Lesson 2

Inference

Now that we have a collection of propositional forms and a means by which to

interpret them as either true or false, we want to define a system that expands these ideas
to include methods by which we can prove certain propositional forms from given
propositional forms. What we will define is familiar because it is similar to what Euclid
did with his geometry. Take, for example, the familiar result,

opposite angles in a parallelogram are congruent.

In other words,

if ABCD is a parallelogram, then B  D,

which translates to:

Given: ABCD is a parallelogram,

Prove: B  D.

To demonstrate this, we draw a diagram,

and then write a proof:

Logic and Set Theory

 Euclid’s geometry consists of geometric propositions that are established by
proofs like the above. These proofs rely on rules of logic, previously proved propositions
(lemmas, theorems, and corollaries), and propositions that are assumed to be true (the
postulates). Using this system of thought, we can show which geometric propositions
follow from the postulates and conclude which propositions are true, whatever it means
for a geometric proposition to be true. Euclidean geometry serves as a model for the
following modern definition.

A logical system consists of the following:

 An alphabet
 A grammar
 Propositional forms that require no proof
 Rules that determine truth
 Rules that are used to write proofs.

Although Euclid did not provide an alphabet or a grammar specifically for his
geometry, his system did include the last three aspects of a logical system. In this lesson
we develop the logical system known as propositional logic. Its alphabet, grammar, and
rules that determine truth were defined.
Consider the following collection of propositions:

If squares are rectangles, then squares are quadrilaterals.

Squares are rectangles.
Therefore, squares are quadrilaterals.

This is an example of a deduction, a collection of propositions of which one is supposed

to follow necessarily from the others. In this particular case,

if squares are rectangles, then squares are quadrilaterals

and

squares are rectangles

are the premises, and

squares are quadrilaterals

is the conclusion. We recognize that in this case, the conclusion does follow from the
premises because whenever the premises are true, the conclusion must also be true.
When this is the case, the deduction is semantically valid, else it is semantically invalid.
Notice that not only do we see that the deduction works because of the meaning
of the propositions, but we also see that it is valid based on the forms of the sentences. In
other words, we also recognize this deduction as valid:

Logic and Set Theory

 If Hausdorff spaces are preregular, their points can be separated.
Hausdorff spaces are preregular.
Therefore, their points can be separated.

Although we might not know the terms Hausdorff space, preregular, and separated, we
recognize the deduction as valid because it is of the same pattern as the first deduction:

pq
p (a )
q

When the deduction is found to work based on its form, the deduction is syntactically
valid, else it is syntactically invalid.
We study both types of validity by examining general patterns of deductions and
choosing rules that determine which forms correspond to deductions that are valid
semantically and which forms correspond to deductions that are valid syntactically.

Semantics

The study of meaning is called semantics. We began this study when we wrote
truth tables. These are characterized as semantic because the truth value of a proposition
is based on its meaning. Our goal is to use truth tables to determine when an argument
form, an example being (a), corresponds to a deduction that is semantically valid.
We begin with a definition.

Let p0 , p1 ,..., pn1 and q be propositional forms.

 If q is a tautology, we write ⊨ q.
 Define p0 , p1 ,..., pn1 to logically imply q if
⊨ p0  p1  ...  pn1  q.

When p0 , p1 ,..., pn1 logically imply q, write

p0 , p1 ,..., pn1 ⊨ q.

and say that q is a consequence of p0 , p1 ,..., pn1. Call the propositional forms
p0 , p1 ,..., pn1 the premises of the implication and q the conclusion.

Notice that if p0 , p1 ,..., pn1 ⊨ q, then for any valuation  , whenever   pi   T for all
i  0, 1,..., n  1, it must be the case that  q   T . Moreover, any deduction with
premises presented by p0 , p1 ,..., pn1 and conclusion by q is semantically valid if
p0 , p1 ,..., pn1 ⊨ q.

Logic and Set Theory

Logic and Set Theory
Syntactics

Although we will return to semantics, it is important to note that using truth

tables to check for logical implication has its limitations. If the argument form involves
many propositional forms or if the propositional forms are complicated, the truth table
used to show or disprove the logical implication can become unwieldy. Another issue is
that in practice, truth tables are not the method of choice when determining whether a
conclusion follows from the premises. What is typically done is to follow Euclid’s
example, basing the conclusions on the syntax of the argument form, namely, based only
on its pattern and structure.
Start with the two propositions,

if squares are rectangles, then squares are quadrilaterals

and

squares are rectangles.

Because of the combined structure of the two sentences, we know that we can write

squares are quadrilaterals.

Logic and Set Theory

This act of writing (on paper or a blackboard or in the mind) means that we have the
proposition and that it follows from the first two. Similarly, if we start with

squares are triangles, or squares are rectangles

and

squares are not triangles,

we can write

squares are rectangles.

Determining a method that will model this reasoning requires us to find rules by
which propositional forms can be written from other propositional forms. Since every
logical system requires a starting point, the first step in this process is to choose which
propositional forms can be written without any prior justification. Each such
propositional form is called an axiom. Playing the same role as that of a postulate in
Euclidean geometry, an axiom can be considered as a rule of the game. Certain
propositional forms lend themselves as good candidates for axioms because they are
regarded as obvious. That is, they are self-evident. Other propositional forms are good
candidates to be axioms, not because they are necessarily self-evident, but because they
are helpful. In either case, the number of axioms should be as few as possible so as to
minimize the number of assumptions. For propositional logic, we choose only three.
They were first found in work of Gottlob Frege (1879) and later in that of Jan
Łukasiewicz (1930).

■ AXIOMS 1.2.8 [Frege–Łukasiewicz]

Let p, q, and r be propositional forms.

 p  q  p  [FL 1]
 p  q  p    p  q   p  r   [FL 2]
 p  q  q  p  [FL 3]

The next step in defining propositional logic is to state when it is legal to write a
propositional form from given propositional forms.

The propositional forms p0 , p1 ,..., pn1 infer q if q can be written whenever p0 , p1 ,..., pn1
are written. Denote this by

p0 , p1 ,..., pn1  q.

This is known as inference.

Logic and Set Theory

 To make rigorous which propositional forms can be inferred from given forms,
we establish some rules. These are chosen because they model basic reasoning. They are
also not proved, so they serve as postulates for our logic.
■ INFERENCE RULES

Let p, q, and r be propositional forms.

To use the Inference Rules, match the form exactly. For example, even though
P  R appears to follow from P  Q  R as an application of simplification, it does
not. The problem is that simplification can only be applied to propositional forms with
the p  q pattern, but P  Q  R is of the form p  q. With this detail in mind, we
make some inferences.

Each inference is justified by the indicated rule.

Logic and Set Theory

 Since it is possible that some propositional forms are not needed for the inference,
we also have the following:

Inference is a powerful tool, but it can only be used to check simple deductions.
Sometimes multiple inferences are needed to move from a collection of premises to a
conclusion. For example, if we write

p  q, p, q  r ,

based on the first two propositional forms, we can write

by DS, and then based on this propositional form and the third of the given propositional
forms, we can write
r
by MP. This is a simple example of the next definition.

 A formal proof of the propositional form q (the conclusion) from the

propositional forms p0 , p1 ,..., pn1 (the premises) is a sequence of propositional
forms,

p0 , p1 ,..., pn1 , q0 , q1 ,..., qm1 ,

Logic and Set Theory

 such that qm1  q, and for all i = 0, 1,…, m-1, either qi is an axiom,

if i = 0, then p0 , p1 ,..., pn1  qi or

if i > 0, then p0 , p1 ,..., pn1 , q0 , q1 ,..., qi1  qi .

If there exists a formal proof of q from p0 , p1 ,..., pn1 , then q is proved or deduced
from p0 , p1 ,..., pn1 and we write
p0 , p1 ,..., pn1 ⱶ q.

If there are no premises, a formal proof of q is a sequence,

q0 , q1 ,..., qm1 ,

Such that q 0 is an axiom, qm1  q, and for all i > 0, either qi is an axiom or

q0 , q1 ,..., qm1  qi

In this case, write ⱶ q and call q a theorem.

Observe that any deduction with premises represented by p0 , p1 ,..., pn1 and conclusion
by q is syntactically valid if p0 , p1 ,..., pn1 ⱶ q.
We should note that although  and ⱶ have different meanings as syntactic
symbols, they are equivalent. If p  q, then p ⱶ q using the proof p , q. Conversely,
suppose p ⱶ q. This means that there exists a proof

p, q0 , q1 ,..., qn1 , q,

so every time we write down p , we can also write down q. That is, p  q. We
summarize this as follows.

■ THEOREM 1.2.14

For all propositional forms p and q, p  q if and only if p ⱶ q.

We use a particular style to write formal proofs. They will be in two-column

format with each line being numbered. In the first column will be the sequence of
propositional forms that make up the proof. In the second column will be the reasons
that allowed us to include each form. The only reasons that we will use are:

 Given (for premises),

 FL1, FL2, or FL3 (for an axiom),
 An inference rule.

An inference rule is cited by giving the line numbers used as the premises followed by the
abbreviation for the rule. Thus, the following proves P  Q  Q  R, P ⱶ Q :

Logic and Set Theory

Despite the style, we should remember that a proof is a sequence of propositional forms
that satisfy the aforementioned Definition. In this case, the sequence is:

P  Q  Q  R, P, P  Q, Q  R, Q.

The first two examples involve proofs that use the axioms.

The next three examples do not use an axiom in their proofs.

Logic and Set Theory

Logic and Set Theory
Logic and Set Theory