 ϕ Assumption
3 ϕ ∨ ∼ϕ 2∨I
4 ∼(ϕ ∨ ∼ϕ) 1R
5 ∼ϕ 2-4 ∼ I
6 ϕ ∨ ∼ϕ 5∨I
7 ϕ ∨ ∼ϕ 1-6 ∼ E
Now this derivation can be seen as a conversion of a metalogical semantical derivation.
Another semantical proof that:  ϕ ∨ ∼ϕ

1 ¬(v(ϕ ∨ ∼ϕ) = t) Assumption

2 v(ϕ) = t Assumption
3 v(ϕ) = t → v(ϕ ∨ ∼ϕ) = t SR-∨
4 v(ϕ ∨ ∼ϕ) = t 23→E
5 ¬(v(ϕ ∨ ∼ϕ) = t) 1 Reiteration
6 ¬v(ϕ) = t 2-5 ¬ I
7 v(ϕ) = t v(ϕ) = f BV
W

8 v(ϕ) = f 6 7 Disjunctive Syllogism

9 v(ϕ) = f → v(∼ϕ) = t SR-∼
10 v(∼ϕ) = t 89→E
11 v(∼ϕ) = t → v(ϕ ∨ ∼ϕ) = t SR-∨
12 v(ϕ ∨ ∼ϕ) = t 10 11 → E
13 v(ϕ ∨ ∼ϕ) = t 1-12 ¬ E
To try to show completeness by exploiting the parallism between the two pieces of reasoning, we
would have to show how to set up all our semantical derivations in a way that would allow them
to be converted to derivations in D(0). There may be a way of making the derivations proving
entailment conform precisely to how a truth-table is constructed. This is a task which will not be
undertaken here.

5 Axiomatic Formulation of Propositional Logic

The earliest formulations of sentential logic were axiomatic. In an axiom system, there is a set of
formulas called axioms and a set of rules of inference are specified. A theorem is defined as an
axiom or what follows from axioms by way of rules of inference. The primary aim of axiomatic
systems is to generate an acceptable list of theorems.

33
5.1 The Axiom System A(0)
The following is one among many possible axiomatizations of Propositional Logic. This version
(which we will call “A(0)”) describes an axiom system using axiom schemata. Rather than speci-
fying axioms, the system shows through the schemata the form that the axioms have. The axioms
themselves are formulas of L0 with the displayed form. Since there are infinitely many formulas of
L0 , there are infinitely many axioms in this system.38

Axioms

A(0) 1 `A(0) ϕ ⊃ (ψ ⊃ ϕ)

A(0) 2 `A(0) (ϕ ⊃ (ψ ⊃ γ)) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))

A(0) 3 `A(0) (∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)

Rule of Inference

Modus Ponens From `A(0) ϕ and `A(0) ϕ ⊃ ψ, infer `A(0) ψ.

None of the axioms of A(0) contains any of the operators ‘∧,’ ‘∨,’ or ‘≡.’ These operators may be
defined contextually.

ϕ ∧ ψ =Df ∼(ϕ ⊃ ∼ψ)

ϕ ∨ ψ =Df ∼ϕ ⊃ ψ

ϕ ≡ ψ =Df ∼((ϕ ⊃ ψ) ⊃ ∼(ψ ⊃ ϕ))

Theorems of A(0) may be converted to definitionally equivalent theorems by making the appro-
priate substitutions. For example, `A(0) ∼ϕ ⊃ ∼ϕ. Based on the definition (and Double Negation,
which holds in A(0)), we may assert that `A(0) ϕ ∨ ∼ϕ.
The following is an example of an axiomatic proof that `A(0) ϕ ⊃ ϕ. (It is easily adapted to prove
`A(0) ∼ϕ ⊃ ∼ϕ.) The theorem is very simple, yet the proof is quite cumbersome.

Axiomatic Proof that: A(0) ϕ ⊃ ϕ

1 (ϕ ⊃ ((ϕ ⊃ ϕ) ⊃ ϕ)) ⊃ ((ϕ ⊃ (ϕ ⊃ ϕ)) ⊃ (ϕ ⊃ ϕ)) A(0) 2

2 ϕ ⊃ ((ϕ ⊃ ϕ) ⊃ ϕ) A(0) 1
3 (ϕ ⊃ (ϕ ⊃ ϕ)) ⊃ (ϕ ⊃ ϕ) 1 2 Modus Ponens
4 ϕ ⊃ (ϕ ⊃ ϕ) A(0) 1
5 ϕ⊃ϕ 3 4 Modus Ponens

The chief advantage of the axiomatic formulation of a system for L0 is that it facilitates proofs
of soundness, completeness, and other metalogical properties and relations. Here, we will use
the axiom system to illustrate the way in which semantical reasoning and derivations work in the
context of Propositional Logic.
38
Alternatively, one could specify three axioms (for this formulation) using formulas of L0 and add a rule of “uniform
substitution” which generates theorems by substituting other formulas for those in the axioms.

34
5.2 A(0) and S(0)
The axiom system A(0) can be proved to be sound and complete relative to the semantical system
S(0). Here we will only prove the weak form of soundness, that all theorems of A(0) are valid in
S(0). The completeness theorem follows the Henkin method outlined above, which works most
easily with the axiomatic formulation.
The proof of weak soundness is by mathematical induction on the number n of steps in the
proof of the theoremhood of an arbitrary formula ϕ.
Basis Step: n = 1.
If there is only one step in the proof of the theoremhood of ϕ, then ϕ is an axiom of A(0). So
we will prove that any axiom schema of A(0) has as instances only formulas valid in S(0).

Semantical proof of Validity of A(0) 1: A(0) ϕ ⊃ (ψ ⊃ ϕ)

1 v(ϕ) = t v(ϕ) = f BV
W

2 v(ϕ) = t Assumption
3 v(ϕ) = t → v(ψ ⊃ ϕ) = t SR-⊃
4 v(ψ ⊃ ϕ) = t 23→E
5 v(ψ ⊃ ϕ) = t → v(ϕ ⊃ (ψ ⊃ ϕ)) = t SR-⊃
6 v(ϕ ⊃ (ψ ⊃ ϕ)) = t 45→E
7 v(ϕ) = f Assumption
8 v(ϕ) = f → v(ϕ ⊃ (ψ ⊃ ϕ)) = t SR-⊃
9 v(ϕ ⊃ (ψ ⊃ ϕ)) = t 78→E
10 v(ϕ ⊃ (ψ ⊃ ϕ)) = t 1 2-6 7-9 E
W

The moves made in the proof can be illustrated through truth-tables. Only the values needed
to get the result are listed.

ϕ ψ⊃ϕ ϕ ⊃ (ψ ⊃ ϕ)
t t t

ϕ ϕ ⊃ (ψ ⊃ ϕ)
f t

35
 Semantical proof of Validity of A(0) 2: A(0) (ϕ ⊃ (ψ ⊃ γ) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))

1 v(ϕ) = t v(ϕ) = f BV
W

2 v(ϕ) = t Assumption
3 v(γ) = t v(γ) = f BV
W

4 v(γ) = t Assumption
5 v(ϕ ⊃ γ) = t 4 SR-⊃
6 v((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ)) = t 5 SR-⊃
7 v((ϕ ⊃ (ψ ⊃ γ)) ⊃ (ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ)) = t 6 SR-⊃
8 v(γ) = f Assumption
9 v(ψ) = t v(ψ) = f BV
W

10 v(ψ) = t Assumption
11 v(ψ ⊃ γ) = f 8 10 SR-⊃
12 v(ϕ ⊃ (ψ ⊃ γ)) = f 11 SR-⊃
13 v((ϕ ⊃ (ψ ⊃ γ)) ⊃ (ϕ ⊃ (ψ ⊃ γ))) = t 12 SR-⊃
14 v(ψ) = f Assumption
15 v(ϕ ⊃ ψ) = f 2 14 SR-⊃
16 v((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ)) = t 15 SR-⊃
17 v((ϕ ⊃ (ψ ⊃ γ)) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))) = t 16 SR-⊃
18 v((ϕ ⊃ (ψ ⊃ γ)) ⊃ (ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ) = t 9 10-13 14-17 E
W

19 v((ϕ ⊃ (ψ ⊃ γ)) ⊃ (ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ) = t 3 4-7 8-18 E

W

20 v(ϕ) = f Assumption
21 v(ϕ ⊃ γ) = t 20 SR-⊃
22 v((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ)) = t 21 SR-⊃
23 v((ϕ ⊃ (ψ ⊃ γ)) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))) = t 22 SR-⊃
24 v((ϕ ⊃ (ψ ⊃ γ)) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))) = t 1 2-19 20-23 E
W

Again, we illustrate the moves in the proof using truth-tables.

ϕ γ ϕ⊃γ (ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ) (ϕ ⊃ (ψ ⊃ γ) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))

t t t t t

ϕ ψ γ ψ⊃γ ϕ ⊃ (ψ ⊃ γ) (ϕ ⊃ (ψ ⊃ γ) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))

t t f f f t

ϕ ψ ϕ⊃ψ (ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ) (ϕ ⊃ (ψ ⊃ γ) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))

t f f t t

36
 ϕ ϕ⊃γ (ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ) (ϕ ⊃ (ψ ⊃ γ) ⊃ ((ϕ ⊃ ψ) ⊃ (ϕ ⊃ γ))
f t t t

Semantical proof of Validity of A(0) 3: A(0) (∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)

1 v(ψ) = t v(ψ) = f BV
W

2 v(ψ) = t Assumption
3 v(((∼ψ ⊃ ϕ) ⊃ ψ) = t 2 SR-⊃
4 v((∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)) = t 3 SR-⊃
5 v(ψ) = f Assumption
6 v(∼ψ) = t 5 SR-∼
7 v(ϕ) = t v(ϕ) = f BV
W

8 v(ϕ) = t Assumption
9 v(∼ϕ) = f 8 SR-∼
10 v(∼ψ ⊃ ∼ϕ) = f 6 9 SR-⊃
11 v((∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)) = t 10 SR-⊃
12 v(ϕ) = f Assumption
13 v(∼ψ ⊃ ϕ) = f 12 SR-⊃
14 v((∼ψ ⊃ ϕ) ⊃ ψ) = t 13 SR-⊃
15 v((∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)) = t 14 SR-⊃
16 v((∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)) = t 7 8-11 12-15 E
W

17 v((∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)) = t 1 2-4 5-16 E

W

Again, the reasoning is illustrated with truth-tables.

ψ (∼ψ ⊃ ϕ) ⊃ ψ (∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)
t t t

ϕ ψ ∼ϕ ∼ψ ∼ψ ⊃ ∼ϕ (∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)

t f f t f t

ϕ ψ ∼ψ ∼ψ ⊃ ϕ (∼ψ ⊃ ϕ) ⊃ ψ (∼ψ ⊃ ∼ϕ) ⊃ ((∼ψ ⊃ ϕ) ⊃ ψ)

f f t f t t

Inductive Hypothesis. Suppose that the formulas proved in all steps before n are valid. It will
be shown that the formula produced by step n itself is valid.
Induction Step. Because we have only one inference rule, Modus Ponens, all that is required,
given the Inductive Hypothesis, is to show that if  ϕ and  ϕ ⊃ ψ, then  ψ. Suppose this is the case.
Then ϕ has the value t on all assignments, as does ϕ ⊃ ψ. By SR-⊃, it follows by easy reasoning that
ψ has the value t on all assignments, i.e.,  ϕ.
This completes the proof of the Induction Step, and we can assert that no matter what the
length of the proof of its theoremhood, if ϕ is a theorem, then ϕ is valid.

37
5.3 A(0) and D(0)
Earlier, we derived all three axioms of A(0) as theorems of D(0). So we can say that all axioms of
A(0) are theorems of D(0). Further, the use of the Rule of Modus Ponens can be simulated in a
derivation of the following form. Given that ` ϕ and ` ϕ ⊃ ψ, ϕ and ϕ ⊃ ψ can be derived from no
assumptions. We can then use ⊃ Elimination to derive ψ. Since the two steps used to derive ψ do
not depend on any assumptions, ψ does not either.

ϕ Derived from no assumptions

...
ϕ⊃ψ Derived from no assumptions
...
ψ ⊃ E (Derived from no assumptions)
These results together show that the derivation system contains the axiom system. That is,
anything that can be proved using the axioms and Modus Ponens can be proved using natural
deduction. Proving the converse, that the axiom system contains the derivation system, is more
difficult. One would have to show that any theorem produced by natural deduction can be proved
axiomatically. This involves giving a method to convert derivations to axiomatic proofs.39
A final consideration is the notion of derivability, as opposed to mere theoremhood, in an axiom
system. We can define a notion of derivability in the system in the following way. Suppose Γ is a
set of formulas of L0 and ϕ is a formula of L0 . A deduction of ϕ from Γ is a sequence of formulas,
with ϕ as the last member, each of which is either an axiom of A(0), a member of Γ, or the result of
the use of Modus Ponens on two previous steps. We can then write: Γ `A(0) ϕ.
This concludes our survey of non-modal Propositional Logic. We are now ready to move on to
study Modal Propositional Logic.

39
See, for example, Richmond H. Thomason, Symbolic Logic: An Introduction, Chapter V.

38