NOTES ON

FORMAL LOGIC
BY

JOHN L. BELL
(COMPILED & EMBELLISHED
BY ROBERT CLIFTON)

Logic is the beginning of wisdom, not the end.

Mr. Spock

On the contrary, I find nothing in logistic but shackles. It does not help us at all
in the direction of conciseness, far from it; and if it requires 27 equations to
establish that 1 is a number, how many will it require to demonstrate a real
theorem?
Henri Poincar

CONTENTS
I. PROPOSITIONAL LOGIC
1. Statements, Arguments and Counterexamples
2. Truth Tables and Testing Validity
3. Tautologies, Contradictions and Satisfiability
4. Exercises
II. EQUIVALENCE
1. Equivalence and Bi-implication
2. Expressive Completeness
3. Arithmetical Representation of Statements and Logical Operations
4. Venn Diagrams
5. Exercises
III. TRUTH TREES
1. Introduction to Trees
2. The Tree Rules
3. Tree Test for Validity
4. Further Applications of the Tree Method
5. Correctness and Adequacy of the Tree Method
6. Exercises
IV. PREDICATE LOGIC
1. Predicates, Relations and Quantifiers
2. Tree Rules for Quantifiers
3. Identity
4. Validity and Interpretations
5. Many-Sorted Logic
6. Functions
7. Exercises
V. THE PROPOSITIONAL CALCULUS
1. Axioms
2. Deductions
3. Soundness
4. Completeness
5. Exercises
APPENDIX A - LOGIC AND THE FOUNDATIONS OF MATHEMATICS
APPENDIX B - HILBERTS PROGRAM & GDELS THEOREMS
APPENDIX C - SOLUTIONS TO EXERCISES

I. PROPOSITIONAL LOGIC
1. Statements, Arguments and Counterexamples
An inference or argument is any list of statements divided into premises and
conclusions. We shall confine our attention to inferences with just one conclusion, for
example,
1. Either this man's dead, or my watch has stopped.
2. This man is not dead.
3. Therefore, my watch has stopped.
Here statements 1 and 2 are premises, and 3 the conclusion.
We shall assume that the constituent statements of inferences are assertive in the
sense that each can be assigned either of two truth values true (t) or false (f) as the
case may be. (An example of a statement that would not fall into this category is
Would you please go to the store.) Granted this, it is natural to declare an inference
valid if its conclusion is true in any case in which its premises are true. (Thus, on the
basis of our usual grasp of the meaning of the terms "or" and "not", the inference above
would count as valid.) A counterexample to an inference is a case in which all its
premises are true but its conclusion is false. Thus an inference is valid provided it has no
counterexamples, and invalid if it has counterexamples.

As an example of an invalid

inference, consider the following:

Either this man's dead, or my watch has stopped.
This man is dead.
Therefore, my watch has stopped.
This inference is invalid, because the situation in which this man is dead and my watch
hasn't stopped constitutes a counterexample.

5
To investigate the validity of inferences we need first to consider how their
constituent statements are formed, and how these are then to be assigned truth values.
As the basic ingredients from which we shall fashion all such statements we shall take
simple declarative sentences of the kind "It is raining", "The cranes are flying", etc. Such
statements will be called elementary statements: we shall assume that elementary
statements can be assigned truth values arbitrarily and entirely independently of one
another.
From elementary statements we obtain compound statements by applying the
syntactical operations "and", "or", "not", "if...then". (In this way we obtain, e.g. statements
such as "It is raining and the cranes are flying", "If it is raining, the cranes are not flying",
etc.) These operations are truth functional in the sense that the truth value of any
compound statement built up from them is unambiguously determined by the truth
values of its constitutive elementary statements.
We shall use capital letters A,B,C,... to denote elementary statements, and
symbols
for "and" (conjunction),
for "or" (disjunction),
for "not" (negation or denial),
for "if...then" (implication).
The symbols A,B,C,... are called statement letters, and the symbols , ,, , logical
operators. "A B" is sometimes read "A implies B", or "B if A", or "A only if B". Also ""
has the inclusive sense of "and\or"; it may also be understood as "unless".
Using these symbols, statements are obtained by starting with the statement
letters which of course count as the simplest kind of statement and applying the
logical operators , , , to these, using parentheses and brackets as necessary to
eliminate ambiguity. So, for example, from the statement letters A, B, C, we obtain the

6
compound statements A B, (A B) C, (A B) C, [(A B) C], etc. We shall
use letters p,q,r,... to denote arbitrary statements (elementary or compound).
Formally, a statement, or propositional statement, may now be defined by means of
the following rules of formation:
1. Any statement letter is a statement.
2. If p and q are statements, so are (p q), (p q), (p q), (p).
Here p,q are called the conjuncts in the conjunction (p q) and the disjuncts in the
disjunction (p q). (Thus, for example, in the disjunction (((A) B) C) the disjuncts
are ((A) B) and C.) Also p is called the antecedent and q the consequent in the
implication (p q).
Strictly speaking, any sequence of statement letters and logical operators that
cannot be generated by repeated application of rules 1 and 2 above will not, for us,
count as a statement. For example, A B is obviously not allowed (If not A, then
or B is gibberish), while ((A) B) is allowed. However, in the interests of brevity
(and keeping our sanity!) we shall bend the rules slightly and feel free to omit
parentheses in statements when no ambiguity is likely to result. For example, instead of
((A) B) we shall write simply A B. For obviously theres no need for the outside
brackets if this statement is meant to be a complete statement and not a component of
some larger statement (like one disjunct of a larger disjunction). Also, by replacing (A)
by A we are agreeing to understand the not operator as only acting upon the
statement letter A and not upon the larger statement A B. If, instead, we wanted to
symbolize the statement Neither A nor B then we would have to write (A B) so
that the scope of the operator covers the entire disjunction. As a general rule of thumb,
insert parentheses only when it will not otherwise be clear to your readers what
statement is being negated, or what statements are being disjoined, conjoined, etc. (For

7
example, A B C is not clear, while A (B C) is; A B C is not clear, (A B)
C is; etc.)

2. Truth Tables and Testing Validity

The rules for computing the truth values of compound statements are as follows.
A B is true if A and B are both true, and false if at least one of A and B is false.
A B is true if at least one of A and B is true, and false if both A and B are false.
A is true if A is false, and false if A is true.
A B is false when A is true and B is false, but true in all other cases.
The least intuitive of these rules is the last one. The idea here is that we want a
statement of the form p q to be false exactly when the truth values of p and q
constitute a counterexample to the validity of the inference from p to q, that is, when p
is true and q is false. In all other cases, p q shall be declared true.
The above rules for computing truth values may be summed up in the form of
truth tables.

For those to whom this seems odd, consider the following down-to-earth example to make it more
palatable. Suppose that, upon leaving for work in the morning, I promise my wife If I go to the store
(p), then I will buy some milk (q). When I arrive back from work in the evening, she asks me whether I
picked up any milk, and I say No. Did I break my promise? That is: should p q be declared false in
this case; a case where, in fact, both p and q turned out to be false? Surely not: I would only have broken
the promise if in fact I did go to the store but did not buy any milk (due to an oversight, or lack of money,
or what have you). Nevertheless, it must be admitted that our definition of the truth-conditions for the
so-called material conditional fails to do justice to all our intuitions about how the if...then...
construction in natural language functions. For example, we are being forced to declare that If New
York is a big apple, then grass is green is true simply on the basis of its consequent being true (which it
is). A more sophisticated treatment of conditionals would involve discussing strict conditionals,
counterfactual conditionals, etc. which are beyond the scope of these notes.)

8
A

B A B A B

AB

f
f

t
f

f
f

t
f

t
t

t
t

Each line under the first two (A,B) columns represents an assignment of truth values
a (truth) valuation. Here there are 22 = 4 valuations. If we had n statement letters
A1 ,...,An there would be 2n valuations.
So far we have laid down rules for forming compound statements from
elementary statement letters. We also have rules for determining the truth values of
any compound statement given the logical operators that occur in it and any truth
valuation of its statement letters. Armed with these rules, we can now test a few
inferences for validity.
It is raining and the cranes are flying.
Therefore, it is raining.
This has the form
A B
A

The inference is valid since, according to the truth table for , whenever the premise
A B is true, so is the conclusion A, thus showing that there are no counterexamples.
It is raining or the cranes are flying.
It isn't raining.
Therefore, the cranes are flying.
This has the form

A B
A
B
Examining the truth table for possible counterexamples we find

A
t
f

prem ises
A
A B
f
t
t
f

B
f
f

conclusion
B
f
f

Notice that we only needed to examine the (two) cases in which the conclusion (B) is
false, since counterexamples cannot arise in any other way. Since neither of these cases

constitutes a counterexample, there are none, and the inference is, accordingly, valid.
If it is raining, the cranes are flying.
It is raining.
Therefore, the cranes are flying.

This has the form

AB
A
B
Examine the truth table for possible counterexamples (conclusion false):
A
t
f

B
f
f

A
t
f

AB
f
t

B
f
f

Neither of these cases constitutes a counterexample, so the inference is valid.

10
If it is raining, the cranes are flying.
The cranes are flying.
Therefore, it is raining.

This one has the form

AB
B
A
The following line in the truth table is a counterexample (in fact the only one)
A

AB

The inference is, accordingly, invalid.

If it is raining, the cranes are flying.
If the cranes are flying, the bears are restless.
Therefore, if it raining, the bears are restless.

This has the form

AB
BC
AC
Now the only possible counterexamples arise when the conclusion A C is false. This
can happen just when A is true and C is false. Therefore we need merely examine the
two lines in the truth table in which this occurs. These are the following:

11
A B
t t
t f

C
f
f

AB
t
f

BC
f
t

AC
f
f

Since in neither of these lines are both premises A B and B C true, neither
constitutes a counterexample, so there are none, and the inference is, accordingly, valid.
We shall use the notation
p1,...,pn |= q
to indicate that the inference from the statements p1,...,pn to the statement q is valid.
Thus the validity of the first three inferences above may be symbolized:
A B |= A; A B, A |= B; A, A B |= B.
We read "p1,...,pn |= q" as "p1,...,pn (logically) imply q" or "q follows from p1,...,pn".

3. Tautologies, Contradictions and Satisfiability

Sometimes conclusions are obtainable without using premises. For example,

consider the premiseless "inference"
Therefore, it is raining or it is not raining.

A A

This "inference" is valid because in its truth table

Note that p |= q is not itself a statement in our logical language, like p q, but rather a kind of metastatement or statement about statementsi.e. the statement that the argument from p as premise to q as
conclusion is a valid argument. However, there is an obvious connection between the expressions p |= q
and p q, namely, the former holds exactly when the latters truth-table has no fs.

12
A
t
f

A A
t
t

the conclusion A A is always true: there are no counterexamples.

A statement which, like AA, is true in all possible cases is called (logically)
valid or a tautology. An inference with a valid conclusion is always valid, regardless of

what its premises are. We shall use the symbol t to stand for a fixed tautologous
statement, which for definiteness we shall take to be the statement AA (although any
tautology would do). The symbol "t" is doing double duty: it indicates both a truth value
and a particular statement. Notice that we then have

p |= t
for any statement p.
A set S of statements is said to be satisfiable or consistent if there is at least one
case in which all the members of S are true, and unsatisfiable or inconsistent if not. This
concept is related to that of validity in the following way.
If p1,...,pn |= q, then the set {p1,...,pn,q} is unsatisfiable, and conversely.

For the unsatisfiability of {p1,...pn,q} is just the assertion that p1,...,pn,q are never
simultaneously true, which amounts to asserting that q is false, i.e. q is true, whenever
p1,...,pn are. In particular, it follows that if {p1,...,pn} is unsatisfiable, then p1,...pn |= q for
any statement q. That is, inconsistent premises yield any conclusion whatsoever.

A single unsatisfiable statement (e.g. A A) is called a contradiction (or

inconsistent). Thus a contradiction is a statement which is always false. Notice that

contradictions are exactly the negations of tautologies. We shall use the symbol "f" to
stand for a fixed contradiction, which for definiteness we take to be the statement

13
A A (although, as in the case of "t", it matters not which particular contradiction we
choose). Notice that we now have, for any statement p,
f |= p .
A statement is said to be contingent if it is neither a tautology nor a contradiction;
so, a contingent statement is one which is true in at least one case, and false in at least
one case. Any statement is either tautologous, contradictory, or contingent: we shall
later develop an efficient technique for deciding which
4. Exercises
Truth Tables and Testing Validity
A1. Use truth tables to determine whether the statements in the left column (jointly)

imply the corresponding statement in the right column.

(a)

A, BC

(b)

(AB)

BA

(c)

A, [(AB)B]

AB

(d)

A (BC)

(AC)B

(e)

A(BC), AB

AC

(f)

AB, [(AC)B]

[(AC)A]

(g)

A(BC)

(AB)C

(h)

A(BC)

(AB)(AC)

(i)

(AB)

AB

(j)

(AB)

AB

(k)

(AB)B

AB

14
(l)

A(AB)

(m)

(AB)C, C, A

(n)

AB, AB

(o)

AA

A2. Symbolize each of the following inferences and use truth tables to determine which

of them are valid.

(a) Silas is either a knave or a fool. Silas is a knave; so, he's no fool.
(b) You may enter only if the Major's out. The Major is out. So you may enter.
(c) There will be a fire! For only if there's oxygen present will there be a fire. And of
course there's oxygen present.
(d) If I'm right, then you're wrong. But if you're wrong, then I can't be right. Therefore
you're wrong.
(e) If I'm right, then you're wrong. But if you're wrong, then I can't be right. Therefore,
I can't be right.
(f) If they retreat provided we attack, then we attack. But they won't retreat. Therefore
we attack.
(g) It's a duck if it walks and quacks like one. Therefore, either it's a duck if it walks like
one or it's a duck if it quacks like one.
(h) You cannot serve both God and Mammon. But if you don't serve Mammon, you'll
starve; if you starve, you can't serve God. Therefore, you can't serve God.
(i) If today's Friday, we must be in Toronto. Today is Friday, but we're not in Toronto.
So we're in London.

15
(j) Computers can think only if they have emotions. If computers can have emotions
then they can have desires as well. But computers can't think if they have desires.
Therefore computers can't think.
A3. Knaves always lie, knights always tell the truth, and in Camelot, where everybody

is one or the other (but you can't tell which by just looking), you encounter two people,
one of whom says to you: "He's a knight or I'm a knave." What are they?
A4. Symbolize this argument and use a truth table (truncated, if you like) to determine

whether it is valid:
If I'm right, then you're a fool. If I'm a fool, I'm not right. If you're a fool, I
am right. So one or other of us is a fool!
A5. Symbolize this set of sentences and determine (using a truth table or otherwise)

whether the set is consistent (i.e. satisfiable):

Either the witness was not intimidated, or if Flaherty committed suicide, a
note was found. If the witness was not intimidated, then Flaherty did not
commit suicide. If a note was found, then Flaherty committed suicide.
A6. Translate the following two arguments into logical notation (defining your

symbols). Then use a truth table (truncated, if you like) to determine whether the
arguments are valid. For the invalid arguments (if any), supply all counterexamples.
(a) You will eat and either I will eat or we shall starve. Therefore, you and I will eat or
we shall starve.(assume that: starve = not eat)
(b) Well win! For if they withdraw if we advance, well win. And we wont advance!
A7. Knaves always lie, knights always tell the truth, and in Camelot, where everybody

is one or the other, you encounter two people, one of whom says to you: "He's a knight
and I'm a knave. What are they?

16
A8. Politicians always lie, taxpayers always tell the truth, and in the US, where

everybody is one or the other (but never both, as we all know!), you encounter two
people, one of whom points to the other and grudgingly declares:
"Im a taxpayer if and only if he is!
What are they?
A9. Using truth tables, determine whether the following arguments are valid.

(a) A (H J)
JH

(b) (D G) G

(c) (G H) (G H)

(G [(A D) A]) D

(G H) (G H)

G D

A
A10. Consider the following argument:

This argument is unsound, for its conclusion is false, and no

sound argument has a false conclusion.
Is this argument sound? (Sound means Valid + True Premises.)

Tautologies, Contradictions and Satisfiability

B1. Classify the following statements as tautologous, contradictory or contingent:

(a) (AB)(BA)

(b) [(AB)B]A

(c) [(AB)C][(AC)(BC)]

B2. Which of the the following assertions is correct and why:

(a) There is a statement that implies every other statement.

(b) Any statement that follows from a satisfiable statement is satisfiable.
(c) Any statement implying a contingent statement is contingent.

17
(d) Any statement that follows from a contingent statement is contingent.
(e) Any statement that follows from a valid statement is valid.
(f) Any statement that implies a valid statement is valid.
(g) All contingent statements imply one another.
(h) No inference with a contradiction as conclusion can be valid.
(i) No statement implies its own negation.
(j) Each of the disjuncts of a valid disjunction is valid.
(k) An implication is valid precisely when the consequent follows from the antecedent.
(l) Any statement implied by its own negation is valid.
(m) Removing a premise from a valid argument cannot affect its validity.
(n) In a valid argument, the conclusion is always consistent with the premises; in a
sound argument it is not. (Note: An argument is sound exactly when it is both valid
and has true premises.)
B3. Determine which of the following five assertions are correct, justifying your answer.

(a) If a statement is not contingent, nor can its negation be.

(b) Every valid argument with a satisfiable set of premises has a satisfiable conclusion.
(c) If a conjunction is a tautology, so is each of its conjuncts.
(d) An invalid argument can always be made into a valid one by adding premises.
(e) The argument from p to q is valid if and only if pq is valid.
B4. Circle the correct answer to each of the questions below.

(a) Identify the statement which is a contradiction in the following:

18
(i) tt

(ii) tf

(iii) ft

(iv) ff

(b) Identify the statement which is valid in the following:

(i) AA

(ii) AA

(iii) AA

(iv) (AA)A

(c) Any argument with an unsatisfiable set of premises must be:

(i) valid and sound

(ii) invalid and sound

(iii) valid and unsound

(iv) invalid and unsound

B5. Provide a one to two sentence answer for these questions.

(a) Explain why each conjunct of a valid conjunction must itself be valid.
(b) Why is it that whenever the pair of statements {p,c} is jointly satisfiable we cant
write p |= c ?
B6. Circle the correct answer to each of the questions below.

(a) Which of the following statements is valid?

(i) A A

(ii) A A

(iii) (A A) A

(iv) A A

(b) Which of the following statements is not equivalent to any of the others?
(i) B B

(ii) B B

(iii) B B

(iv) B B

(c) Which of the following statements is a contradiction?

(i) t

(ii) f t

(iii) t f

(iv) t (t f)

(d) Any argument that concludes with a tautology must be:

19
(i) valid + sound

(ii) valid + unsound

(iii) valid

(iv) sound

(e) At least one of the disjuncts of a valid disjunction must be:

(i) valid

(ii) sound

(iii) consistent

(iv) contingent

(f) The consequent of an inconsistent conditional cannot be:

(i) unsatisfiable

(ii) satisfiable

(iii) a conjunction

(iv) inconsistent

B7. Indicate whether each of the following statements is true or false.

(a) If a statement is not valid, its negation must be.

(b) If a statement fails to logically imply an other, it must imply the negation of that
other.
(c) A statement that logically implies an other cannot imply the negation of that other.
(d) If a set of statements is satisfiable, so is each statement in the set.
(e) If each statement in a set is satisfiable, so is the set.
(f) You cannot make a valid argument invalid by adding more premises.
(g) You cannot make an invalid argument valid by removing premises.
(h) Sound arguments can never have f as their conclusion.
B8. Indicate whether each of the following statements is true or false.

(a) A tautologous conjunction must have a tautologous conjunct.

(b) A contradictory disjunction must have a contradictory disjunct.

20
(c) If neither a statement nor its negation is valid, then both must be consistent.
(d) If a conditional is unsatisfiable, its consequent must be too.
(e) A contingent statement can logically imply both a statement and the negation of that
statement.
(f) No subset of a set of satisfiable statements can be unsatisfiable.
(g) Every statement logically implies at least one other statement with which it is not
equivalent.
(h) You can never make an invalid argument into a sound one by dropping some of its
premises.
(i) You can never make a valid argument into an unsound one by adding more premises
to it.
(j) Some statements are equivalent to every statement that logically implies them.
B9. Using truth tables (where necessary), decide if the following sets of sentences are

satisfiable.
(a) {A B, B C, A C}
(b) {[J (H L)], L (J H), H (J L)}
(c) {(J J)H, J, H}
(d) {A, B, C}
(e) {(A B) (C B), A, B}
B10. True or False?

(a) A conjunction with one valid conjunct must itself be valid.

(b) An implication with a valid consequent must itself be valid.

21
(c) A disjunction with one unsatisfiable disjunct must itself be unsatisfiable.
(d) A sentence is valid iff its negation is unsatisfiable.
(e) An implication with a valid antecedent must itself be valid.
B11. Using truth tables, determine whether the following are valid.

(a) (F H) (H F)

(b) A [(BA) C]
II. EQUIVALENCE

1. Equivalence and Bi-implication

Two statements are called (logically) equivalent if they take the same truth values
in all possible cases. For example, consider the truth tables for the statements A B,
B A:
A

AB

B A

Since A B and B A have the same truth value on every line of the table, they are
equivalent.
We write p q to indicate that the statements p and q are equivalent. We may
think of as a kind of equality between statements. We leave it to the reader as an
exercise to show that for any statements p,q the assertion that p q amounts to the
same thing as:
p |= q and q |= p
In connection with , we can define a new logical operator" " called biimplication (or if and only if) as follows:

22
A
t
t

B
t
f

A B
t
f

f
f

t
f

f
t

Thus A B has value "t" exactly when A and B have the same truth value. It follows
from this that p q holds when and only when the statement p q is valid. The
statements p and q are called the components of p q.
It is easy to check the following equivalences:
p p
p q (p q)
p q (p q)
p q p q
p q (p q) (q p)
We see from these equivalences that and are in a natural sense expressible in terms
of {,,} and hence in terms both of {,} and {,}. The question now arises as to
whether every possible truth function is so expressible. The answer, as we shall see, is
yes.
2. Expressive Completeness

We begin by describing what is meant by the disjunctive normal form of a truth

function. Let us suppose we are given a truth function H of n statement letters A1,...,An.
Thus, for each valuation of A1,...An, a corresponding (truth) value of H(A1,...,An) is
obtained. (For example, H(A1, A2, A3) = (A1 A2) A3 defines a truth function, since
truth values for A1, A2, and A3 fix a truth value for (A1 A2) A3 via the truth tables
for and .) We display this situation in the form of a truth table:

23
A 1 A 2 ... A n
t
t ... t
t
t ... f
f

f ...

H(A 1,...,A n )
*
*
*

Assume first that at least one of the entries in the H column is "t". For each valuation of
A1,...,An in which a "t" appears in the H column we form the conjunction A1* ... An*
where each Ai* is Ai if the given valuation assigns t to Ai and Ai if not. Notice that this
conjunction is true precisely under the given valuation and no other. Now we form the
disjunction of all these conjunctions arising from the "t" cases of the given truth table.

The resulting statement is called the disjunctive normal form (d.n.f.) of the given truth
function. Clearly, its truth table is identical to that of the given truth function.
It remains to consider the case in which the given truth function always takes the
value "f". Here we may take the disjunctive normal form to be, e.g., A1 A1.
Since d.n.f.s contain only the logical operators , , , it follows from all this that
every possible truth function can be expressed in terms of , , , and so every statement is
equivalent to one whose only logical operators are these. We sum this up by saying that the

set {, } is expressively complete. Moreover, since is expressible in terms of , and

in terms of , , we may infer that each of the sets {,} and { } is expressively
complete.

Before proceeding further let us determine a d.n.f. in a practical case. Suppose

we are given, for instance, the truth table

24
A B

H(A,B,C)

t f f
all remaining lines

t
f

The d.n.f. here is, writing A for A etc. and omitting the ""s,
ABC ABC ABC
The question arises as to whether there are single logical operations (involving
just two statement letters) which are expressively complete. We shall see that there are
exactly two of these.
We define the logical operators ("Sheffer strokes") "|" nand and "" nor
by means of the following truth tables.
A
t
t

B
t
f

AB
f
t

f
f

t
f

t
t

A B
f
f
f
t

Clearly, A|B (A B) and AB (A B) (hence nand is short for not and and
nor short for not or!).
First, we show that | and are each expressively complete. To do this it suffices
to show that and are both expressible in terms of |, and and in terms of .
(Why?)
Clearly A|A (A A) A, so is expressible in terms of |. Now
A|B (A B) A B,

25
so
A|B A B A B.
Hence, recalling that A A|A, we see that
A B (A|A)|(B|B),
and so is expressible in terms of |.
Similarly, A AA and A B (AA)(BB). Therefore | and are each
expressively complete.
We next show that | and are the only expressively complete logical operations
on two statement letters.
For suppose that H(A,B) is expressively complete. If H(t,t) were t, then any
statement built up using only H would take the value t when all its statement letters
take value t. So A would not be expressible in terms of H. Therefore H(t,t) = f.
Similarly, H(f,f) = t. So we obtain the partial truth table
A
t
t

B
t
f

f
f

t
f

H(A,B)
f

If the second and third entries in the last column are t,t or f,f, then H is | or . If they are
f,t, then H(A,B) A; and if they are t,f, then H(A,B) B. So in both of these cases H
would be expressible in terms of . But clearly is not expressively complete by itself,
since the truth function t is not expressible in terms of it. So H is | or as claimed.
3. Arithmetical Representation of Statements and Logical Operations

26
Statements and logical operations can be nicely expressed within binary
arithmetic: the arithmetic of 0 and 1.

First, we describe the rules of binary arithmetic. We suppose given the two
numbers 0,1 and two operations "+" (plus) and "." (times) on them subject to the
following rules (only one of them may be unfamiliar!):
0+0=1+1=0

0.0 = 0.1 = 1.0 = 0

0+1=1+0=1

1.1 = 1

We shall think of statements as determining binary functions (that is, functions

taking just the values 0 and 1) as follows. Statement letters A,B,C,... will be regarded as
variables taking values 0,1: we think of 1 as representing the truth value t and 0 as

representing the truth value f.

Then the operation "" corresponds to "." and the

operation "" to the operation "1 + " of adding 1.

Given this, how do we interpret "" and ""? We argue as follows.
A B (A B)
= 1 + (1 + A).(1 + B)
= 1 + 1 + A + B + A.B
= 0 + A + B + A.B
= A + B + A.B
And
A B A B
= 1 + A + B + (1 + A).B
= 1 + A + B + B + A.B
= 1 + A + A.B
In this way, any statement p gives rise to a binary function called its binary
representation which we shall denote by the same symbol p. In that case, tautologies are

27
those statements whose binary representations take only value 1, and contradictions
those statements whose binary representations take only value 0.
When, for example, is p q a tautology? Exactly when the corresponding
binary representation 1 + p + p.q is constantly 1. But this is the case precisely when 0 =
p + p.q = p.(1 + q), that is, when at least one of p and 1 + q is 0, in other words, if p = 1,
then 1 + q = 0, i.e. q = 1. But this means that the value of p never exceeds the value of q:
we shall write this as p q. It follows that
p |= q p q is a tautology p q
(where we have written "" to indicate equivalence of assertions). That is, in the binary
representation, |= corresponds to . By the same token,

p q p = q.
That is, in the binary representation, corresponds to =.
The binary representation sheds light on expressive completeness. For example,
the expressive completeness of {,} translates into the assertion that any binary
function can be expressed in terms of the operations "." and "1 + ", while the expressive
completeness of "|" translates into the assertion that any binary function can be
expressed in terms of the single binary function 1 + x.y.
4. Venn Diagrams

Venn diagrams are a convenient method of depicting logical relationships. We

represent the various truth valuations of statement letters by points in a rectangle
(which itself may be thought of as a kind of "logical space"). Then, for a given statement
p, the collection of valuations making p true is represented by a circle within the
rectangle:

28

This circle is called the region corresponding to p, or simply the region of p.

It is clear that the region corresponding to any tautology is the whole rectangle, and
that corresponding to any contradiction is the empty region. For a conjunction p q the
corresponding region is the shaded portion in the figure below, that is, the intersection

of the regions corresponding to p and q. For a disjunction p q the corresponding

region is that covered by the union of both circles. For a negation p the corresponding
region is that lying outside the region of p: its complement.

The relation of logical implication corresponds to the relation of inclusion between

regions: p |= q is equivalent to the region of p being

29

included in the region of q. Finally, the relation of equivalence corresponds to the relation

of identity between regions: p q is equivalent to asserting that the regions of p and q

coincide.

It is now straightforward to illustrate the following rules of equivalence by

means of Venn diagrams:
1.
2.

p p
pq qp

3. p (p q) p

pq qp
p (p q) p

4.

pff

pfp

5.

ptp

ptt

6. p (q r) (p q) (p r) p (q r) (p q) (p r)
7.

(p q) p q

(p q) p q.

Here on line 1 we have the law of double negation, on line 2 the commutative laws, on line 3
the absorptive laws, on line 6 the distributive laws, and on line 7 de Morgan's laws.
For example, consider the Venn diagram immediately below.

30
Clearly the shaded region that corresponding to (p q) is the union of the region
outside that of p with the region outside that of q. This latter is the region
corresponding to p q. This verifies the first de Morgan law. The remaining laws
may be similarly verified.
5. Exercises
Equivalence
A1. Which of the following pairs of statements are equivalent?

(a)

(AB)A

AB

(b)

A(AB)

AB

(c)

A(AB)

AB

(d)

(AB)

AB

(e)

(AB)

(AB)(AB)

(f)

A(BC)

(AB)C

(g)

A(BC)

(AB)C

(h)

A(AA)

(i)

(AA)A

A2. (a) Indicate which of the following statements are valid:

(i) tt
(ii) ff
(iii) (tf)(f|t)
(iv) (pp)

31
(v) (pp)
(vi) (pq)(qp)
(vii) (pq)(qp)
(viii) (pq)(pq)
(ix) (p(pq))(q(pq))
(b) Which of the above statements (i)-(ix) are equivalent to each other?
Expressive Completeness
B1. Find statements involving the operators , , and the statement letters A, B, C that

have the following truth tables (1), (2), (3):

A
t
f
t
f
t
f
t
f

B
t
t
f
f
t
t
f
f

C
t
t
t
t
f
f
f
f

|
|
|
|
|
|
|
|

(1)
t
t
t
f
f
f
f
t

(2)
t
t
t
f
t
f
t
f

(3)
f
t
f
f
t
t
f
t

B2. The logical operator called exclusive disjunction is defined by specifying that pq

is true when exactly one of p, q is true, and false otherwise.

(a) Define in terms of and .
(b) Show that is associative, in the sense that p(qr) (pq)r, for any statements
p, q and r.
(c) Show that {,,t} is an expressively complete set.

32
(d) What are the truth conditions for p1p2 ... pn ? (That is: when would you
regard such an expression as true and when not?)
B3. (a) Show that the pair {,} is expressively complete.

(b) Show that the single truth function f(A,B,C) = (AB)C is expressively
complete. (Hint: One approach is to show that {, } are both expressible in terms
of the function f and invoke (a).)
B4. (a) Show that cannot be expressed in terms of alone. (Hint: any statement

containing exactly A, B, takes value t in at least one case where A and B have
opposite truth values.)
(b) Show that can be expressed in terms of alone.
(c) Show that cannot be expressed in terms of alone. (Start by showing that any
statement containing just the logical operator must take truth value t in at least two
cases.)
B5. Find statements in , , that have the following truth functions f, g, h.

A
t
f
t
t
f
t
f
f

B
t
t
f
t
f
f
t
f

C
t
t
t
f
t
f
f
f

f(A, B, C)
t
t
f
f
f
f
f
t

g(A, B, C)
f
t
t
t
f
t
t
t

h(A, B, C)
t
f
f
f
f
t
t
t

B6. Show that the truth function h(A, B, C) determined by (AB)C is expressively

complete.
B7. Find disjunctive normal forms for the following statements:

(a) (A B) (A C)

(b) A [(B A) C]

33
B8. (a) Explain why {f,} must be an expressively complete set.

(b) Out of the 16 possible binary logical operators one could define, exactly how many
can be expressed in terms of alone? (This ones moderately difficult!)
Binary Representations and Venn Diagrams
C1. (i) Find the binary representations of the following statements, (ii) using those

representations classify each statement as valid, contradictory or contingent, and (iii)

draw the Venn diagram corresponding to each statement:
(a) AB

(b) AB

(c) AB

(d) (AB) (BA)

(e) (AA) (BB)

C2. Find the binary representations of the following statements and draw their Venn

diagrams:
(a) A(BB)

(b) (AA)(AA)

(c) (AB)(AB)

(d) p(qr)

(e) [p(qr)](ss)

(f) (CD)|(CD)

C3. Find the binary representations of the following statements and draw their Venn

diagrams:
(a) (AA)(AA)

(b) [(A|A)|(A|A)]

(c) [(AB)(AB)]

(d) (pq)r

(e) [(pq)r](ss)

(f) (CD)(CD)

C4. Find the disjunctive normal forms, binary representations and Venn diagrams for

the following statements:

(a) A(BA)
(b) [(AB)A]C

34

III. TRUTH TREES

1. Introduction to Truth Trees

To test an inference for validity it suffices to conduct an exhaustive search for

counterexamples. If none are found, then the inference is valid. Truth trees are an
efficient and elegant device for unearthing counterexamples.

35
Consider, for example, the (valid) inference
A B
A
B
To obtain its tree form, we start by listing its premises and the negation of its conclusion:
A B
A
B
These statements will be true in exactly the cases in which there are counterexamples to
the original inference. Now we continue, generating a tree-like structure:

A B
A
B

Here the statements A, B, A, B, A B occupy positions, or as we shall call them,

nodes1in the tree. The statement occupying the top node is a disjunction and requires

analysis: A B is true in all those cases in which A is true and all those cases in which B
is true, and in no other cases. We indicate this by writing A and B at the ends of a fork
at the foot of the tree. At the same time we tick the statement AB, using , to indicate

We shall often identify a node in a tree with the statement occupying it.

36
that all its t cases have been taken into account. Ticking a statement2 is, accordingly,
equivalent to erasing it. Finally we write "" at the foot of each path through the tree in

which a statement occupies one node and its negation another. Such paths are called
closed. In this particular tree all paths are closed; under these conditions the tree itself is

said to be closed. And, as we shall see, the inference is then valid.

Why is this? Because the procedure was designed so that when we ticked a
statement, we displayed all the possible ways in which that statement can be true. The
various paths then represent all the ways in which the initial statements (i.e., the
statements with which we began the tree) could possibly be true; that is, each path
represents a potential counterexample to the original inference. In the case of a closed
path, the possibility it represents does not really exist. Accordingly, if all paths are
closed, then it is impossible for all the initial statements of the tree to be
(simultaneously) true, in other words, there are no counterexamples to the original
inference and so it is valid.
In contrast, observe what happens when we test an invalid inference, e.g.,
A B
A
B

In this case the tree looks like this:

We shall often use the locution"to tick a given node" as a synonym for "to tick the statement occupying
the given node".

37
A B
A
B

The left-hand path is not closed, that is, it is open and represents a genuine
counterexample to the inference in question. To describe it, note which statement
letters, with or without , occupy nodes in the path. In this case they are A, B, and the
corresponding counterexample is that in which A is true and B is false:
A
t

B
f

A B
t

A
t

B
f

We next describe the various tree rules.

2. The Tree Rules
Disjunction
pq
p

Tick a disjunction occupying a node and write the disjuncts at the end of a fork drawn
at the foot of each open path containing the ticked node.

38
Negated conjunction
(pq)

Tick a negated conjunction occupying a node and write the negations of the conjuncts at
the end of a fork drawn at the foot of each open path containing the ticked node. (For: a
conjunction is false exactly when some conjunct is false. Notice that, by de Morgans
law, this rule is nothing but the disjunction rule in disguise.)
Conjunction

pq

p
q
Tick a conjunction occupying a node and write the conjuncts in a column at the foot of
each open path containing the ticked node. (Justification: a conjunction is true exactly
when both conjuncts are true.)

39
Negated disjunction

(pq)

p
q
Tick a negated disjunction occupying a node and write the negations of the disjuncts in
a column at the foot of each open path containing the ticked node. (Justification: a
disjunction is false exactly when both disjuncts are false.)
Implication
pq
p

Tick an implication occupying a node and write the negation of the antecedent and the
consequent at the ends of a fork drawn at the foot of each open path containing the
ticked node. (For: an implication is true exactly when the negation of the antecedent is
true, or the consequent is true, or both.)
Negated implication
(pq)
p
q

Tick a negated implication occupying a node and write the antecedent and the negation
of the consequent in a column at the foot of each open path containing the ticked node.
(For: an implication is false exactly when the antecedent is true and the consequent
false.)

40
Bi-implication
pq

p
q

p
q

Tick a bi-implication occupying a node and draw a fork at the foot of each open path
containing the ticked node. At the ends of each of these write in columns the
components, and, respectively, the negations of the components, of the ticked node.
(For: a bi-implication is true exactly when both components are true, or both are false.)
Negated bi-implication

(p q)
p
q

p
q

Tick a negated bi-implication occupying a node, and draw a fork at the foot of each
open path containing the ticked node. At the ends of these write in columns the first
component and the negation of the second, and, respectively, the negation of the first
and the second. (For: a bi-implication is false exactly when one component is true and
the other false.)
Double negation

p
p
Erase double negations. (For: the negation of a statement is false exactly when the
statement is true.)

41
We summarize these rules as follows:

Negation
Affirm ed

Conjunction

Disjunction

pq

pq

q
Negated

(p q)

(p q)

p
q

Im plication
Affirm ed

pq
p

Negated

(p q)

Bi-im plication
(p q)
p

(p q)

42
When applying a tree rule of the form
p
q
r

p is called the premise, and {q,r} the list of conclusions, of the application. Similarly, when
applying a tree rule of the form
p
q
r

q'
r'

p is called the premise, and {q,r}, {q',r'} the lists of conclusions, of the application.
3. Tree Test for Validity

To test an inference for validity, write its premises and the negation of its
conclusion in a column and apply the tree rules to all unticked lines of open paths,
ticking lines to which rules are applied, until the tree is finished, i.e. until the only
unticked nodes in any remaining open paths are statement letters and their negations.
A tree obtained in this way is called a [finished] tree associated with the given inference.
If any such tree is closed, i.e. if all its paths are closed, the original inference is valid.
We now give some examples of the use of this test.

43
(A B) C
A
C

(A B) C
A
C
A B

A
B

Tree closed, so inference valid.

A B
CA
A
CB

A B
CA
A
(C B)
A

B
C
C

Tree closed, so inference valid.

(A B)
A
B

(A B)
A
B
A

Tree closed, so inference valid.

44
A B

A B

CD

CD

(B D)

(B D)

A C

(A C)
A
C
A

D
B

Tree closed, so inference valid.

AB
BC
AC

AB
BC
(A C)
A
C
A

Tree closed, so inference valid.

45
AB

AB

(A C) (B C)

[(A C) (B C)]
A C
( B C)

( A C)
B C

B
C
A

A
B

A
B

A
C
C

B
A
B

A
B

Tree closed, so inference valid.

4. Further Applications of the Tree Method

A. Counterexamples from the associated tree. Any open path remaining in a

finished tree associated with an inference determines a counterexample to it (and so establishes
its invalidity). And conversely, any counterexample is determined by an open path in any such
tree.

For example, here is an invalid inference:

A B
CD
C
Consider the following finished tree associated with this inference:

46
AB
CD
C
A
C

B
D

Each open path in this tree determines a counterexample to the given inference. For
example, the left-hand open path, nodes of which are occupied by A, C, D, but by
neither B nor B, determine as counterexamples all cases in which A, C, D are t, f, t
respectively, regardless of the truth value of B. That is, we obtain two counterexamples
A: t, B: t, C: f, D: t, and A: t, B: f, C: f, D: t. Similarly, the right-hand open path
determines as counterexamples all cases in which B, C, D are t, f, t respectively,
regardless of the truth value of A. In total we get the three distinct counterexamples
ABCD: ttft, tfft, ftft.
These are all the counterexamples to the given inference.
In this connection we observe that the open paths in the other finished tree
associated with the above invalid inference, viz.,
A B
CD
C
C

D
A

of course determine exactly the same counterexamples as were obtained before.

Recall that a set of statements is satisfiable or consistent if there is at least one case
in which all the members of the set are true.

47

B. Tree test for satisfiability. Given a set S of statements, start a tree with the
members of S in a column. Then S is satisfiable precisely when there is an open path through the
finished tree. Each open path determines a truth valuation that makes all the members of S true.

We illustrate this by the following example. Consider the set of statements

{A B, (A B)}.
The relevant finished tree is
AB
(AB)
A
A

B
B A

There are two open paths in which the statement letters (negated or unnegated) A, B;
A, B respectively, occupy nodes. Thus the valuations making the given set of
statements true are AB:tf or ft.
C. Tree test for logical validity. To determine whether a given statement is logically
valid, start a tree with its negation. Then the given statement is logically valid precisely when
the resulting finished tree is closed.

For example, consider the statement (A B) (A B). To test for logical

validity, we construct the following tree:

48
[(AB) (AB)]
(AB)
(A B)
A
B
A

Since this (finished) tree is closed, the statement in question is logically valid.

D.

Tree test for contradictions.

To test whether a given statement is a

contradiction, start a tree with the (unnegated) statement.

Then the statement is a

contradiction precisely when the resulting finished tree is closed.

For example, to test whether the statement A (A B) (AB) is a

contradiction, construct the following tree:

49
A [(AB) (AB)]
A
(AB) (AB)
AB
AB
A

B
A

Since this (finished) tree is closed, the given statement is a contradiction.

5. Correctness and Adequacy of the Tree Method

We conclude this chapter with some arguments designed to justify the claims we
have made concerning the use of trees in establishing validity and satisfiability.
First, let us call a tree rule R correct if whenever the premise of R is true under a
given valuation, then all the statements in at least one of R's lists of conclusions are also
true under the valuation. And let us call R complete if the converse holds, that is, the
premise of R is true under a given valuation whenever all the statements in at least one
of R's lists of conclusions is true under the valuation.
Clearly, all the tree rules we have introduced are correct and complete in the above
senses.

Next, we observe that the process of constructing a finished tree always

terminates. For the tree starts with a finite number of statements, each of which has

finite length (taking the length of a statement to be the total number of symbols in it),
and it grows downward by a process of choosing an unticked statement occupying a

50
node of an open path, ticking it and adding at the foot of the path some finite number of
statements, each of which is shorter than the ticked one. Eventually the point must be
reached at which all unticked statements occupying nodes of open paths have lengths 1
or 2 (i.e., are statement letters or their negations) and the process ends.
Given a set S of statements, let us say that a tree starts with S if it has S as its
initial set of statements. Now we can establish the

Correctness of the tree method. If a set S of statements is satisfiable, there will be an

open (complete) path through any tree that starts with S.

To prove this, observe first that, if all the statements occupying nodes in a path P
of a tree are true under a given valuation, then P is open. For if there is a valuation
making all statements occupying nodes in P true, then both a statement and its negation
cannot both occupy nodes in P, otherwise the (alleged) valuation would have to make
both a statement and its negation trueimpossible. It follows that P cannot contain
both a statement and its negation, which is just to say that path P is open.
Now suppose that under some valuation V all the members of S are true.
Consider the following property of a tree T.

(*)

T starts with S and contains a (complete) path P such that all statements

occupying nodes of P are true under V.

By the observation above, any tree satisfying (*) contains an open path.

51
We claim that, if T has property (*), so does any tree T* obtained from T by
applying a tree rule. For suppose that (a) all the statements occupying nodes in a
certain path P through T are true under V and (b) we extend T to T* by applying a tree
rule to one of its statements. Clearly we may assume that this statement is in P, for if
not, then P is unaffected and is a complete path of T*. Accordingly in the transition
from T to T* the path P is extended to a new path, or extended and split into two new
paths, by applying some tree rule. Since any tree rule is correct, all the statements
occupying nodes in the new path, or all those occupying nodes in at least one of the
new paths (each of which extends the path P), are true under V. But this shows that T*
has property (*), as claimed.
It follows that any tree T starting with S has property (*), and hence contains an
open path. For any tree T starting with S can be built up (or rather, down!) by starting
first with the tree with a single path consisting of the statements in Swhich has
property (*) by definitionand then applying tree rules, one after another (finitely
many times), until tree T results. By the argument of the previous paragraph, at each
stage of the tree building process, property (*) is preserved, therefore the end result
the tree Twill have that property too (and so must contain an open path, which is
what we needed to show).
As an immediate consequence of this, we obtain the
Inference correctness of the tree method. If a finished tree associated with an
inference is closed, then the inference is valid.

Now we prove the converse of the above correctness result, that is, the
Adequacy of the tree method. If there is an open path through a finished tree
starting with a given set S of statements, then S is satisfiable.

To prove this, let T be a finished tree starting with S and containing an open path
P. We are going to show how to define a truth valuation V on the statement letters that

52
figure in tree T such that the sentences in set S all come out true under V. Consider the
single statement letters that occur in path P (not negated statement letters, just the nonnegated elementary statements in P). Let V be the valuation that assigns all those
statement letters value t, and all the statement letters that do not occur in path P (i.e. that
occur somewhere else in the tree T) the truth value f. (If there are any other statement
letters left out of this assignment, let them take any truth value you want.) We claim
that all statements occupying nodes of P are true under V, not just the nodes containing
statement letters.
To show this first notice that all statements of lengths 1 or 2 occupying nodes of P
are true under V. For those of length 1 are statement letters and are accordingly true
under V by definition. And any one of length 2 is a negation A of a statement letter A;
since P is open, A cannot occupy a node of P, and so is false under V. Thus A is true
under V.
Now suppose that, if possible, some statement occupying a node of P is false
under V. Let p be such a statement of shortest length. Then by the above the length of p
must be at least 3, so a tree rule, R say, may be applied to p. Since T is finished, some
list L of conclusions obtained by applying R to p is already part of P.

But each

statement in L is shorter than p, and so must be true under V. Since R is complete, it

follows that the premise p of the specified application of R is also true under V.
Therefore the falsity of p is refuted, and the claim above follows.
Given the (now established) truth of the claim that all statements occurring in
path P are true under V, it follows (in particular) that the initial statements in the set S
are true under V (since that set clearly lies in all paths of the tree, P included). It
follows, then, that the set S is satisfiable, and hence the tree method is adequate in the
sense spelled out above.

53
As an immediate consequence, we obtain the converse of validity correctness,
that is, the

Inference adequacy of the tree method. If an inference is valid, then any finished
tree associated with it is closed.
6. Exercises
Tree Test for Validity
A1. Use the tree method to determine whether the following arguments are valid. In

the invalid cases, find all counterexamples:

(a)

(AB)C
AD

(b)

A(BC)
(AC)B

B(CD)

(c)

(BA)C
(BD)(AE)
(DE)C
AB

(d) If Holmes has bungled or Watson is windy, Moriarty will escape. Thus Moriarty
will escape unless Holmes bungles.
(e) Moriarty will not escape unless Holmes acts. We shall rely on Watson only if
Holmes does not act. So if Holmes does not act, Moriarty will escape unless we rely on
Watson.
(f) Moriarty will escape only if Holmes bungles. Holmes will not bungle if Watson's
to be believed. So if Watson's to be believed, Moriarty won't escape.
A2. Use the tree method to determine which of the following inferences is valid. In the

invalid cases, supply all counterexamples.

(a)

A (B C)

(A B) (A C)

54
(b) (A B) (A B)
AB
AB
(c) (A B) C
AD
B C D
(d) (A B) C
AC
A3. Use the tree method to determine which of the following inferences are valid. In

the invalid cases find all counterexamples.

(i) (A B) C
A D

(ii) A (B C)

(iii )

(A C) B

(B A) C
(B D) (A E)

B (C D)

(D E) C
AB

A4. Define the new logical operations [A, B, C] and A*B by

[A, B, C] = (A B) C
A*B = A B
Devise the simplest tree rules you can for these operations and their negations. Use the
rules you have devised to determine which of the following inferences are valid:
(i) [A, B, C]

(ii) [A, B, C]

A*B

A*B
C*B

A5. Use the tree method to determine which of the following inferences are valid. In

the invalid cases find all counterexamples.

55
(i) (A B) C

(ii)

A D

(A B) C

(iii) A (B C)

(AC) (BC)

(A C) B

B (CD)
A6. Define the logical operations {A, B, C}, A*B, and AB by

{A, B, C} = (B A) (B C)
A*B = A B
AB = (A B)
Devise the simplest tree rules you can for these operations. Use these rules to test the
validity of the following inferences:
(i) AB

(ii) A*B

A*B

AB

(iii)

{A, B, C}

(iv) {A, AB, C}

AC

A*C
B

A7. Determine which of the following arguments are valid. In the invalid cases, supply

all counterexamples.
(a) AB

(b) (qr)p

(c) p(qr)

qq

qq

AC

rp

pr

CB

p(qr)

(qr)p

A8. Determine which of the following arguments are valid. In the invalid cases, supply

all counterexamples.
(a) AB

(b) (qr)p

(c)

It's a duck if it walks

BC

qq

and quacks like one.

CD

rp

Either it's a duck if it walks like one

AD

pp

or it's a duck if it quacks like one.

56
A9. Translate the following arguments into logical notation (indicating what elementary

sentences your symbols refer to) and then determine whether each argument is valid. If
not, indicate the total number of counterexamples.
(a) If Dumb knows that hes dumb, then hes dumb. If he knows that hes dumb, then
he at least knows something. If Dumb knows something, then hes not dumb after all!
Therefore, Dumbs not dumb.
(b) Canadas economy will fail if Quebec does separate. If Canadas economy wont
fail, then the market will get the jitters if Quebec does separate. The market will get the
jitters even if Quebec doesnt separate. So, the market will get the jitters and Canadas
economy will fail.
A10. Use the tree method to determine whether the following argument is valid; if not,

supply one counterexample.

Either scientists dont know what they are talking about, or the sun will eventually
burn out and Earth will become dark and cold. If scientists dont know what they are
talking about, then Mars is teeming with life. If Earth becomes dark and cold, then
either the human race will migrate to other planets or will die out. Mars is not teeming
with life, but the human race will not die out. Therefore, the human race will migrate to
other planets only if Mars is teeming with life.
A11. Use the tree method to determine whether the following argument is valid. If not,

supply the exact number of counterexamples.

It will not be the case that both the Representatives and the Senators will pass the
bill. If either the Representatives or the Senators pass it, the voters will be pleased; but
if both pass it, the President wont be pleased. The President wont be pleased if and
only if Dole rejoices. Therefore, Dole wont rejoice.
Further Applications of the Tree Method

57
B1. Use the tree method to determine which of the following sets of statements are

satisfiable. In the latter cases, supply all the satisfying valuations.

(a) A,B, (A B)
(b) A,B, A B
(c) A, B, A B.
B2. Use the tree method to determine which of the following statements are

tautologies.
(a) (A B) A
(b) A (A B)
(c) (A B) A B
B3. Use the tree method to determine which of the following statements are

contradictions.
(i) [ [ (A B) A] A]

(ii)

[ (A B) B] A B

B4. In the land of knights and knaves, knights always state the truth and knaves

falsehoods. Punch and Judy are two inhabitants of this land. From their assertions in
each case use the tree method to deduce as much as you can about their statuses.
(i) Punch: Judys a knight
Judy:

Were not both knaves.

(ii) Punch: If Judys a knave, we both are.

Judy: Either hes a knight, or Im a knave.
B5. Use the tree method to determine which of the following pairs of statements are

equivalent.

58
(i)

A (B C)

(A B) C

(ii)

(A B) C

A (B C)

(iii) (A B)
(iv) (A (B C))
(v)

A (B C)

A B
A (B C)
A B C

B6. Classify each of the following statements as tautologous, contradictory or

contingent.
(a) ((AB)B)A
(b) (p q) p
(c) B(CC)
(d) (p(pq))q
B7. Knaves always lie, knights always tell the truth, and in Camelot, where everybody

is one or the other, you encounter two people, one of whom says to you:
(i) "He's a knight and I'm a knave." What are they?
(ii) What if that person had said: "If he's a knave, then so am I"?
(iii) How about if that person had said: "I'm a knight, and, then again, I'm not; though
he's a knave if I am" ?
B8. Classify each of the following statements as tautologous, contradictory or

contingent.
(a) ((AB)B)B
(b) (p q) q

59
(c) (BB)(CC)
(d) [(pq)p)]q
B9. Knaves always lie, knights always tell the truth, and in Camelot, where everybody

is one or the other, you encounter three people, Lancelot, Arthur and Merlin, who say to
you:
Lancelot: Merlin's a knave.
Arthur: Either Lancelot or Merlin is a knave.
Merlin: If I'm a knave, they are too.
What are they?
B10. Circle the tautologies that occur below:

(a) A A

(b) A (B A)

(d) A (A A)

(c) A (B A)

(e) (A B) (B A)

B11. Circle the inconsistent sets of sentences that occur below:

(a) B A , B , A

(b) A A , A A

(d) A B , B C , A C

(c) A (B B) , A

(e) (A (B C)) , C

B12. In the land of knights and knaves, where knaves always lie, knights always tell the

truth, and everybody is either one or the other (clearly no one can be both!), you
encounter two people, Dumb and Dumber, both of whom speak to you. In each case
below, determine as much as you can about their individual identities.
(a) Dumb: If Im a knave, we both are. Dumber: Hes a knight or Im not.
(b) Dumb: Dumber is a knight if and only if 2+2 is 4.
4!

Dumber: Come on, 2+2 is not

60
B13. These puzzles concern a land populated by saints and sinners. Saints always tell

the truth; sinners always lie. You are a traveler in this strange land and must try to
identify those you meet as saints or sinners.
You encounter two people, Mutt and Jeff, one or both of whom speak to you.
What can you deduce in each case, using the tree method, about whether they are saints
or sinners?
1. Mutt: Im a saint.
2. Jeff: Mutt is a saint.
3. Mutt: Jeff's a sinner.
4. Jeff: Either Im a saint, or Im not.
5. Mutt:Im a saint, and, then again, Im not.
6. Jeff: If Mutt is a sinner, so am I.
7. Jeff: Neither of us is a saint.
8. Mutt: Were not both saints.
9. Mutt: Im a sinner if and only if Jeff's a saint.
10. Jeff: Mutt is a saint, and Im a sinner.
11. Mutt: Im a sinner unless Jeff's a saint.
12. Mutt: If either of us is a sinner, I am.
13. Mutt: Jeff's a sinner.
Jeff: Were not both sinners.
14. Mutt: Im a saint if and only if Jeff's a sinner.
Jeff: Mutt is a sinner.

61
15. Mutt: Jeff's a saint.
Jeff: At least one of us is a sinner.
16. Mutt: Im a saint if and only if Jeff is.
Jeff: Mutt is a saint.
17. Mutt: If Im a sinner, we both are.
Jeff: Either hes a saint, or Im a sinner.
18. Mutt: Jeff's a saint if and only if his brother is.
Jeff: Unfortunately, my brother's a sinner.
19. Mutt: Jeff and his brother are both saints.
Jeff: Well, Im a saint, but my brother isnt.

At this point, you meet three curious looking people in the land of knights and knaves.
What can you deduce about their status?
20. Curly: Larrys a sinner.
Moe: Either Curly or Larry is a sinner.
Larry: If Im a sinner, they are too.
21. Curly: Were all saints.
Moe: Well, Im a saint, but Larrys a sinner.
Larry: No, the other two are both sinners.
22. Curly: ThatMoes a saint.
Moe: No,were all sinners.
Larry: Curly, Moe, and their cousins are all sinners.

62
23. Curly: Well, at least were not all of us sinners.
Moe: Curly is.
Larry: If Curly is, Moe is too.
24. Curly: If Moes a saint, Larry is too.
Moe: Well, Larrys a sinner if Curly's one.
Larry: But Curly and Moe arent both sinners.
25. Curly: If any of us are saints, Larry is.
Moe: But Larrys a sinner.
Larry: And Im a sinner if and only if Moe's one.
26. Curly: If Moes a sinner, Larry is too.
Moe: If Larrys a sinner, so is Curly.
Larry: If Moes a saint, we all are.
B14. Determine which of the following sets of statements are (jointly) satisfiable, in each

case describing the satisfying valuations:

a) A B
B C
B (C D)

(b) (B A)
A C
B C

(c) D B
A B
(D A)
D

B15.

Using the tree method, determine which of the following statements are

tautologies. In the non-tautologous cases, supply all the truth valuations that make the
statement false.
(i) ((A B) B) A

63
(ii)

A (B (BA))

B16. Using the tree method, determine which of the following sets of statements are

satisfiable. In the satisfiable cases, supply all the satisfying valuations.

(i) A B, B C, (C D) B
(ii) (B A), A C, B C
B17. Knaves always lie, knights always tell the truth, and in Camelot, where everybody

is one or the other, you encounter three people, Lancelot, Arthur and Merlin, who say to
you:
Lancelot: Merlin's a knave.
Arthur: Either Lancelot or Merlin is a knave.
Merlin: If I'm a knave, they are too.
Use the tree method to determine as much as you can about each person's identity.
B18. We return for one last visit to the land of Camelot where everyone is either a

knight (always speaking the truth) or a knave (always uttering falsehoods). Sir Lancelot
is searching for his mistress Queen Guinevere, and happens upon King Arthur and his
band of merry men. When Lancelot asks of Guineveres whereabouts, Arthur becomes
jealous and is in no mood to give Lancelot a straight answer. So he instructs Merlin to
cast a spell upon his men so that each, in turn, responds to Lancelot as follows:
Sir Karl the Pauper: Guinevere is in Camelot today.
Sir Loin of Beef: Sir Karl is a knight, but Sir Rob is most certainly a knave.
Sir Rob of Cliff Town: Hey, Im a knave if and only if Sir Loin is!
Sir Lee Fellow: Yah! If any of us are knights, Sir Rob is.

64
Does Arthur succeed in hiding Guineveres present whereabouts, or do his men
inadvertently disclose her location to Arthurs rival in love? Use the tree method to find
out.
B19. A certain island is populated entirely by heroes and scoundrels; the former always

tell the truth, the latter invariably lie.

(a) You encounter four people on the island who say to you:
Dean: If Im a hero, so is Stan.

Stan: If Im a hero, so is Jerry.

Jerry: If Im a scoundrel, Ollie isnt.

Ollie: Those others are all liars!

Determine as much as you can about their individual identities.

(b) Believe it or not, I once went to the Island myself in search of buried treasure. I
dont remember the details too clearly, but I do recall encountering Dean and Jerry.
Dean, I remember, told me:
Jerry is a hero and there is buried treasure on the island,
but I cant quite recall what Jerry said. All I remember is that he said exactly one of the
following:
Dean is a scoundrel and there's no buried treasure on the island
or

Dean is a scoundrel and there is buried treasure on the island.

Nevertheless, I do remember being able to figure out whether treasure was buried on
the island and the identity of both speakers. What were their identities? Was treasure
buried on the island?
(c) Suppose you were to go to the Island of Heroes and Scoundrels and wished to find
out whether or not there is gold on the island. You meet Dean (not knowing his

65
identity) and you are allowed to ask him only one question, which must be answerable
by Yes or No. What question could you ask him that would allow you to figure out if
there is buried treasure on the island? (This ones tricky and there may be more than
one question that could do the job.)
B.20 Finally, here's a toughie. On a certain island, rumoured to contain buried treasure,

live three gnomes, identical in appearance, of whom it is known that one invariably tells
the truth, one always lies, and the third answers "yes" or "no" at random. You arrive on
the island and, encountering the three gnomes, ask them a total of two questions, each
addressed to one gnome at a time, and to which the answer is a simple "yes" or "no".
What questions would you ask that would allow you to figure out if there is buried
treasure on the island? (Hint: the answer to the first question must enable you to
"eliminate" the gnome who answers at random.)

66

IV. PREDICATE LOGIC

1. Predicates, Relations and Quantifiers

The concept of validity (which we shall call propositional validity) that we have
employed up to now is restricted in that it does not cover a large class of arguments
which are clearly logically correct. Consider, for example, the following argument:
1. All Cretans love all animals.
2. All horses are animals.
3. Epimenides is a Cretan.

4. Someone loves all horses.

This argument, while patently not propositionally valid, is still, given the usual reading
of the terms "all" and "some", logically correct. Its correctness also of course derives
from our grasp of the grammatical structure of the statements constituting it, which
involve in an essential way predicates "(is a) horse", "(is an) animal", "Cretan"and
relations "loves".

In order to symbolize this argument and others like it we need to enlarge our
logical vocabulary. Thus, as in algebra, it is natural to introduce variables x,y,z,... to refer
to arbitrary individuals, and then to write, for example, "Ax" for "x is an animal", "Cx"
for "x is a Cretan", "Hx" for "x is a horse", "Lxy" for "x loves y", and "e" for "Epimenides".
The symbols A, C and H are predicate symbols, L is a relation symbol, and e a name. Finally
we introduce two symbols and called the universal and existential quantifier,

67
respectively: the expression "x" will symbolize the phrase "for all (or any) x", and "x"
the phrases "for some x", or, equivalently, "there exists x".
To put our argument in symbolic form, we first write it in the following way:
1'. For any individual x, if x is a Cretan, then for any individual y, if y is an animal, then x loves
y.
2'. For any individual x, if x is a horse, then x is an animal.
3'. Unchanged.

4'. For some individual x, for all individuals y, if y is a horse, then x loves y.
Now 1'- 4' can be symbolized directly in terms of our enlarged logical vocabulary thus:
1". x[Cx y(Ay Lxy)]
2". x(Hx Ax)
3". Ce
4". xy(Hy Lxy).
The logical system associated with the enlarged vocabulary of variables,
predicate and relation symbols, names, and quantifiers is called predicate or
quantificational logic.

Statements formulated within this vocabulary will be called

predicate or quantificational statements.

2. Tree Rules for Quantifiers

In order to be able to employ the tree method to test arguments within predicate
logic (such as the one above) for validity we need to formulate new tree rules governing
the quantifiers. In the case of we shall be guided by the usual meaning of generality,
namely, that whenever we assert that all individuals under consideration have a certain
property, then, given any individual, that individual has, or, as we shall sometimes say,
instantiates the property. We call this the principle of universal instantiation.

corresponding tree rule may be formulated thus:

The

68

Given a statement of the form vp(v) occupying a node of an open path

of a tree,
(1) if a name n appears in the path, write p(n) at its foot unless that statement
already occupies a node of the path (in which case, writing p(n) once more
in the path would be redundant);
(2) if no name appears in the path, choose some name n and write p(n) at its foot.
Do not tick the line vp(v).
UI.

In writing p(v) we have indicated that the statement p contains an occurrence of the
variable v; this done, we have written p(n) for the result of substituting "n" for "v" at
each occurrence of the latter in p.
Let us observe this rule in action. Consider the argument:
1. Juliet loves all who love Romeo.
2. Romeo loves himself.

3. Juliet loves herself.

The argument may be symbolized as follows, using "r" as a name for Romeo, "j" for
Juliet, and writing "L" for "loves":
1. x(Lxr Ljx)
2. Lrr
3. Ljj.

Strictly speaking, by "occurrence of v" here we mean free occurrence, that is, an occurrence of v not
within a context of the form "vq(v)" or "vq(v)". We shall always tacitly assume that this is the kind of
occurrence in question.

69
As usual, we start off with the premises of the argument followed by the negation of its
conclusion, and then continue so as to obtain a closed tree in the following way:
1.

x(LxrLj x)

2.

Lrr

3.

Lj j

4.

LrrLj r

5.

Lrr

6.

7.

(UI applied to 1)
Lj r

(from 4)

Lj rLj j (UI applied to 1 again)

Lj r

Lj j (from 6)

In this example we used the UI rule twice to obtain lines 4 and 6:

1. x(Lxr Ljx)
4. Lrr Ljr
6. Ljr Ljj

xp(x)
p(r)
p(j)

Both applications were made to the same node, 1, and in both the variable v was "x",
and p(v)that is, p(x)the statement "x(Lxr Ljx)". The two applications differed,
however, in respect of the name substituted for x: in the first case it was "r" and in the
second "j". In the first case we obtained p(r) by substituting "r" for "x" in p(x), and in the
second p(j) by substituting "j" for "x" in p(x).
From the fact that we had to apply UI twice to the same statement 1. it should
now be apparent why we do not tick a statement to which UI has been applied. Indeed,
in this example we had to continue to apply it with every name actually appearing in
the path in question before the path (and the tree) finally closed.

70
Let us now consider an example of an application of UI in which no names are
initially given.

Here the tree method will be used to test satisfiability rather than

validity. Consider the conditions:

All unicorns are fleet.

No unicorns are fleet.

Using the obvious notation, these apparently conflicting hypotheses concerning

unicorns are expressible as the statements occupying the first two nodes of the
following tree, which tests their joint satisfiability:
x(UxFx)
x(UxFx)
UaFa
UaFa
Ua
Ua

Fa
Fa

Ua

Fa

The third node here results by applying UI to the first node, at the same time
introducing the new name a. Once this name has been introduced into the path, it must
be used in any subsequent application of UI in that path, in particular, in the
application yielding the fourth line from the second.
We note that the tree is finished since no further applications of UI can be made,
and it has 3 open paths. Each of these open paths may be regarded as representing a
possible domain or universe of discourse in which all the statements occupying lines in it
are true. In general, the objects constituting the domain associated with an open path
correspond to the names appearing in that path. In our example, there is only one such
name"a"present, so that each domain of discourse has exactly one element, which
we take to be named by "a". Since the statement Ua occurs in each path, the statement

71
"a is not a unicorn" holds in each domain of discourse. In the second open path the
statement Fa appears, so the statement "a is not fleet" holds in the associated domain
of discourse. The third open path contains the statement Fa, so "a is fleet" holds in the
associated domain of discourse. On the other hand, the first path contains neither Fa
nor Fa, so in the corresponding domain of discourse a can be fleet or not indifferently.
In fact, since the object named by "a" is the sole individual in each domain of discourse,
we see that in each of these contexts the statement Ua has the stronger meaning that
nothing is a unicorn. Thus each domain of discourse represents a "world" in which no

unicorns exist, so that any assertion about all unicorns, including our two conditions
above, automatically come out true, and are therefore jointly satisfiable there (contrary
to what one might naively expect).
We turn now to the existential quantifier . First, we note that there is a simple
connection between and . To see what it is, imagine that we have a domain of
discourse consisting of three people, named by a, b, c, say. Consider the two statements
Someone (in our domain of discourse) is Canadian
Everyone (in our domain of discourse) is Canadian

xCx
xCx

It is clear that in our domain of discourse the statement xCx is equivalent to the
disjunction
Ca Cb Cc
and the statement xCx to the conjunction
Ca Cb Cc.
Therefore the negated statement xCx is equivalent to
(Ca Cb Cc),

72
which by de Morgan's law is equivalent to
Ca Cb Cc.
But this last statement asserts that each, and so every, individual in our domain of
discourse satisfies C; in other words, it is equivalent to the statement xCx.
It is evident that the correctness of this line of reasoning is independent both of
the nature of the predicate C and of the number of indviduals in the domain of
discourse. Thus we may draw the general conclusion that, for any statement p(v),
writing for equivalence as usual,
vp(v) vp(v).
An analogous argument shows that also
vp(v) vp(v).
Thus, in our example above, negating the statement "someone is Canadian" is
equivalent to asserting "everyone is nonCanadian" and negating the statement
"everyone is Canadian" to asserting "someone is nonCanadian".
All this justifies the following
RULE FOR NEGATED QUANTIFICATION

If a statement beginning with v (or v)

occupies a node of an open path, tick
it and write at the feet of all open paths
containing that node the same statement
with v in place of v (or with v in place
of v) in front.

v p(v) v p(v)
|
|
v p(v)
v p(v)

73
We now require a rule for the existential quantifier. This is the rule of existential
instantiation:

EI
Given an unticked statement of the form vp(v) occupying a node of an open
path, check to see whether it contains a node occupied by a statement of the form
p(n). If not, choose a name n that has not been used anywhere in the path and write
the statement p(n) at its foot. When this has been done for every open path in
which the statement vp(v) occupies a node, tick the node occupied by the given

statement:
vp(v)
|
p(n) (n new)
It is important to observe in applying this rule that the name n introduced not be already
present in the path. This is imperative because we want n to name an individual about
which we assume nothing except that it satisfy p; individuals that have already been named

may have properties that conflict with this supposition. For example, consider the
following (true) premises:
Someone is Canadian
Nixon is not Canadian

xCx
Cn.

Were we allowed to use the old name n instead of being forced to introduce a new one,
we would be able to generate a closed tree from these premises:
xCx
Cn
Cn

where we have (incorrectly!) applied EI to the first node to obtain the third. This would
mean that the premises are not jointly satisfiable, in other words, that from the assertion
"someone is Canadian", we would be able to infer "Nixon is Canadian". Using the same

74
line of reasoning, we would in fact be able to infer "everyone is Canadian". Incorrectly
applied, EI can lead to absurdities such as these.
Correctly applied, on the other hand, EI leads in our example to

xCx
Cx
Ca
where a is a new name, denoting, as it were, an "archetypal Canadian", whose identity is
not further specified.
Armed with these new rules for quantifiers, let us return to the argument with
which this chapter began, and see if the associated tree closes. Here it is:

75
x[Cxy (AyLxy)]
x(HxAx)
Ce
xx (HyLxy)
Cey (AyLey)
Ce

y (AyLey)
xy (HyLxy)
y (HyLey)
y (HyLey)
(HaLea)
Ha
Lea
HaAa
Ha

Aa

Aa Lea
Aa

Lea

It does close.
3. Identity

We frequently need to assert that two names refer to the same, or different, things,
as, for instance, in the (correct) argument

1. Yesterday I was home

2. Monday I was out.
3. Yesterday was not Monday.

76

Writing "a" for "yesterday", "b" for "Monday", and "Hx" for "I was home on day x", we
still lack a way of symbolizing statement 3. We rectify this by introducing the symbol
"=" called the identity or equality symbol, which we agree is to be written in between
variables or names, as in x = y, n = x or m = n. Similarly, we introduce the diversity or
inequality symbol , so that the statements x y, n x, m n serve as abbreviations for

(x = y), (n = x), (m = n), respectively.

Now our argument may be symbolized
Ha
Hb
ab
If we negate the conclusion of this argument so as to investigate its validity in the
customary arboreal fashion, we start off with
Ha
Hb
a=b
Now if a and b are truly identical, then any property possessed by a should be shared
by b (and vice versa), so from the first and third of these statements it should be
permissible to infer

Hb

As a result the tree will close.

This idea leads us to introduce the following

77
RULE FOR IDENTITY

If an open path contains a full line of the form m = n and also a full line p in which one
of the names m, n appears one or more times, write at the foot of the path a statement q
obtained by replacing one or more of the occurrences of that name in p by the other
name, provided that q does not already appear in that path as a full line:
m=n
p

|
q

It is also a characteristic feature of identity that every individual is self-identical.

We formulate this idea as a tree rule by closing any path which terminates with a
statement of the form n n. Thus we introduce the

RULE FOR NONIDENTITY (OR DIVERSITY)

Close any path that contains a line of the form n n:

n n

These rules enable us to establish the four basic laws of identity, viz., substitutivity,
reflexivity, symmetry, and transitivity.
Substitutivity

p(a)
a=b
p (b)
The validity of this inference is confirmed by the fact that the following tree closes:

78

p(a)
a=b
p(b)
p(b)

where we have used the rule for identity to obtain the fourth line from the first two.
Reflexivity

a=a
The validity of this inference follows immediately from the rule for nonidentity.
Symmetry

a=b
b=a
The correctness of this inference follows from the closed tree
a=b
ba
bb

in which the third statement is obtained from the first two by the rule for identity, and
closure from the rule for nonidentity.
Transitivity

a=b
b=c
a=c
The correctness of this inference results from the fact that the following tree is closed:

79
a=b
b=c
ac
ab

where the last statement arises from the second and third by the rule for identity.
4. Validity and Interpretations

The reader will observe that we have been calling an argument formulated in
predicate logic valid, or correct, if the tree beginning with the argument's premises and
negated conclusion closes. In consonance with this, we call a predicate statement valid if
the tree starting with its negation closes. Further, we say that a set of predicate
statements is satisfiable or consistent if any tree starting with that set of statements
contains an open path. Now these are, of course, purely formal definitions in which, by
contrast with the corresponding definitions for propositional logic, the concepts of truth
and falsity do not figure. Nevertheless, with a little insight it is possible to reformulate
these definitions for predicate statements in terms of truth and falsity. We give a brief
outline.
The key idea we require is that of an interpretation of the vocabulary of predicate
logic, a concept which generalizes that of a valuation of statement letters. For this
purpose we shall allow our logical vocabulary to include, in addition to names and
predicate symbols, relation symbols linking an arbitrary number of individuals: thus
relation symbols may be binary, ternary , ... , n-ary. An interpretation I of our logical
vocabulary now consists of:
(1) A nonempty set A called the universe of I;
(2) an assignment, to each name, of a definite element of A (which we shall call the
interpretation under I of that name);

80
(3) an assignment, to each n-ary relation symbol, of a definite n-ary relation among the
elements of A (which we shall call the interpretation under I of that relation).
Once an interpretation of our logical vocabulary has been fixed, it becomes possible to
assign a definite truth value to each predicate statement based on that vocabulary by
giving the identity symbol, the logical operators, and the quantifiers, their natural
meanings. This is best conveyed by an example.
Suppose our vocabulary contains one binary relation symbol L and one name a.
Consider the interpretation I whose universe is the set of natural numbers {1,2,3,...}, a is
interpreted as the number 1, and L is interpreted as the "less than" relation <. Let us
determine the truth values of the statements
(1) xLax (2) xLax (3) xyLxy

under I. To do this we use I to "translate" these statements into assertions possessing

truth values as follows. The "translation" of (1) is:
"There is a number x such that 1 < x"
which is obviously true. The interpretation of (2) is
"For all numbers x, 1 < x",
which is clearly false since it is not the case that 1 < 1. Finally, the interpretation of (3) is
"For any number x, there is some number y such that x < y",
which is obviously true.
This example indicates the method by which any interpretation can be made to
yield a definite truth valuetruth or falsityto each predicate statement p: we call this
the truth value of p under I.

81
Now we can define an argument A to be valid if, for any interpretation I,
whenever all the premises of A are true under I, so is its conclusion. A statement p is
valid if it is true under any interpretation. And a set of statements S is satisfiable if there

is an interpretation under which all the statements in S are true.

It can be shown that the two definitions of validity, that based on trees and that
based on interpretations, are equivalent. This is also the case for satisfiability.
5. Many-Sorted Logic

In English (and other languages) there are different quantifiers for different types
of domain, for example, various universal quantifiers.

Domain

Places

Quantifier Everywhere

Times

People

Things

Always

Everyone

Everything

It is convenient to introduce similar devices into our formal logical notation. The
method is best illustrated by an example.
Consider the following vocabulary:
Px: x is a person
Qx: x is a politician
Tx: x is a time
Fxyz: x can fool y at (time) z
Then the statement
There is someone who can fool only himself and all politicians all of the time.

may be symbolized in our customary notation as

x[Px y[Py [z(Tz Fxyz) (x = y Qy)]]].

82
This rather involved expression may be simplified by introducing different sorts of
letter to indicate individuals satisfying P ("persons") or T ("times"). Thus, if we agree to
use letters x,y for persons, and letters t,u for times, the statement above assumes the
simpler "many sorted" form:
xy[t[Fxyt (x = y Qy)]].
The advantage here is that we no longer need to employ explicit predicates to
restrict the "range" of the variables. Notice that in order to transcribe this many-sorted
statement back into its original "one-sorted" form we need to replace "x" by "x(Px
...)", "y" by "y(Py ...)" and "t" by "z(Tz ...)".
6. Functions

Relationships such as motherhood or fatherhood have the property that each

individual determines a specific, indeed unique, individual (one's mother or father,
respectively) with respect to which it stands in that relationship. The introduction of
devices called function symbols into our vocabulary will enable us to give symbolic
expression to this fact.
Thus consider, for example, the relation M of motherhood on the domain
of discourse consisting of all persons. We introduce the function symbol m to stand for
"mother of", so that mx is to be read "mother of (person) x". Then there are two
equivalent ways of expressing the statement "y is the mother of x", viz.,
Mxy and y = mx.
Thus, for example, if "a" names Liza Minnelli, then "ma" names Judy Garland.
Names and variables are noun-like terms, and function symbols may be applied
to terms of this sort to yield new terms. Thus we may write, for example,
mmx = mother of mother of x = maternal grandmother of x.

83
Similarly, if in addition we introduce the function symbol f for "father of", then
mfx = mother of father of x = paternal grandmother of x
etc.
In general, we may introduce a function symbol in connection with a relation R
precisely when R has the two following properties:
Existence: for any x, there exists y such that Rxy
Uniqueness: for any x,y,z, if Rxy and Rxz, then y = z.
When these conditions are satisfied, then for any x there is a unique y such that Rxy, and
so we can introduce a function symbol f with the meaning that, for any x, fx denotes this
uniquely determined y. Thus, for any x and y, the following conditions are equivalent:
y = fx and Rxy.
Function symbols may also be employed in trees, where such terms as fa, fma,
etc. are counted as names. However, in doing this we must at the same time insist that
when the EI rule requires us to introduce a new name, it must be a simple one, i.e. a
new letter not already used. To illustrate, we establish the validity of the inference
x (fa = x)
x (fx = a)
(An example of this form of argument in English is: Everybodys Adams father,
therefore Adams everybodys father.) The tree for this inference is
x (fa = x)
x (fx = a)
x (fx a)
fb a

84
fa = fb
fa = a
fb = a

The tree is closed and the inference valid. Notice that in the fourth line EI was applied
to the third line, introducing a new letter "b". Also notice that both the names fb and a
have been substituted in for the variable x in the first line.
7. Exercises
Trees and Translations Involving Quantifiers
A1. Using the tree method, determine which of the follwing inferences are valid.

(i) x(Px Qx)

xPx xQx
(iv) xPx xQx
x (Px Qx)

(ii) xPx xQx

x(Px Qx)
(v) x(Px Qx)
xPx xQx

(iii) x(Px Qx)

xPx xQx
(vi) xPx xQx
x(Px Qx)

A2. Symbolize the following arguments, and determine whether they are valid (always

try to choose natural predicate letters and name letters!):

(a) All logicians are neurotic. No vegetarians are neurotic. Therefore, no vegetarians
are logicians.
(b) Every Greek who loathes a Trojan is feared by all. Achilles loathes a Trojan, so, if
Achilles is Greek, everybody fears him.
(c) Alma has a brother who has no brother, so she's no one's brother.
(d) I'll be home before four o'clock. Therefore there's a time before four o'clock that I'll
be home before.
A3. Use the tree method to determine whether the following argument is valid:

85
xFxxGx
xy(FxGy)
A4. Using the following translation key:

Ex = x is an epic

Sxy = x is shorter than y Wxy = x wrote y

b = Beowulf

c = the Odyssey

d = Homer

symbolize each of the following assertions:

(i) The Odyssey is an epic.
(ii) Homer didnt write Beowulf.
(iii) The Odyssey and Beowulf are not the same length.
(iv) Homer wrote an epic which is longer than Beowulf.
(v) Of all the epics that there are, Homer wrote the longest.
A5. Symbolize the following arguments (using the given symbols), and determine

whether they are valid using any reliable method:

(a) If anyone can learn physics, you can. Anyone who can learn logic can learn physics.
Dr. Rob can learn logic. So you can learn physics! [Use Px: x can learn physics; u:
you; Lx: x can learn logic; d: Dr. Rob]
(b) No acrobats are clumsy. Therefore, if Alma is a waiter, then if all waiters are
clumsy, Alma is no acrobat. [Use Ax: x is an acrobat; Cx: x is clumsy; Wx: x is a
waiter; a: Alma]
(c) All dogs are cats. Therefore, whoever loves a dog loves a cat. [Use Dx: x is a dog;
Cx: x is a cat; Lxy: x loves y]
A6. Symbolize the sentences in the following arguments and, using the tree method,

determine which ones are valid.

86
(i) Everything has a cause. If the world has a cause, then there is a God. Hence, there is
a God.
(ii) If everyone litters, the world will be dirty. Hence, if you litter, the world will be
dirty.
(iii) All love all lovers. Romeo loves Juliet. Therefore, I love you.
(iv) Any barber in Seville shaves exactly those men in Seville who do not shave
themselves. Hence, there is no barber in Seville.
A7. Symbolize the following (sets of) sentences and, using the tree method, determine

in each case whether they are jointly satisfiable.

(i) Any reasonable person can understand logic and is fit to vote. But Joe doesnt
understand logic and yet is fit to vote.
(ii) There is a barber who shaves exactly those who do not shave themselves.
(iii) Everybody loves all lovers. You love yourself, but you dont love me.
(iv) There are at least three objects in this box, and exactly one out of every two is black.
A8. Translate into logical notation:

(a) Some know all.

(b) Some know all who know them.
(c) Some know all who know themselves.
(d) All who know some know some who know all.
(e) No one who knows someone Alma knows, knows all who know Alma.
(f) All who know everyone Alma knows know some who know Alma.
A9. Using the tree method, determine which of the following arguments is valid:

87
(a) There is someone who is going to pay for all the breakages. Therefore, each of the
breakages is going to be paid for by someone.
(b) No student in the statistics class is smarter than every student in the logic class.
Hence, some student in the logic class is smarter than every student in the statistics
class.
(c) Any person who is not mad can understand logic. None of Wagners sons can
understand logic. No mad persons are fit to vote. Therefore, none of Wagners sons is
fit to vote.
A10. Translate the following into logical notation (use:- Lxy = x likes y, m = me or I):

(a) I like myself.

(b) Someone likes me.
(c) No one likes me.
(d) Everyone likes someone.
(e) I like myself and no one else.
(f) Someone likes everyone who likes me.
(g) I dislike anyone who dislikes me.
(h) Someone likes everyone I dislike.
(i) Someone who likes me likes everyone.
(j) Everyone who likes someone I like likes no one I dislike.
A11. Translate the following into logical notation

[Assume the domain is persons, and use: Txy = x is taller than y, r = Rob):
(a) Rob is taller than everyone.

88
(b) Everyone is taller than someone.
(c) Noones taller than everyone.
(d) Everyones taller than everyone else.
(e) Rob is taller than no more than one person.
(f) No two people are taller than eachother.
(g) If anyones taller than Rob, Rob is.
(h) Someone is taller than everyone Rob isnt taller than.
(i) Someone taller than Rob is taller than everyone taller than Rob.
(j) Everyone taller than someone is taller than someone taller than everyone.
A12. Demonstrate the validity or invalidity of each of the following two arguments by

first translating them using the given symbols, and then doing their trees.
For translation, use:
Cx = x is a chimpanzee

Bx = x will get a banana

f = fred

Sxy = x can solve y

Txy = x is trying harder than y

b = barney

Px = x is a problem
(i) Not all chimpanzees are trying equally hard. No chimpanzee tries harder than
himself. Therefore there are at least two chimpanzees.
(ii) Fred and Barney can solve exactly the same problems. If Fred can solve even one
problem, then he will get a banana. Fred will not get a banana. Therefore Barney can't
solve any of the problems, and he won't get a banana either.
A13. Using the tree method, determine which of the following pairs of statements are

equivalent:

89
(i) x Px xQx

xx (Px Qy)

(ii) xPxxQx

yx (Px Qy)

A14. Define !x by writing !xPx for x[Px y(Py y=x)]. State, in simple language,

the meaning of !xPx.

Determine which of the following inferences are valid:
(i) _________
x!y (x = y)
(ii)

!x Px

xy(Pyy=x)
(iii) !x(Ax Bx)
!xAx !xBx
(iv) xy (x = y)
!x (x = x)
A15. Demonstrate the validity or invalidity of each of the three arguments below by

first translating them using the given symbols, and then doing their trees.
(a) Tweety bird despises cats.

No cats despise Tweety bird.

Sylvester is a cat.

Therefore, Tweety bird despises someone who despises him. (use:- Dxy: x despises y; t:
Tweety bird; Cx: x is a cat; s: Sylvester)

(b) Any good logic teaching assistant helps all and only those who dont help
themselves. Hence there arent any good logic teaching assistants! (use:- Gx = x is a
good logic teaching assistant; Hxy = x helps y)

(c) I'll finish this exam before four oclock. For any pair of times, one later than the
other, there is a time in between them. So there's a time before four oclock that I'll
finish this exam before. (use:- Ex = I'll finish this exam at time x; x<y = time x is
earlier than time y; f = four oclock)

90
A16. Translate the following into logical notation (use:- Lxy: x loves y):

(a) Some love anyone.

(b) No one loves all who love them.
(c) Some love all who love themselves.
(d) All lovers love some who love all.
(e) Nobody who loves somebody loves somebody who loves nobody.
(f) All who love all love all lovers.
A17. Demonstrate the validity or invalidity of each of the three arguments below by

first translating them using the given symbols, and then doing their trees.
(a) Ben loves cats. No cats love Ben. Whitey is a cat. Therefore, Ben loves someone
who doesnt love him. (use:- Lxy: x loves y; b: Ben; Cx: x is a cat; w: Whitey)
(b) Theres a set containing all and only those sets which are not members of
themselves. Therefore, every set is a member of itself. (use:- Sx = x is a set, x y = x is
a member of y)

(c) Everyone loves lovers. Romeo loves Juliet. So Fred loves Wilma. (assume domain
= persons and use:- Lxy = x loves y x; r = Romeo; j = Juliet; f = Fred; w = Wilma)
Interpretations and Counterexamples
B1. Translate the following arguments into logical symbols and determine whether they

are valid. If not, specify one counterexample.

(i) Theres something thats tasty if its a chocolate bar. So theres a tasty chocolate bar.
(ii) Some like it hot and some dont. Those who like it hot like Marilyn Monroe, and
those who dont dont like her! Therefore, everybody either likes her or doesnt.

91
(iii) If anyone is taller than Rob, Gurpreet is. If Gurpreet is taller than Rob, anyone is.
So it isnt the case that theres someone taller than Rob and someone not.
B2. Consider the domain consisting of points and straight lines in a given plane, with

the following vocabulary for describing it:

Px = x is a Point

Lx = x is a straight Line

Oxy = (point) x lies On (line) y

Determine the truth value of each of the following statements, providing a brief
justification of your answer in each case.
(a) xy[(LxLy)z(PzOzxOzy)]
(b) xy[(PxPyxy)zw[(LwOxwOyw)z=w]]
B3. Knaves always lie, knights always tell the truth, and in Camelot, where everybody

is one or the other, you encounter some people, among them King Arthur who says to
you:
"Exactly one out of every two of us is a knave"
Choose names for any other people you might need to refer to and specify an
interpretation (i.e. case) in which Arthur is a knight (if indeed its possible for him to
be a knight, given what he says). Also, specify an interpretation in which Arthur is a
knave (again, only if thats possible).
B4. Consider the following scenario involving three objects, two predicates F and G, and

a relation R:
Domain = {1,2,3}

F = {2,3}

G = {1}

R = {(1,1),(2,2),(1,3),(2,3),(3,3)}

Which of the following is true, and which is false, under this scenario?
(i) x(FxGx)
(ii) x(FxGx)

(iv) xyRyx
(v) yx(RxyGx)

92
(iii) x(FxGx)
B5. Consider the domain consisting of the positive whole numbers {1,2,3,...} with the

following vocabulary for describing it:

Ex: x is even

Ox: x is odd

x>y: x is greater than y

x<y: x is less than y

xy: x is unequal to y

Determine the truth value of each of the following statements, providing a brief
justification of your answer in each case.
(a) x[Oxz(Ez(x>z))]
(b) xy((xy)(x<y))
B6. Using the tree method, or otherwise, determine whether the following sets of

statements are satisfiable. For each of the satisfiable sets, supply an interpretation in
which all the statements are true.
(a) xyPxy
xyz(Pxz Pxy)

(b) xyPyx

(c) xyz ((Pxy Pyz)Pxz)

xy(Pxy Pyx)
xyz((Pxy Pyz) Pxz)

x Pxx
xy(Pxy Pyx)

B7. Symbolize the following arguments, and determine whether they are valid using

any reliable method. If not, give a counterexample.

(a) A person is famous if and only everyone has heard of him or her. So, all famous
people have heard of each other.
(b) Tweety bird hates cats. No cats hate Tweety bird. Sylvester is a cat. Therefore,
Tweety Bird hates someone who hates him. [use:- Hxy: x hates y; t: Tweety bird; Cx: x
is a cat; s: Sylvester)
(c) Logic students are taller than business students. Exactly one out of every pair of
students is a logic student. So some student is taller than some other.

93
B8. Symbolize the following arguments, and determine whether they are valid using

any reliable method. If not, give a counterexample.

(a) Exactly one professor lives in Talbot College. Professor Bell lives in Talbot College.
Professor Bell is a toposopher. Therefore, every professor who lives in Talbot College is
a toposopher. [For translation, assume domain is professors and use: Cx = x lives in
Talbot College, b = Bell, Tx = x is a toposopher.]

(b) Exactly one out of every pair of balls is red. Exactly one ball is red. So exactly one
ball isnt red. [Assume domain is balls.]
(c) Stallone can outgun everybody who can outgun anyone he can. Therefore, Stallone
can outgun himself and noone else! [For translation, assume domain is persons.]
B9. Here is a small world:

and symbols for describing it (with any variables restricted to ranging over the above
nine inhabitants):
Domain = The shapes with names a, b, c, d, e, f, g, h, i
Sx = x is a square

Lxy = x is directly left of y Cxy = x is in the same column as y

94
Tx = x is a triangle Axy = x is directly above y Rxy = x is in the same row as y
Bx = x is black
Interpreting each of the formulae below as a statement about this world, state whether
it is true or false of the world, and, if it is false, briefly state why (referring to any of the
shapes above by name, if you need to):
(i) xTx
(ii) xLxx
(iii) x(TxSx)
(iv) x(TxBx)
(v) yxLxy
(vi) xy(TxTyCxy)
(vii) x[Txy((Rxyxy)Ty)]
(viii) x((SxyAxy)Bx)
(ix) xy((SxSyxy)(Cxy(BxBy)))
(x) yx(BxSxxy)
B10. Using the indicated key: symbolize (1) through (4); translate (5) through (8) into

clear English (not just logical jargon); and say whether each of these statements (1)-(8) is
true or false, briefly justifying your answer.
Key
Domain: statements

Ixy = x logically implies y Exy = x is logically equivalent to y

(1) Every statement implies some statement or other.

95
(2) Some statements are equivalent to anything that implies them.
(3) Statements with the same implications are equivalent.
(4) Some statements imply all and only what implies them.
(5) xyIxy
(6) xy(IxyExy)
(7) xy(IxyExy)
(8) xy(IxyzIyz)
B11. Here is a small world:

Left

Right

and a vocabulary describing it (variables restricted to the five inhabitants):

Fx : x wears a feather

Lxy: x is left of y

Gx : x wears glasses

Rxy: x is right of y

Hx: x wears a hat

Txy: x is taller than y

Ixy: x is identical to y

96
Which of the following quantified formulae are true and which are false of this small
world? If a formula is false, describe a minimal change of the world that would make it
true (e.g. take somebodys feather away, move people around, etc.. but dont move
anybody into or out of the world).
1. x(Fx yRxy)

4. x(Gx y(Gy Rxy)

2. xy (IxyLxy)

5. y(x (Fx Lyx) Hy)

3. xy [Ixy (Txy Tyx)]

B12. Here is another small world:

Domain = {Mum, Pop, Junior}

F = {Pop, Junior}
G = {Mum}
R = {(Pop, Pop), (Mum, Mum), (Mum, Junior), (Pop, Junior)}
Which of the following is true, and which false, when interpreted as assertions about
this world:
(a) x(Fx Gx)
(b) x(Gx Fx)
(c) xy (Fx Gy)
(d) x[y Ryx Fx]
(e) x[yGy Fx]
(f) xy Rxy
(g) x(yRxy zRxz)
(h)xy(Fx Gy)

97
(i) x(Fx Gx yRxy)
(j) xyRxy xyRxy
(k) yx(Fx Gy)
B13. Using any method you like, provide, for each of the following statements, an

interpretation (i.e. case, scenario) which makes it true, and one which makes it false.,
including the domain.
(a) xyzRzyx
(b) xyRxyyxRxy
(c) yz(PzzyQy)
B14. Here is a small world (a quack optometrists eye chart!):

and symbols for describing it (with any variables restricted to ranging over the above
eleven inhabitants):
Dx: x is the letter d (or D)

Cx: x is capitalized

Ex: x is the letter e (or E) Bx: x is bold-faced

Fx: x is the letter f (or F)

Rxy: x is right of y
Lxy: x is directly left of y

Ux: x is underlined
Ix: x is italicized

Interpreting each of the formulae below as a statement about this world, state whether
it is true or false of the world:
(i) xUx
(ii) x(BxIx)

98
(iii) x(FxBx)
(iv) x[Bxy(RyxUy)]
(v) xy[Lxy(UxUy)]
(vi) x(CxyLyx)
(vii) x(ExCxBxy[(EyCyBy)y=x])
(viii) x(Dxy[(DyRyx)Cy])
B15. Here is ANOTHER quack optometrists eye chart:

E D

and symbols for describing it:

Dx: x is the letter d (or D)

Cx: x is capitalized

Rxy: x is somewhere right of y

Ex: x is the letter e (or E)

Bx: x is bold-faced

Lxy: x is directly left of y

Fx: x is the letter f (or F)

Ux: x is underlined
Ix: x is italicized

Interpreting each of the formulae below as a statement about this world, state whether
it is true or false of the world:
(i) xCx
(ii) x(UxIx)
(iii) x(FxBx)
(iv) xy(Lyx(FyBy))
(v) x[Bxy(RyxUy)]

99
(vi) xy[Lxy(UxUy)]
(vii) x(CxyLyx)
(viii) x(ExCxBxy[(EyCyBy)y=x])
(ix) x(Dxy[(DyRyx)Cy])
(x) x(yLxyzRxz)
Functions
C1. Symbolize each of the following, using f as a function symbol for the father of

and m as a function symbol for the mother of:

(i) a is bs paternal grandmother
(ii) a is a father
(iii) a is bs full sibling
(iv) a is bs grandmother
(v) a is a grandfather
(vi) a is bs first cousin.
C2. Let f and m be symbols for the functions "the father of" and "the mother of", and

read xPy as "x is the parent of y" and Mx as "x is male". For each of the statements
below, explain the precise relationship asserted between a and b as concisely as you can
in English.
(a) ab fa=fb ma=mb
(d) (fa=fb mamb) Mb

(b) aPfb

(c) ab faPb maPb Ma

(e) Mb x(xPa fb=fx mb=mx b=fa)

(f) Ma x(xPa fbPx mbPx bPa).

100

V. THE PROPOSITIONAL CALCULUS

In this chapter we describe a formal systemthe propositional calculusfor

proving propositional statements and as a result obtain a purely syntactical

characterization of valid propositional inferences and tautologies. To set up the system

we choose certain tautologies as axioms and lay down a certain rule of inference which
will enable us to construct deductions.
In what follows we shall omit the logical operator in forming statements, and the
rules governing in constructing trees. (Of course, we can always define by p q

(p q) (q p) if we wish.)
1. Axioms

The propositional calculus (PC) has as axioms all statements of the form (1)(10)
below.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)

p (q p)
[p (q r)] [(p q) (p r)]
(p q) p
(p q) q
p (q p q)
p (p q)
q (p q)
(p r) [(q r) (p q r)]
(p q) [(p q) p]
p p.

101
The sole rule of inference for PC is called modus ponens (Latin: "affirming mood"):
p, p q
q

MP

In words, from p, p q, infer q.

2. Deductions

Let S be a set of statements. A deduction from S is a finite sequence p1,...,pn of

statements such that, for any i = 1,...,n, pi is either (a) an axiom, (b) a member of S, or (c)
inferrable using MP from earlier members of the sequence, i.e., there are numbers j,k < i
such that pk is pj pi.
A deduction from the empty set of statements is called simply a proof. A
deduction (or proof) with last statement p is called a deduction (or proof) of p. We write S
| p to indicate that p is deducible from S, i.e. that there is a deduction of p from S. If S is
empty, so that p is provable, i.e. there is a proof of p, we write just | p, and call p a
theorem of PC.

Example. | p p.
The following is a proof of the statement p p.
1.
2.
3.
4.
5.

(p ((p p) p)) ((p (p p)) (p p)) (Ax.2)

p ((p p) p) (Ax.1)
(p (p p)) (pp) (MP on 1,2)
p (p p) (Ax.1)
p p (MP on 3,4)
We now prove the important

Deduction Theorem For any set S of statements and any statements p,q:

S,p | q if and only if S | p q.

102
Proof. First suppose that S | p q. Then there is a deduction D of p q from S.

Clearly, if we add the sequence p,q to D, the result is a deduction of q from S,p.
Therefore S,p | q.
Conversely, suppose that S,p | q. Then there is a deduction r1,...,rn of q from S,p
(so that q is rn). We claim that S | p ri for any i = 1,...,n.
Suppose that the claim were false. Then there is a least number k such that it is
not the case that S | p rk. There are then 4 possibilities: (1) rk is an axiom; (2) rk is in

S; (3) rk is p; (4) rk is deducible using MP from some ri and rj with i,j < k, where rj is ri
rk.
We show that in each of these 4 cases we have S | p rk. This will contradict
the assertion that the claim is false, and it must accordingly be true.
Case (1). rk is an axiom. In this case the sequence of statements rk, rk (p rk), p rk

is a proof of p rk, so that S | p rk.

Case (2). rk is in S. In this case the same sequence of statements as in case (1) is a

deduction of p rk from S.
Case (3). rk is p. Here we have | p rk by our Example above, so a fortiori

S | p rk.
Case (4). For some i,j < k rj is ri rk. Since k was assumed to be the least number for

which it is not the case that S | p rk, and i,j < k, we must have S | p ri and
S | p rj, i.e., S | p (ri rk). By axiom 2,
(p (ri rk)) ((p ri) (p rk)).
Hence, applying MP,
S | (p ri) (p rk)

103
and applying it once more,
S | p rk.
We have obtained a contradiction in each case, so the claim is true. In particular, taking i
= n, we get S | p rn, i.e. S | p q. This completes the proof.
3. Soundness

Our next result is the

Soundness Theorem for the Propositional Calculus.
Any theorem of PC is a tautology.
Proof. Note first that, if a valuation satisfies both p and p q, then it satisfies q.

Thus if both p and p q are tautologies, so is q. In other words, MP leads from

tautologies to tautologies.
It is also not hard to show that any axiom of PC is a tautology. For example, we
may use the tree method to establish this for Axiom 8:
[(p r)((q r)(p q r)]
pr
[(q r)(p q r)]
q r
(p q r)
pq
r
p
p

104

Hence any deduction in PC consists entirely of tautologies, and the theorem follows.
As an immediate consequence of this, it follows that PC is consistent in the sense
that for no statement p do we have both | p and | p.
We are next going to establish a strengthened version of the Soundness Theorem
by employing the Deduction theorem.
Strengthened Soundness Theorem for PC.

If S | p, then S |= p.

Proof. Suppose S | p, where S = {s1, s2 ,..., sn}. The trick is simply to apply the

Deduction theorem to s1, s2 ,..., sn | p and carry each of the statements in the sequence
s1, s2 ,..., sn over to the right-hand side of the | sign so that the (unstrengthened)
Soundness theorem, which weve already proved, can be invoked.
Thus, applying the Deduction theorem n times in succession to S | p yields:
| s1(s2( (sn-1(snp))))

(*)

(For example, if n=3 the first application of the Deduction theorem yields
s1, s2 | s3p,
the next application yields
s1 | s2(s3p),
and the final application yields
| s1(s2(s3p)). )
But by the (unstrengthened) Soundness theorem, we can infer from (*) that:
|= s1(s2( (sn-1(snp))))

(**)

105
(**) simply asserts that the nested conditional above that we have concocted by
applying the Deduction theorem is a tautology. But it is clear from the truth table for
that this could not be so unless there is never a case where all the statements in the set S
= {s1, s2 ,..., sn} are true and p is false. So (**) cannot be correct unless S |= p, which is
what we set out to prove.
Our final task will be to prove the converse of the Strengthened Soundness
theorem.
4. Completeness

Let us call a set S of statements formally inconsistent if S | p and S | p for

some statement p. We now establish the following facts:
Fact A. S is formally inconsistent if and only if S | q for all statements q.
Fact B. S | p if and only if {S, p} is formally inconsistent.
Proof of A. Clearly, if S | q for all statements q then S is formally inconsistent.

To establish the converse, we begin by showing that | p (p q). First, note that
the following sequence qualifies as a deduction of q from p,p:

p
p
p (q p)
p (q p)
q p
q p
(q p) ((q p) q))
(q p) q
q

106
q q
q
It follows that p,p | q. Two applications of the deduction theorem now give
| p (p q) as claimed.
Now if S is formally inconsistent, we have S | p and S | p. Since
| p (p q),
two applications of MP yield S | q. This proves A.
Proof of B. If S | p, then S, p | p and S, p | p, so S, p is formally

inconsistent.
Conversely suppose that S, p is formally inconsistent. Then by Fact A,
S,p | p. So by the deduction theorem S | p p. Now we have
p p,p p | p
as the following deduction shows:
p p
(p p) ((p p) p)
pp
(p p) p
p
Since | p p, it follows that p p | p. So, substituting p for p, we get
p p | p. But p p is an axiom, so an application of MP yields
p p |- p. But we have already observed that S | p p, so another application
of MP yields S | p as required. This proves B.
We now sketch a proof of the

107
Theorem. The initial set of statements of any closed tree is formally inconsistent.
Proof (sketch). Let us define the depth of a tree to be the length of its longest

path. Suppose that the assertion of the theorem is false. Then there is a closed tree with
a formally consistent (i.e., not formally inconsistent) set of initial statements. Among
these choose one, T say, of least depth, d say. Then T is a closed tree whose set S of
initial statements is formally consistent. We shall derive a contradiction from this.
There are two cases to consider.
Case 1: d = 1. In this case T is identical with S. Since T is closed there must be some

statement p for which both p and p are in S. Clearly S is then formally inconsistent.
Case 2: d > 1. In this case, by assumption, the set of initial statements of any closed tree

of depth < d is formally inconsistent. Now examine the statements at level 2 of T. We

claim that however these statements were obtained, we can always conclude that S is
formally inconsistent.
For example, suppose that the statements at level 2 of T arise by applying the rule to a statement in S of the form p q. Then T starts thus:
S
p

If in T we fuse S with p and expunge q as well as all nodes following it, we get a closed
tree (recall that T was assumed closed) of depth < d with S,p as its set of initial
statements. But then S,p is formally inconsistent. Similarly, S,q is formally inconsistent.
Since p q is in S, it follows that S is formally inconsistent. For if r is any statement, we
have S,p | r and S,q | r so that S | p r and S | q r. Two applications of MP and

108
Axiom 8 now yield S | p q r; but since p q is in S, MP yields S | r. Since this
holds for any statement r, S is formally inconsistent.
Similar arguments work for the other rules; in all cases we are able to conclude
that S is formally inconsistent.
We have shown that assuming the theorem false leads to a contradiction. So the
theorem is proved.
As a consequence of this, we finally obtain the
Completeness Theorem for PC. If S |= p then S | p.
Proof. If S |= p, then by inference adequacy any finished tree T associated with

the inference of p from S is closed. It follows from the previous theorem that the set S,
p of initial statements of T is formally inconsistent. Hence, by fact B , S | p.
5. Exercises
The Propositional Calculus

Axioms:
1. p(qp)

6. p(pq)

2. [p(qr)][(pq)(pr)]

7. q(pq)

3. (pq)p

8. (pr)[(qr)((pq)r)]

4. (pq)q

9. (pq)[(pq)p]

5. p(q(pq))

10. pp

Rule of Inference:

p,..., pq, ...,q (Modus Ponens)

A1. The following is a purported deduction of p from p. Verify that it is or is not by

identifying the origin of each statement in the sequence.

p, (p((pp)p)), p(pp),

109
(p((pp)p))((p(pp))(pp)), ((p(pp))(pp)),
(pp), (pp)((pp)p), pp, (pp)p,
p(pp), pp, p, p
Of course, there is a very simple deduction of p from p. What is it?
A2. The sequence 1.-14. below (see over) allegedly establishes that:

pp, q | qp
Check to see whether this is so by justifying each statement below with the words in
the initial set, modus ponens, or axiom # so-and-so (filling in the relevant axiom
number). If a particular statement cannot be justified, say so!
1. [p((pp)p)][(p(pp))(pp)]
2. p((pp)p)
3. q
4. (p(pp))(pp)
5. p(pp)
6. q(p(qp))
7. pp
8. pp
9. (pp)[(pp)p]
10. (pp)p
11. p
12. pp
13. p
14. qp
A3. There are two sequences of statements below (set aside in two separate columns),

each purporting to be a deduction from the set of statements S = {p, qr}. Identify the

110
origin of each statement in each sequence, and thus discern whether or not these
sequences really are deductions from S.
p

qr

qr

p[(qr)(p(qr))]

p(rp)

(qr)(p(qr))

rp

p(qr)

qp

A4. By the Completeness and Strengthened Soundness theorems for the propositional

calculus, each concept on the left below corresponds to one on the right and vice-versa.
Match them up.
tautology

deduction

unsatisfiable

theorem

valid argument

proof

valid argument without premises

formally inconsistent

A5. State the theorems below first in symbols and then in your own words.

(i) (Strengthened) Soundness Theorem

(ii) Completeness Theorem
(iii) What is the point of introducing the propositional calculus and proving these
theorems?
A6. Indicate whether each of the following statements is true or false.

(a) A statement deducible from its negation cannot be a theorem.

(b) Consistency of the propositional calculus follows from the completeness theorem.
(c) If p | (qp), then the pair {p,q} is formally inconsistent.
(d) The conclusion of a proof cannot be formally inconsistent.

111
(e) A theorem cannot be deduced from a formally inconsistent set of statements.
(f) Assuming every tautology is a theorem, completeness of the propositional calculus
follows from the deduction theorem.
(g) a | c implies a,b | b c for all statements a, b and c.
(h) The propositional calculus would not be sound unless it employed the modus
ponens rule.
For the last two questions, call a set of statements maximally consistent if and
only if it is formally consistent but not a subset of any other formally consistent set of
statements.
(i) No maximally consistent set of statements can contain all theorems.
(j) Every maximally consistent set of statements must contain either p or p, for any
statement p.
A7. Why does p | r imply p | (qr) for any statements p, q and r?
A8. (a) Assuming the Completeness and Strengthened Soundness theorems, prove the

Deduction theorem.
(b) Use the Deduction theorem to show directly (i.e. without explicitly constructing a
deduction sequence) that | (pp) and that p | p.
A9. (i) For any statements p, q and set of statements S, S | p and p | q implies S | q.

Why?
(ii) Let S1, S2, ...,Sn be n sets of statements, and let S be the set of statements {p1,p2,...,pn}.
Show that if Si | pi for all i=1 to n and S | p for some statement p, then S1S2 ...Sn |
p.

112
(iii) Show that if S | p and S | q, then S | (pq). (You are not allowed to assume the
completeness theorem!)
A10. (a) By relying on the Completeness and Strengthened Soundness theorems, prove

that a set S of statements is formally inconsistent if and only if S | p for all statements p.
(b) Without relying on Completeness and Soundness, show that S | p implies that the
set T=S{p} is formally inconsistent.

113
APPENDIX A
LOGIC AND THE FOUNDATIONS OF MATHEMATICS

Mathematics, the traditional science of form and quantity, and logic, the
traditional science of reasoning, are among the oldest of human intellectual endeavours.
However, it is only in the last century or so that the connections between the two have
been explicitly recognized and systematically developed, leading to the enrichment of
both. In this appendix, we shall take a look at how and why this came about, and
describe the effect that logical analysis has had on the foundations of mathematics.
Medieval scholars divided learning into two categories. The first category was
the quadrivium, or fourfold way to knowledge, comprising the mathematical arts
arithmetic, geometry, astronomy and music. The second was the trivium, or threefold
way of eloquence, comprising the verbal arts grammar, rhetoric and logic. Logic was
thus regarded by the schoolmen as an essentially linguistic discipline, having little or no
bearing on mathematics. In any case, the official scholastic view was that logic had
been perfected by Aristotle, so that any further contributions to the discipline could be
no more than embellishments on the Aristotelian edifice. Although in the seventeeth
century Leibniz had expressed the desirability of transforming logic into a universal
scientific language, little progress was made in this respect until the middle of the
nineteenth century when the English mathematicians George Boole (1815-1864) and
Augustus de Morgan (1806-1871) and the American philosopher Charles Sanders Peirce
(1839-1914) took the first steps in extending logic beyond its Aristotelian horizons.
Original as these contributions were, however, the major impetus behind the
transformation of logic was furnished by the later appearance of difficulties in the
foundations of mathematics.
Towards the end of the nineteenth century, the German mathematician Georg
Cantor (1845-1918), in the course of his mathematical research, had come to reject the

114
received idea that the actual infinite is an inadmissible concept in mathematics, and
proceeded to build a mathematical theory of infinite totatilitiesthe so-called set theory.
Cantor held that there is no difference in principle between finite and infinite sets, as is
revealed in his 1895 definition of the set concept:
By a set we understand every collection to a whole of definite, well-differentiated
objects of our intuition or thought.

The problem, of course, is to determine exactly which collections consititute legitimate

wholes. Traditionally, only finite collections were admitted to be wholes in this sense,
since, it was held, infinite collections fail to accord with the time-honoured dictum that
a whole must always be greater than any of its (proper) parts.
The curious way in which the principle that a whole must always exceed its parts
is violated by infinite collections is strikingly conveyed by a fable attributed to the
German mathematician David Hilbert (1862-1943). In Hilberts tale, he finds himself the
manager of a vast hotel, so vast, indeed, that it has an infinite number of rooms. Thus
the hotel has a first, second, ..., nth,... room, ad infinitum. At the height of the tourist
season, Hilberts hotel is full: each room is occupied. (Note that we are tacitly assuming
the existence of an infinite collection of occupants.)

Now a newcomer seeking

accommodations shows up. Alas, says Hilbert, I have not a room to spare. But the
newcomer is desperate. At that point an idea occurs to Hilbert. He telephones the
occupant of each room and tells him to move to the next one; thus the occupant of room
1 moves to room 2, that of room 2 to room 3, etc. This leaves the original occupants
housed (one to a room, as before), only now room 1 is vacant, and the relieved
newcomer duly takes possession. In this way we see that the whole set of rooms is in
some sense no larger than the part obtained by removing the first room.
The fable does not end here, however. Hilbert is about to switch on the no
vacancy sign when a vast assembly of tourists desirous of accommodation descends on

115
the hotel. A quick tally reveals that the assembly is infinite, filling Hilbert with dismay
(and the reader, no doubt, with incredulity). But now another idea occurs to him. He
tells each occupant that he is to move to the room assigned double the number of his
present one: thus the occupant of room number 1 proceeds to room number 2, that of
room number 2 to room number 4, etc. This again leaves all the original occupants
housed, only now each of the infinite set of rooms carrying odd numbers is vacant.
Thus each newcomer can be accommodated: the first in room 1, the second in room 3,
the third in room 5, etc. Clearly this procedure can be repeated indefinitely, enabling an
infinite number of infinite assemblies of tourists to be accommodated.
Hilberts tale shows that infinite sets are intriguing paradoxical but not, be it
noted, that they are contradictory.

(Indeed, if, for example, the physical universe

happens to contain infinitely many starswhich is at least possiblethen it can serve

as Hilberts hotel, with the stars playing the role of rooms.) Set theory as originally
formulated does contain contradictions; however, they arise not from the admission of
infinite totalities per se, but are rather the result of countenancing totalitites consisting of
all entities of a given kind. This is best illustrated by the infamous Russell paradox,

discovered by Bertrand Russell (1872-1970) in 1901.

Russells paradox arises in the following way. It starts with the truism that any
set is either a member of itself or not. For instance, the set of all men is not a member of
itself (since it is not a man), while the set of all possible sets is a member of itself (since it
is, presumably, a set). Now consider the set consisting precisely of all those sets which
are not members of themselves: call this set R (for Russell, of course). Is R a member of
itself or not? Suppose it is. Then it must satisfy the defining condition for inclusion in
R, i.e., it must not be a member of itself. Conversely, suppose R isnt a member of itself.
Then it fails to satisfy the defining condition for inclusion in R, i.e., it must be a member
of itself. We have thus arrived at the unsettling, indeed contradictory, conclusion that R

116
is a member of itself if and only if it isnt. We also note that whether R is finite or
infinite is immaterial to this argument.
Russells paradox has a purely linguistic counterpart due to Kurt Grelling (18861941) which is, perhaps, even more unsettling since it appears to strike at the very fabric
of language. Call an (English) adjective autological if it is true of itself and heterological if
not.

For instance, the adjectives polysyllabic, English are autological while

palindromic, French are not. Now, mischievously, we enquire as to whether the

adjective heterological is autological or not. A moments thought should reveal that
heterological is autological if and only if it isnt autological.
Another principle of set theory which occasioned much dispute when first
formulated is the so-called axiom of choice. In its simplest form, the axiom asserts that, if
we are given any collection S of sets, each of which has at least one member, then there
is a set M containing exactly one member from each set in S. We can imagine the set M
as being formed by choosing one member from each set in S. When there are only
finitely many sets in S, or if S is infinite and we have a definite rule for choosing a
member from each set in S, no difficulty is encountered. The problem arises when S
contains infinitely many sets, and we have no rule for choosing a member from each:
how can we justify the procedure of making infinitely many arbitrary choices (and
forming a set from the result)?
anecdote.

The difficulty is well illustrated by a Russellian

A millionaire possesses an infinite number of pairs of shoes, and a

corresponding number of pairs of socks. In a fit of eccentricity, he requests his valet to

select one shoe from each pair. When the valet asks for instructions as to how to
perform the selection, the millionaire suggests that the left shoe be chosen from each
pair. The following day the millionaire proposes that the valet select one sock from each
pair. When asked as to how this operation should be carried out, the millionaire is at a
loss for a reply, since, unlike shoes, there is no intrinsic way of distinguishing one sock

117
of a pair from the other. In other words, the selection of the socks must be truly
arbitrary.
One curious consequence of the axiom of choice is the so-called paradoxical
decomposition of the sphere, formulated in 1924 by the Polish mathematicians Stefan

Banach (1892-1945) and Alfred Tarski (1902-1983). In its most striking form, this states
that a solid sphere can be cut into finitely many (later shown to be reducible to 5) pieces
which can themselves be rearranged to form two solid spheres of the same size as the
original. (Of course, the use of the phrase can be cut is metaphorical, but this does not
detract from the wierdness of the result.) Note that this assertion, like the situation
obtaining at Hilberts Hotel, and unlike Russells paradox, is counter-intuitive but not
contradictory.
The perplexities surrounding the foundations of set theory are collectively
designated by historians of mathematics as the third crisis in the foundations of
mathematics. (For the record, the first two were the Pythagorean discovery of irrational
numbers c. 450 B.C. and the scandalously shaky state of the foundations of the calculus
throughout the seventeenth and eighteenth centuries.) Attempts to resolve this crisis
took several different forms, but they all required the deployment of subtle kinds of
logical analysis of mathematical concepts and reasoning, thereby occasioning the rise of
mathematical logic as a powerful new discipline.

Three principal tendencies were to emerge concerning the foundations of

mathematics and the resolution of the crisis: Logicism, Intuitionism, and Formalism.
Logicism, associated with (among others) Gottlob Frege (1848-1925), Russell and

Alfred North Whitehead (1861-1947), pivots on an attempt to reduce mathematics (i.e.

set theory) to pure logic, the formers inconsistencies being dissolved (it was hoped) by
proper formulation in terms of unimpeachable logical concepts.

In Russell and

Whiteheads approach, Cantors set theory is replaced by the theory of types, in which a

118
logical distinction is drawn between a set and its members, the former being, so to
speak, of higher type than the latter.

The paradoxes are avoided by scrupulous

observance of the so-called vicious circle principle, according to which no totality can
contain members defined in terms of itself. (For example, the totality of all sets violates
the vicious circle principle since it contains itself as a member and is therefore excluded
as a legitimate totality.) The logicist programme culminated in the publication in 1910
of the monumental, and formidably recondite, Principia Mathematica of Russell and
Whitehead.

That the term recondite is apposite here may be deduced from the

following extract from a review of the work which appeared in a 1911 number of the
Spectator:
It is easy to picture the dismay of the innocent person who out of curiosity looks
into the later part of the book. He would come upon whole pages without a single
word of English below the headline; he would see, instead, scattered in wild
profusion, disconnected Greek and Roman letters of every size interspersed with
brackets and dots and inverted commas, with arrows and exclamation marks
standing on their heads, and with even more fantastic signs for which he would
with difficulty so much as find names.

The complexity of Russell and Whiteheads theory, allied with doubts that, despite their
Herculean efforts, they had truly succeeded in reducing mathematics to pure logic,
resulted in a certain lack of enthusiasm on the part of mathematicians for the logicist
programme.
Intuitionism, the creation of the Dutch mathematician L. E. J. Brouwer (1882-

1966), hinges on the conviction that a mathematical concept is admissible only if it is

adequately grounded in intuition. For Brouwer, the source of mathematical ideas is
ultimately to be found, not in the external world, but in our own intuitive awareness.
(In particular, Brouwer shares with Kant the view that the natural numbers arise from
our intuitive grasp of temporal succession.) As a result, for intuitionism a mathematical

119
object may be said to exist only if it can be constructed (in thought) in some definite
way, the construction constituting the proof of existence. Moreover, an infinite totality
must never be treated as if it were a completed whole, but rather as if it were
continually growing in time. These principles, if accepted, have profound implications
for logic and mathematics. For example, the first implies that the classical logical law of
the excluded middle is no longer generally valid, that is, we can no longer assert, of any
given proposition A, that either A or its negation not A holds.

According to the

intuitionist view, in order to be able to correctly assert this disjunction, we must actually
be in possession of a proof of A or a proof of not A: however, there are many
mathematical assertionsfor example the statement that every even number is the sum
of two primesfor which we possess neither.

As for the second principle, it

necessitates the outright rejection of Cantors theory and the reconstruction of

mathematics along entirely different lines.
The radical nature of the intuitionist program and the fact that it apparently
entailed sacrifice of large parts of classical mathematics caused it to appear less than
attractive to the majority of mathematicians.
Formalism, the brainchild of David Hilbert, had as its aim the provision of a new

foundation for mathematics by reducing it, not to logic, but to the manipulation of
formal symbols. Hilbert believed that the only completely reliable parts of mathematics
are those which involve nothing more than mechanical reasoning about surveyable
domains of concrete objects, in particular, mathematical symbols considered as marks
on paper. Propositions referring only to such parts of mathematics he regarded as real
propositions while all other mathematical propositions are to be considered ideal
statements, rather like ideal points or lines at infinity in projective geometry. (For
example, 2+2=4 is a real proposition, while there exists an infinite set is an ideal one.)
Thus Hilberts real propositions correspond to the verifiable statements of the logical
positivists, while the ideal statements are, strictly speaking, meaningless. The central

120
objective of Hilberts formalist program was to show by strictly concrete and
unimpeachable means that the classical use of ideal propositionsin particular, those of
Cantors set theory (suitably adjusted to avoid the contradictions already mentioned)
would never lead to falsehoods among the real propositions. In short, the aim was to
prove classical mathematics consistent. Once this was achieved, mathematicians would
be able to roam freely within Cantors paradise (as Hilbert called it) without fear of a
sudden descent into the inferno of contradiction.
Hilbert intended to establish the consistency of classical mathematics by setting it
out as a purely formal system of symbols, devoid of meaning, and then showing that no
proof in the system leads to a false assertion, e.g., 0=1. This, in turn, was to be done by
replacing each (ideal) proof of a real proposition by a real concrete proof. Since, clearly,
there is no concrete proof of the real (false) proposition 0=1, this leads to the conclusion
that classical mathematics is consistent.
However, in 1931 the Austrian logician Kurt Gdel (1906-1976) shattered
Hilberts program by demonstrating, through his famous Incompleteness Theorem, that
there would always be real propositions provable by ideal means which are not
provable by concrete means. He achieved this by means of an ingenious modification
of the ancient Liar paradox (attributed to the Greek philosophers Epimenides and
Eubulides). To obtain the Liar paradox in its most transparent form, one considers the
sentence This sentence is false. Calling this sentence A, an instants thought reveals
that A is true if and only if A is false; in short, A asserts its own falsehood. Now Gdel
showed that, if in A one replaces the term false by the phrase not concretely
provable, then the resulting statement B is true (i.e. provable by ideal means) but not
concretely provable. This is so because B actually asserts its own concrete unprovability.
Gdel showed, moreover, that the consistency of arithmetic cannot be proved by concrete
means. Thus, the soundness of even such an apparently concrete and perspicuous part

121
of mathematics as arithmetic must in some sense remain an article of faith. (See the next
appendix for further discussion of Gdels theorems.)
Although Logicism, Intuitionism and Formalism are unacceptable as complete
accounts of the foundations of mathematics each one embodies an important partial
truth concerning the nature of mathematics: Logicism, that mathematical truth is
intimately connected with logical truth; Intuitionism, that mathematical activity
proceeds by the performance of mental constructions; and Formalism, that the results of
these constructions are presented symbolically.
In practice, most mathematicians regard set theory as constituting an adequate
foundation for their work. This became possible when, in the first few decades of this
century, set theory was axiomatized in such a way as to avoid the evident
contradictions by suitably restricting the formation rules for sets. (Any residual doubts
concerning the acceptability of the axiom of choice were dispelled in 1938 when Gdel
established its consistency with respect to the remaining axioms of set theory.)
Mathematicians find set theory acceptable not solely for the pragmatic reason that it
enables mathematics to be done but also because it accords with the unspoken belief of
many of them that mathematical objects actually exist in some sense and mathematical
theorems express truths about these objects. This is a version of Realism, also termed,
somewhat inaccurately, Platonism.

122
APPENDIX B
HILBERTS PROGRAM AND GODELS THEOREMS

In response to the inconsistencies which had made their appearance in set theory
(see Appendix A), Hilbert proposed to set up a precise formal language for
mathematics and then prove that the resulting formalization is consistent, i.e. does not
lead to contradiction. To ensure that the consistency proof convinced the maximum
possible number of mathematicians, it was supposed to involve only elementary
mathematical notions, and ideally should not transcend school arithmetic.
In 1931 Gdel showed that Hilberts program could not be carried out even for
arithmetic! He showed, in fact, that if arithmetic is consistent, then any consistency

proof must transcend arithmetic itself. And, a fortiori, this applies also to any branch of
mathematics (such as set theory) which is stronger than arithmetic.
We shall sketch proofs of Gdels result, obtaining along the way an important
theorem of Tarski on the undefinability of mathematical truth.
First we set up a precise formal language L for arithmetic as follows.
following symbols:
Numerical variables x1 , x2 , x3 , ...
Numerical constants 0 , 1 , 2 , ...
Numerical function symbols +,
Equality symbol =
Logical operators , , , , , ,
Punctuation symbols.
An expression of L is a finite sequence (string) of symbols of L.

L has the

123
Terms or noun-like expressions of L are obtained as follows.

(i)

Any variable or constant symbol is a term.

(ii)

If t,u are terms, so are t u, t + u.

(iii) An expression is a term if and only if it follows that it is one from

finitely many applications of (i) and (ii).
Examples of terms: x1 + 2 , 6, x1 x5 + 27.
Formulas (or grammatical assertions) of L are obtained as follows.

(a) If t,u are terms, then t = u is a formula.

(b) If A and B are formulas, so are A, A B, A B, A B, A B,
xA, xA for any variable x.
(c) An expression is a formula if and only if it follows from finitely
many applications of (a) and (b) that it is one.
Examples of formulas: x1 + 2 = 3, x1[x1 + 2 = 3], x1[x1 + 2 = 3].

Formulas assume truth values (truth or falsehood) when their constituents are
interpreted arithmetically in the obvious way. We observe that, e.g. the first formula in
the examples above is true or false depending on the value assigned to the variable x1,
while the second or third is simply true or false independently of the value assigned to
the variables. A formula of this latter kind in which each variable x occurring in it is
governed by a quantifier x or x is called a sentence.
We next assign code numbers to the expressions of L as follows. Suppose that the
symbols of L, excluding variables and constants, are k in number. To these symbols we
assign, in some arbitrary but fixed manner, the label 0, ..., k-1. Then to each variable xn
we assign the label k + 2n and to each numeral (i.e. constant) n the label k + 2n + 1. Thus

124
each symbol s has been assigned a label which we shall denote by s*. Now each
expression s1...sn is assigned the code number
s*
1

. 3s2 ... pnsn

where pn is the nth prime number. It follows, by the fact that every natural number has
a unique prime factorization, that distinct expressions are assigned distinct code
numbers by this procedure. The expression with code number n will be denoted by An.
A property P(a1,...,an) of natural numbers is called expressible in

L if there is a

formula A(x1,...,xn) of L such that, for all numbers m1,...,mn,

P(m1,...,mn) holds A(m1,...,mn) is true.
A property of expressions of

is called expressible in

if the corresponding

property of their code numbers is expressible. It can be shown without much difficulty
that the property of being a (code number of a) formula of L is expressible in L.
But what about the property of being a true sentence of L? We shall see that this
property is not expressible!
Since the assignment of code numbers to expressions of

L is evidently effective,

we should be able to compute, for a given expression Am(x1) with code number m, and
any number n, the code number of the expression Am(n) obtained by substituting n for
x1 in Am(x1).

Thus it is plausible to assume (and, indeed, provable) that this is

expressible in L in the sense that there is a formula

s(x1, x2) = x3

125
for some term s of L such that, for any numbers m, n, and p,
(s (m,n) = p ) is true p is the code number of the expression Am(n).
Now let S be any collection of sentences of L: S that may be thought of as a set of
true arithmetical statements in L.

We can now prove:

Theorem 1. Suppose that S satisfies the following conditions:

(1) Each member of S is true.

(2) The property n is the code number of a member of S is expressible in L.
Then there is a true sentence G of L such that G S (and G S).

Proof. By (2) there is a formula T(x1) of L such that, for all n,

T(n) is true n is the code number of a sentence of S

An S.
Write B(x1) for T (s(x1, x1)) and let m be the code number of B. Then
B is Am .
Next, let p be the natural number such that
p = s(m, m)
is a true sentence. Then, by definition, p is the code number of the sentence Am(m).
Now write G for Am(m). Then p is the code number of G, so that

126
G is Ap.
So we have
G is true Am (m) is true
B(m) is true
T(s(m, m)) is true
T(p) is false
Ap S
G S.
(Thus G asserts I am not in S.) It follows that G is true, for
G is false G S G is true.
So G S, and since G is true, G S also. This completes the proof of the theorem.
By taking S to be the collection of all true sentences in this theorem, we
immediately obtain:
Tarskis Theorem in the Undefinability of Truth. The property of being a true sentence of

L is not expressible in L.
Notice that the appropriate sentence G in Tarskis theorem asserts I am not in
the set of true sentences, i.e., I am false. So Tarskis theorem is closely connected
with the well-known Liar Paradox.
Now one can introduce the notion of a proof from S and that of a formula provable
from S in such a way that:

if each member of S is true, so is each sentence provable from S,

if the property of being a member of S is expressible in L, so is the property of being

a formula provable from S.

127
One now obtains immediately from Theorem 1
Godels 1st Incompleteness Theorem (weak form). Let S be a set of true sentences of

such that the property of being a member of S is expressible in L. Then S is incomplete, i.e. there
is a (true) sentence G of S such that neither G nor G is provable from S.

Here the sentence G asserts I am unprovable.

By refining the argument it is possible to strengthen this result as follows. Let us
call S consistent if no formula of the form A A is provable from S. Let P(x1,...,xn) be a
property expressible by a formula A(x1,...,xn). P is said to be S-definite if, for any natural
numbers m1, . . . , mn
P(m1,...,mn) holds A(m1,...,mn) is provable from S
P(m1,...,mn) fails A(m1,...,mn) is provable from S.
Let Q(m1,m2,m3) be the substitution property of natural numbers, i.e.
Q(m1,m2,m3) holds s(m1, m2) = m3 is true.
Then one can prove
Godels 1st Incompleteness Theorem (strong form). Suppose that S is consistent, the
property of being a member of S is expressible and the property Q and that of being an Sprovable formula are both S-definite. Then S is incomplete.

Sketch of proof. As before, we take G to be a sentence of

L which asserts its

own unprovability from S. For any A, let us write S | A for A is provable from S.

128
Suppose that S | G. Then because of the assumptions on S,
S | G is provable from S.
But the assertion G is provable from S is just G, so we get
S | G,
contradicting the supposed consistency of S. Therefore, not (S | G).
Now suppose that S |

G. Then since S is consistent, we get not (S |

G) and

by the assumptions on S it follows that

S | G is unprovable from S.
But G is the assertion G is unprovable from S, so we get
S |

G,

again contradicting the consistency of S. Hence not (S |

G). This completes the

proof.
In the proof of this last theorem we showed that
S is consistent not (S | G).

(*)

Now the assertion on the left-hand side of (*) can be expressed as a sentence ConS of L
as follows. Let no be the code number of the sentence 0 = 1 and let P(x1) be the formula
of L expressing x1 is the code number of a sentence provable from S. Then ConS may
be taken to be the sentence P(no).
The implication (*) can now be formalized in S yielding a proof in S of the
implication
ConSG

129
(observing that G is essentially the formalization in S of the assertion not (S | G)).
Now suppose that
S | ConS .
Then since, as we have seen,
S | ConS G,
it follows that S | G. But, by the 1st incompleteness theorem, if S is consistent, then not
(S | G). Accordingly, we have
Godels Second Incompleteness Theorem. Under the same conditions as the strong form of
the first incompleteness theorem, the sentence ConS formalizing S is consistent is not
provable from S.

Conclusion: if we classify as theological any domain of thought which rests on

faith, i.e. on unproved assumptions, then mathematics is a branch of theology which

contains a rigorous demonstration that it must be so classified!

130
APPENDIX C
SOLUTIONS TO EXERCISES

CHAPTER I

Truth Tables and Testing Validity

A1. (a) invalid; (c), (e), (g), (i), (k), (m), (o) all valid
A2. (a) K F

(c) F O

invalid

invalid

(e) R W
W R

(g) (W Q) D
(W D) (Q D)

(i) F T
F T
L

valid

valid

valid!

A3. Both knights

A4. With 'I'm right' = IR, 'You're a fool' = YF, and 'I'm a fool' = IF, the argument

symbolized is:
IRYF
IFIR
YFIR
IF YF
To check for counterexamples, and hence validity, we only need look at the cases in the
truth table where the conclusion comes out false. That means we only look at cases
where both IF and YF are false (otherwise IF YF comes out true!), and that can happen
in two ways according to what the truth value of IR is. So the truncated truth table for
this argument looks like:

131

IF

YF

IR

IF YF

IRYF

IFIR

YFIR

The second row exhibits a case where the premises are all true but conclusion false, so
the argument is invalid.
A5. With 'The witness was not intimidated' = W, 'Flaherty committed suicide' = F, and

'A note was found' = N, the set of sentences is {W (FN), WF, NF}. Fairly
quickly one can see that W = true, F = false, and N = false is a satisfying truth valuation!
A6. (a) Y [I (Y I)], therefore (Y I) (Y I). valid
A7. Both knaves.
A8. I = cant tell! He = taxpayer for sure.
A9. (a) valid; (c) valid
A10. NO.

The arguments certainly valid, and its second premise is true.

But

supposing premise one is true (i.e. for its conclusion is false) leads to a contradiction:
for since we would then have a sound argument, the conclusion would have to be true,
contradicting premise ones truth (and also the arguments soundness!). So premise one
cant be true! So in fact the conclusion is true, and therefore the argument is unsound (as
its conclusion claims!).
Tautologies, Contradictions and Satisfiability
B1. (a), (c) tautologous
B2. (a) correct; (c) incorrect; (e) correct; (g) incorrect;

(i), (k) correct; (m) incorrect

132
B3. (a) Correct. If a statement is not contingent it is either a tautology or contradiction,

therefore its negation is either a tautology or contradiction, hence its negation can't be
contingent.
(c) Correct. For a conjunction to be a tautology, it must come out true under all possible
valuations; and since it's a conjunction, that means each of its conjuncts has to come out
true under all possible valuations, otherwise the entire conjunction will come out false
under some valuation. Hence all the conjuncts must also be tautologies.
(e) Correct. If pq is valid that means it's a tautology, which means there is never any
case where it comes out true. By the truth-conditions for '', that means there can
never be any case where p comes out true and q false (otherwise pq would be false).
But if there is never any case where p is true and q false, the argument from p as
premise to q as conclusion faces no counterexamples, and so must be valid. Conversely,
suppose the argument from p to q is a valid one. Then there is never any case where p
is true and q is false (otherwise we'd have a counterexample!). But since pq is false
only in such a case, there is never any case where pq comes out false, which means
that it is a tautology, i.e. valid.
B4. (a) (ii); (c) (iii)
B5. (a) If there were a case where a conjunct comes out false, the conjunction would

have to come out false in that case too, and so couldnt be valid!
B6. (a) (iv); (c) (iv); (e) (iii)
B7. (a) false; (c) false; (e) false; (g) true
B8. (a) true; (c) true; (e) false;

(g) false - tautologies only imply tautologies!; (i) false - just add a contradiction!
B9. (a) satisfiable; (c) unsatisfiable; (e) satisfiable

133
B10. (a) false; (c) false; (e) false.
B11. (a) Not valid

CHAPTER II

Equivalence
A1. (a), (c), (e) equivalent; (g) inequivalent; (i) equivalent
A2. (a) (ii), (iii), (v), (vii)-(ix) are all valid, the rest invalid.

(b) Just the tautologies, which are of course all equivalent to each other.
Expressive Completeness
B1. (1) ABCABCABCABC

(3) ABCABCABCABC

B2. (a) p q (p q)

(c) Use the facts: {, } is expressively complete and A A t

B3. (a) Just use the fact that p q p q and that {, } are expressively complete.
B4. (a) To prove the hint: show that if p and q are statements in the letters A and B

taking value t in at least one case where A and B have opposite truth values, then p
q has exactly the same property (i.e. it too takes value t in a case where A and B have
opposite truth values). Using the hint, any statement using just A, B and takes
value t in a case where A and B have opposite values. But in such a case, AB is
false! So you could never express it using just A, B and .
(c) Use: if p(A, B), q(A, B) both are true in at least two cases, p(A, B) q(A, B) has the
same property. Then, since A B only takes value t in one case, youre done!

134
B5. f: ABC ABC ABC
B6.

h(A, A, A) A,

h: ABC ABC ABC ABC

h(A, h(A, A, A), h(C, C, C)) A C

B7. (a) ABC ABC ABC ABC

B8. (a) Since the set {,} is expressively complete and those logical operators can be

expressed in terms of {f,} as: A Af, A B (AB) B

Binary Representations and Venn Diagrams

C1.

(i)
(a)

1+A+B+A.B

(ii)
contingent

(iii)

(c)

1+A+B

contingent

(e)

tautology

C2.

(a)

Binary
1

(c)

(e)

1 + pq + pqr

Venn
A

(only the
indicated space is empty)
C3.

(a)

Binary
1

(c)

(e)

r+p+pr+pq+pqr

Venn
A

....shade everything except the

region inside of q and outside

135
note: this = 0 exactly
when q=1 and r=0 so....

of r!

C4. (a) AB; A+AB; shade only whats inside circle A and outside circle B.

CHAPTER III
Tree Test for Validity
A1. (a) valid (c) invalid; counterexample: A=t, B=C=D=E=f

(e) E A, R A A (E R) invalid, counterexample: A=E=R=f

A2. (a) valid; (c) valid
A3. (i) valid (iii) invalid c.examples: A true, all other statement letters false

[A, B, C]

A4. [A, B, C]

A*B

(A * B)

A
B

A5. (i) valid; (iii) invalid, c. examples: A=t, B=C=f

A6. Rules:

{A, B, C}

A*B

A B

AB

A
B

(i) valid; (iii) invalid

A7. (a) valid; (c) valid
A8. (a) valid; (c) valid
A9. (a) K D

Invalid, one counterexample

(i) valid

136
KS

[Note that the argument fails simply because Dumb could

S D

be so dumb that he doesnt even know it!]

A10.

S = scientists dont know what they are talking about

B = the sun will eventually burn out
E = Earth will become dark and cold
M = Mars is teeming with life
H = the human race will migrate to other planets
D = the human race will die out
Argument has H=E=B=true, S=M=D=false for a counterexample
A11. Invalid, exactly 4 counterexamples

Further Applications of the Tree Method

B1. (a) unsatisfiable; (c) satisfiable, A=t, B=f
B2. (a) tautology; (c) tautology
B3. (i) contradiction
B4. (i) P & J are both Knights or both Knaves.
B5. (i), (iii) equivalent; (v) inequivalent
B6. (a) contingent; (c) contingent
B7. (i) Both knaves; (iii) I=knave, He=cant tell! (this ones tricky)
B8. (a) contingent; (c) contradiction

137
B9. Lancelot is a knave, Arthur a knight and Merlin a knight.
B10. (a), (b) and (e) are the tautologies, the others arent.
B11. Theyre all inconsistent!
B12. (a) Both knights.
B13.
1. No information derivable.
3. M is a saint if and only if J is a sinner.
5. M is a saint.
7. M is a saint and J a sinner.
9. J is a sinner.
11. M and J are both saints.
13. M is a sinner and J a saint.
15. Inconsistent.
17. M and J are both saints.
19. M is a sinner and J and his brother are not both saints.
21. C is a sinner and the others arent both saints.
23. C and L are saints and M a sinner.
25. M is a saint and the others are sinners.
B14. (a) Satisfiable in 4 ways:

A
f
f

B
t
f

C D
f f
f f

138
f
t

f
t

f
f

t
f

(c) Unsatisfiable
B15. (i) Non-tautologous; A=f, B=t only way to make false.
B16. (i) Satisfiable in one way:
A B
f f

C D
f t

B17. L=Lancelot a knight

A=Arthur a knight
M= Merlin a knight

LM
A (L M)
M [M (L A)]
L
M

L
M
A
L M

A
LM
L

M
M

M (L A)

[M(LA)]
M
L A
L

A
LM

L
L
M
M

A
L M

L
M

M(LA) [M(LA)]
M
M
LA
LA
L

From 3 open branches we see: A is true and M and L have opposite truth values.
B18. Guinevere is in Camelot today - Lancelot is not deceived!

139
B19. (a) Dean, Jerry, Stan are heroes, and Ollie is a scoundrel.

(c) One question you could ask is Is it the case that youre a knight if and only if there
is no buried treasure on the island? It is easy to verify (with the tree method) that if
Dean answers Yes, theres no gold on the island; and if he answers No, there is!

CHAPTER IV
Trees and Translations Involving Quantifiers
A1. (i) valid; (iii) valid; (v) invalid
A2. (a) x(Lx Nx), x(Vx Nx) x(Vx Lx), valid.

(c) x(Bxa yByx) Bax, invalid.

A3. valid
A4. (i) Ec; (iii) Sbc Scb; (v) x(Ex Wdx y(EySyx))
A5. (a) xPxPu, x(LxPx), Ld Pu; valid.

(c)x(DxCx) x(y(DyLxy)z(Cz Lxz)); valid.

A6.

(i) xCx

(iii) w[xLwx zLzw]

CwG valid
G

Lrj
Liy

A7. (i) {x[Rx(Lx Vx)], Lj Vj} satisfiable

(iii) {w[xLwxzLzw], Lyy, Lym} unsatisfiable

A8. (a) xy Kxy

(c) xy[KyyKxy]

valid

140
(e) x[y(Kxy Kay) y(KyaKxy)]
A9. (a) x[Hx y[ByPxy]]

y[Byx(Hx Pxy)]

(c) x(Sx Ux)

x(Wx Ux)
x(Sx Vx)

valid

x(Wx Vx)
valid
A10. (a) Lmm; (c) xLxm; (e) Lmm x(Lmx x=m); (g) x(Lxm Lmx);

(i) x(Lxm yLxy)

A11. (a)xTrx; (c) xyTxy; (e) xy((Trx Try) x=y); (g) xTxrTrr;

(i) x(Txr y(Tyr Txy))

A12. (i) xy (( Cx Cy x y) (Txy Tyx))

x(Cx Txx)
xy (Cx Cy x y)

Valid.

A13. (i) equivalent

A14. !xPx means theres exactly one thing with property P. (i) valid; (iii) invalid
A15. (a) x(CxDtx)

x(Cx Dxt)
Cs
x(Dtx Dxt) NOT Valid.

(c)

x(x < f Ex)

xy(x<y z(x <z z <y))

x(x <f y(y < x Ey))
Valid.

A16. (a) xyLxy; (c) xy(Lyy Lxy); (e) x(yLxy z(Lxz wLzw))

141
A17. (a) x(Cx Lbx)

(c) xy (zLyz Lxy)

x(Cx Lxb)

Lrj

Cw

Lfw

x(Lbx Lxb)

Valid.

Valid.

Interpretations and Counterexamples

B1.

(i) x (Cx Tx)

x (Tx Cx)

Not valid. Counterex: Domain = {1}, C= { }, T = { }

(iii) x Txr Tgr

Tgr y Tyr
(x Txr x Txr)

Valid. No counterex.

B2. (a) False, because parallel lines dont intersect.

B3. For the knight case, you can take:

Domain={Arthur, Lancelot}, Knaves={Lancelot}, Knights={Arthur}

For the knave case, you can take:
Domain={Arthur, Lancelot}, Knaves={Lancelot, Arthur}, Knights={ }
B4. (i) false; (iii) true; (v) true
B5. (a) False, because the number 1 is odd but there are no positive even numbers less

than it!
B6. (a) With Domain = {1} and Pxy taken to mean x=y, trivially satisfiable!

(c) unsatisfiable
B7. (a) x(Fx yHyx) xy((Fx Fy) Hxy), valid.

142
(c) xy((Lx By) Txy), xy(xy (Lx Ly)), xy(xy Txy) not valid.
Counterex: Domain = {a}, L={ }, B={ }, T={ }
(always shoot for a simple counterexample by trial and error first!!!)
B8. (a) x(Sx y(Cyy=x)), Cb, Tbx(CxTx), valid.

(c) x(y(OsyOxy)Osx) Oss x(Osx x=s), not valid,

counterex: Domain ={s,a}, O ={(s,s), (a,s), (s,a)}
B9. (i) False, because shape a is not a triangle.

(iii) False, because g is neither a triangle nor square.

(v) False, because g has nothing left of it.
(vii) True ; (ix) True
B10. (1) xyIxy, which is true because every statement implies itself!

(3) xy(z(Ixz Iyz) Exy), which is true because if x and y have the same
implications, then since x implies itself, y must imply x too, and also (by the same
argument) x must imply y, which means x and y must be equivalent.
(5) This says: Every statement fails to imply some statement, which is equivalent
to saying: No statement implies all statements. But contradictions do! So the
stated claim is false.
(7) This says: Some statements are equivalent to anything they imply Yes, thats
right: just consider any tautology!
B11. 1. False. To make it True, remove as feather.

3. False. To make it True, increase as and cs heights to the second line.

5. False. To make it True, give ds hat to b and remove ds feather.

143
B12. (a) false; (c) true; (e) true; (g) true; (i) true; (k) false
B13. (a) Case where true: Domain = {1}, R = { };

Case where false: Domain = {1}, R= {(1,1,1)}

*Note: These are obviously not the only possible answers!
(c) Case where true: Domain = {1,2}, P = {1,2}, Q = {1,2};
Case where false: Domain = {1}, P= { }, Q = { }
B14. (i) False; (iii) True; (v) True; (vii) True
B15. (i) false (iii) false (v) false (vii) false (ix) true

Functions
C1. (i) a=mfb; (iii) fa=fb ma=mb ab; (v) x (a=fmx a=ffx)
C2. (a) a and b are siblings

(c) a is bs brother; (e) b is as father

CHAPTER V

The Propositional Calculus

A1. Justification for the sequence: in initial set, Axiom 1, Axiom 1, Axiom 2, Modus

Ponens, Modus Ponens, Axiom 9, Axiom 10, Modus Ponens, Axiom 1, Modus Ponens,
in initial set, Modus Ponens.
So, Yes a legit deduction of p | p. Simple deduction is: p, p p, p !
In initial set Ax.10

MP

144
A2. 1. axiom 2; 2. axiom 1; 3. in initial set; 4. modus ponens; 5. axiom 1; 6. axiom 5; 7.

modus ponens; 8. in initial set; 9. axiom 9; 10. modus ponens; 11. modus ponens; 12.
axiom 10; 13 modus ponens; 14. You aint foolin me!
A3. For first sequence, justification is: in S, in S, Axiom 5, MP, MP; so the sequence is

indeed a legitimate deduction from S.

A4. tautology goes with theorem

unsatisfiable goes with formally inconsistent

valid argument goes with deduction
valid argument, no premises goes with proof
A5. (i) S | p S |= p. Every deduction in the propositional calculus generates a valid

argument!
(iii) To show that logic (at least propositional logic without quantifiers) can be
completely captured through formal rules of symbol manipulation without any
reference to the external (and potentially problematic) notion of truth or meaning.
A6. (a) false; (c) true; (e) false; (g) true; (i) false
A7. Because if p | r, then clearly p,q | r for any q, from which it follows by the

deduction theorem that p | (qr).

A8. (a) The Deduction theorem says: S, p | q if and only if S | p q.

Assuming completeness and soundness, all we have to show is:

S, p |= q iff S |= p q (*)
because we could then argue as follows:
S, p | q S, p |= q S |= p q S | p q.

145
Argument for (*). Assuming S, p |= q, it follows that {S, p, q} is unsatisfiable. So if all
statements in S are true, p q can't be false; because if it were, both p and q would
have to be true as well, contradicting the unsatisfiability of {S, p, q}.

So if all

statements in S are true, so is p q, which is just to say that S |= p q. Now we need

to argue the other way around. If S |= p q, then theres never a case where all the
statements in S are true and p q is false, i.e. p true, q true. So {S, p, q} is
unsatisfiable, which implies S, p |= q.
A9. (i) If S | pq then there is a legitimate deduction sequence of form S, , p.

Similarly, if p | q then a legitimate sequence p, , q exists. So now just concatenate

these 2 sequences to yield the following legitimate sequence: S, , p, p, , q which
justifies S | q !
(iii) First note that p,q | p q which is justified by the sequence: p, q, p (q (p
q)), q (p q), p q. Now suppose S | p, i.e. S, , p is a legitimate deduction
sequence, and S | q, i.e., S, ,q is legitimate. Since p,q, , p q is legitimate (as we've
just shown), concatenating these three sequences gives S, , p, S, , q, p, q,, p q,
which also is a legitimate deduction sequence, thus establishing S | (pq).
A10. (a) By completeness and (strengthened) soundness, all you have to argue is that:

q: S |= q and S |= q if and only if

p: S |= p

But thats clear: if the left-hand side is true, then S cant be satisfiable, which means that
the right-hand side is true. (The argument from right to left is trivial.)