LOGIC

Lecture 9: Proofs for Predicate Logic I

PREDICATE LOGIC
 Quick review of QL grammar:
 Every atomic formula (predicate followed by
names/variables) is a WFF of QL.
 If ϕ is a WFF, then ~ ϕ is a WFF.
 If ϕ is a WFF and if ψ is a WFF, then (ϕ& ψ) is a
WFF.
 If ϕ is a WFF and if ψ is a WFF, then (ϕ v ψ) is a
WFF.
 If ϕ is a WFF and if ψ is a WFF, then (ϕ → ψ) is a

 If ϕ is a WFF and if ψ is a WFF, then (ϕ ↔ ψ) is a

WFF.

WFF.
PREDICATE LOGIC
 Quick review of QL grammar:
 If ϕ is a WFF, χ is a variable, and ϕ contains no χ-
quantifiers, then ∀χϕ is a WFF.
 If ϕ is a WFF, χ is a variable, and ϕ contains no χ-
quantifiers, then ∃χϕ is a WFF.
 Nothing else is a WFF.
PREDICATE LOGIC
 Quick review of scope and bound/free
variables:
 The scope of a quantifier is the string of symbols
it ‘attaches to’. Quantifiers ‘attach’ like negation.
 Thus, the scope of the universal quantifier in
∀xFx&Gx is Fx – the scope of the quantifier in
∀x(Fx&Gx) is (Fx&Gx).
 A variable is bound if it is inside the scope of a
quantifier with a matching variable (e.g. ∀x binds
x’s). Otherwise it is free.
 WFFs containing free variables do not have a
truth value.
PREDICATE LOGIC
 Quick review of truth-conditions for QL:
 QL interpretations consist of a domain, an
assignment of names to entities in the domain,
and an assignment of extensions to predicates.
 An existentially quantified sentence is true if and
only if there is at least one entity in the UD such
that, if its name is substituted for occurrences of
the relevant variable in the quantifier’s scope,
the result is a true sentence.
 A universally quantified sentence is true if and
only if every entity in the UD is such that, if its
name is substituted for occurrences of the
relevant variable in the quantifier’s scope, the
result is a true sentence.
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 None of these are equivalent. Let’s construct
interpretations to show this.
 Our UD for each will be: {Alice, Bill, Carl}.

 We’ll use the names: Alice = a, Bill = b, Carl =

c.
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves
everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 An interpretation where only the first
sentence is true:
 Extension(L)= {<Alice,Bill>, <Bill,Carl>,
<Carl,Carl>}
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves
everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 This shows us that 1 is not equivalent to 2, 3,
or 4 – remember to be equivalent the first
must entail the second and the second the
first. 1 doesn’t entail any of the rest.
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves
everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 An interpretation where only the third
sentence is true:
 Extension(L) = {<Alice,Alice>, <Bill,Bill>,
<Bill,Carl>}
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 Sentence 2 entails sentence 3, but not 1 or 4:
 Extension(L) = {<Alice,Alice>, <Alice,Bill>,
<Alice,Carl>}
 (this wouldn’t have changed if we had made Bill
love everyone, or Carl)
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves
everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 But since 3 does not entail 2, they are not
equivalent.
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 Sentence 4 entails sentence 1 but not 2 or 3:
 Extension(L) = {<Alice,Alice>, <Bill,Alice>,
<Carl,Alice>}
 (this wouldn’t have changed if we had made Bill
loved by everyone, or Carl)
PREDICATE LOGIC
 Let’s get in a little more practice with nested
quantifiers.
 ∀x∃yLxy – For everyone, there is someone that
person loves.
 ∃x∀yLxy – There is someone who loves
everyone.
 ∀x∃yLyx – Everyone is loved by someone.
 ∃x∀yLyx – There is someone who is loved by
everyone.
 But 1 does not entail 4. So no equivalence in
any pair.
PREDICATE LOGIC
 Our main goal for today will be to begin to
extend our natural deduction system to QL.
PREDICATE LOGIC
 The same rules that applied for SL still apply
for QL. So the following is a successful proof:
 1. ∀xFx & Ga
 2. ∀xFx → ∃yGy
 3. ∀xFx &E 1
 4. ∃yGy →E 2,3
 5. ∃yGy v Rb vI 4
PREDICATE LOGIC
 All we need to add are intro/elim rules for our
two quantifiers.
 (Actually, we will also add a further logical
symbol to QL next time, plus intro/elim rules for
it – identity, or =).
 Two of the quantifier rules are easy. The
other two are hard.
PREDICATE LOGIC
 First we need a new bit of terminology:
 SUBSTITUTION INSTANCE: The result of
substituting a constant for each instance of a
variable in a WFF, and removing the variable’s
quantifier.
 So, e.g., Fa is a substitution instance of ∀xFx. So
is Fb.
 Fa v Ga is a substitution instance of ∀x(Fx v Gx).
 Fa v Gb is a substitution instance of ∀x(Fx v Gb).
 ∃yGby is a substitution instance of ∀x∃yGxy.

 Fav Gb is NOT a substitution instance of ∀x(Fx v

Gx). (must substitute same constant)
PREDICATE LOGIC
 We will write this formally as:
Φ [ χ / a]
 This indicates the result of replacing all
occurrences of χ in Φ with a.

 There are several notational variants for

substitution. Ichikawa uses an arrow, but
that’s easy to confuse with →.
 You shouldn’t need to actually write this for
purposes of HW/tests, so don’t worry about
notation.
PREDICATE LOGIC
 Now we can give our first quantifier rule:
UNIVERSAL ELIMINATION.
 m. ∀χΦ
 n. Φ [ χ / c ] ∀E m

 In English: if you have a line whose main

connective is a universal quantifier, you can
write down any substitution instance of that
line (c here stands for any constant - any
name).
PREDICATE LOGIC
 The basic idea: if I know that e.g. everything
is big, then any particular thing I pick is going
to be big.
 Or: if I know that everything in the UD

satisfies a given WFF, then I know that any

member of the UD I pick will satisfy the WFF.
PREDICATE LOGIC
 Here’s how it looks on a few particular cases:
 1. ∀xFx
 2. Fa ∀E 1

 1. ∀x(Fx & Gx)

 2. Fb & Gb ∀E 1

 1. ∀x(Fx & Gb)

 2. Fa & Gb ∀E 1

 1. ∀x∃yLxy
 2. ∃yLay ∀E 1
PREDICATE LOGIC
 You can NOT use the rule like this:
 1. ∃xFx
 2. Fa ∀E 1

 1. ∀x(Fx & Gb)

 2. Fa & Ga ∀E 1

 1. ∀x(Fx & Gx)

 2. Fa & Gx ∀E 1 (free variables = bad!)

 1. ∀x∃yLxy
 2. Laa ∀E 1
PREDICATE LOGIC
 You can use any name to substitute. In some
cases, you’ll use one that’s already in the
proof:
 1. Fa → Ga
 2. ∀xFx
 3. Fa ∀E 2
 4. Ga →E 1,3

 But in other cases you’ll want to introduce a

new name, that doesn’t appear anywhere.
We’ll see why in a little bit.
PREDICATE LOGIC
 Our other easy rule is EXISTENTIAL
INTRODUCTION.
 m. Φ
 n. ∃χΦ [[ c / χ ]] ∃I m

 The notation here is a bit weird. In English: if

you have a WFF, you can replace any or all
instances of a given constant in the WFF with
a variable and add an existential quantifier.
 Sort of a ‘reverse’ substitution instance!
 In simpler English: If I know e.g. Mary is tall, I
can infer something is tall.
PREDICATE LOGIC
 This rule lets me do things like this:
 1. Fa
 2. ∃xFx ∃I 1

 1. Fa & Ga
 2. ∃x(Fx & Gx) ∃I 1

 1. Fa & Ga
 2. ∃x(Fx & Ga) ∃I 1 (I do NOT need to
replace every instance of the name – no free
variable problem)
PREDICATE LOGIC
 I can NOT do this:
 1. Fa
 2. ∀xFx ∃I 1

 1. Fa & Gb
 2. ∃x(Fx & Gx) ∃I 1
PREDICATE LOGIC
 Let’s do a practice proof with these two new
rules.
 From Gb&Fa and ∀x(Fx → Gx), show ∃x(Fx &

Gx).
PREDICATE LOGIC
1. Gb&Fa (∃x(Fx & Gx))
2. ∀x(Fx → Gx)
3. Fa &E 1
4. Fa → Ga ∀E 2
5. Ga →E 3,4
6. Fa & Ga &I 3,5
7. ∃x(Fx & Gx)∃I 6
PREDICATE LOGIC
 Let’s do one more, a bit more complex: From
∀x∀y(Gxy → Fy), show that Gab→∃xFx.

1. ∀x∀y(Gxy → Fy)
2. Gab want ∃xFx
3. ∀y(Gay → Fy) ∀E 1
4. Gab → Fb ∀E 3
5. Fb →E 2,4
6. ∃xFx ∃I 5
7. Gab→∃xFx →I 2-6
PREDICATE LOGIC
 We’ll introduce one more rule today, and
then leave the final quantifier rule for next
time.
 UNIVERSAL INTRODUCTION:
 m. Φ
 n. ∀χΦ [c / χ]* *: where c does not occur in
an undischarged
assumption (including
premises)
 In English: You may form a universally
quantified sentence from any sentence,
PROVIDED the constant you are replacing is
arbitrary.
PREDICATE LOGIC
 A constant is arbitrary if it is not part of the
premises, or part of an assumption of an
open subproof.
 For instance, this is not an acceptable move:

1. Fa
2. ∀xFx ∀I 1

 Obviously, if the premise tells us that a is F,

we can’t directly infer that everything is F!
PREDICATE LOGIC
 But suppose I had this:
1. ∀x(Fx&Gx)
2. Fa&Ga ∀E 1

 Here, my use of ∀E introduced a new name

that was not part of the premises.
 Because it is not in the premises, it is an

arbitrary name. We could have just as easily

used b, or c, or d.
 It’s a bit like saying: “Let’s pick a member of
the UD at random, without looking, and call
that thing a”
PREDICATE LOGIC
 But suppose I had this:
1. ∀x(Fx&Gx)
2. Fa&Ga ∀E 1

 All we know about a is that a is F and a is G –

but we know that about EVERY member of
the UD!
 We don’t have any ‘special’ or ‘extra’

information about a.
 (Compare with a case where the premises told
us something about a)
PREDICATE LOGIC
 But suppose I had this:
1. ∀x(Fx&Gx)
2. Fa&Ga ∀E 1
3. Fa &E 2
4. ∀xFx ∀I 3

 Since a is arbitrary, we can use ∀I. Note that

the reasoning we go through in this proof to
get to line 3 would work for ANY element of
the UD.
PREDICATE LOGIC
 But suppose I had this:
1. ∀x(Fx&Gx)
2. Fa&Ga ∀E 1
3. Fa &E 2
4. ∀xFx ∀I 3

 Since we could go through the same

reasoning for every member of the UD, we
simply allow ourselves the ‘shortcut’ of
moving straight to ∀xFx.
PREDICATE LOGIC
 We have to be very careful with this rule.
Every time you use it, you must first make
sure that the constant you are replacing does
not appear in:
 The premises
 The assumption of an open subproof.
 It IS okay to use the rule on a constant that
appears in the premise of a closed subproof
(a ‘discharged’ assumption).
PREDICATE LOGIC
 Example:
1. Fa&~Fa for reductio
2. Fa &E 1
3. ~Fa &E 1
4. ~(Fa&~Fa) ~I 1-3

5. ∀x~(Fx&~Fx) ∀I 4

 This is okay, because a does not show up in

the premises (there are none!); and the
subproof assumption it appears in is
DISCHARGED (no longer active, since
subproof is closed).
PREDICATE LOGIC
 Example:
1. Fa&~Fa for reductio
2. Fa &E 1
3. ~Fa &E 1
4. ~(Fa&~Fa) ~I 1-3

5. ∀x~(Fx&~Fx) ∀I 4

 It shows up in lines 2, 3, and 4 – but that’s

ok. We just have to make sure it’s not in the
premises/assumptions.
PREDICATE LOGIC
 Example:
1. Fa&~Fa for reductio
2. Fa &E 1
3. ~Fa &E 1
4. ~(Fa&~Fa) ~I 1-3

5. ∀x~(Fx&~Fx) ∀I 4

 Lines 2-3 aren’t ‘active’ anymore, since the

subproof is closed.
 And line 4 doesn’t give us ‘special’ info about

a – the proof on 1-3 could work with any

name.
PREDICATE LOGIC
 Example:
1. Fa&~Fa for reductio
2. Fa &E 1
3. ~Fa &E 1
4. ~(Fa&~Fa) ~I 1-3

5. ∀x~(Fx&~Fx) ∀I 4

 So in this proof, a really is arbitrary.

PREDICATE LOGIC
 We’ll leave existential elimination until next
time. It’s a bit trickier, since it requires a
subproof.
PREDICATE LOGIC
 Practice proofs if time:
 From ∀x(Fx→Gx) and ∀xFx, prove ∀xGx.

1. ∀x(Fx→Gx)
2. ∀xFx

3. Fa→Ga ∀E 1
4. Fa ∀E 2
5. Ga → 3,4
6. ∀xGx ∀I 5
PREDICATE LOGIC
 Practice proofs if time:
 From Fa, prove ∀x(Fa v Fx).

1. Fa
2. Fa v Fb vI 1 (can’t use a!)
3. ∀x(Fa v Fx) ∀I 2
PREDICATE LOGIC
 Practice proofs if time:
 From ~∃xFx, prove ∀x~Fx.

1. ~∃xFx
2. Fa (for reductio)
3. ∃xFx ∃I 2
4. ~∃xFx R1
5. ~Fa ~I 2-4
6. ∀x~Fx ∀I 5