LOGIC

Lecture 10: Natural deduction II, Identity

PREDICATE LOGIC
 Let’s review the rules we have so far for
natural deduction in QL.
 Remember, all the rules we had in SL still

apply as well.
PREDICATE LOGIC
 UNIVERSAL ELIMINATION:
 m. ∀χΦ
 n. Φ [ χ / c ] ∀E m

 This is the rule that allows us to create

SUBSTITUTION INSTANCES of universally
quantified sentences. E.g.:
 1. ∀x(Fx v Gx)
 2. Fa v Ga
PREDICATE LOGIC
 UNIVERSAL ELIMINATION:
 m. ∀χΦ
 n. Φ [ χ / c ] ∀E m

 Any name can be substituted, even ones

already used in the proof.
 You must substitute it for all occurrences of

the bound variable (to avoid free variables).

PREDICATE LOGIC
 EXISTENTIAL INTRODUCTION:
 m. Φ
 n. ∃χΦ [[ c / χ ]] ∃I m

 This rule lets you replace a name with a

variable + existential quantifier. E.g.:
 1. Fa→Ga
 2. ∃x(Fx→Gx)
PREDICATE LOGIC
 EXISTENTIAL INTRODUCTION:
 m. Φ
 n. ∃χΦ [[ c / χ ]] ∃I m

 You do NOT need to replace all instances of

the name you are replacing. E.g., this is fine:
 1. Faa
 2. ∃xFax
PREDICATE LOGIC
 UNIVERSAL INTRODUCTION:
 m. Φ
 n. ∀χΦ [c / χ]* *: where c does not occur in
an undischarged
assumption

 This rule allows you to replace a name with a

variable and a universal quantifier. But it is
slightly more tricky than ∃I, since you must
make sure that the name you are replacing is
ARBITRARY.
PREDICATE LOGIC
 UNIVERSAL INTRODUCTION:
 m. Φ
 n. ∀χΦ [c / χ]* *: where c does not occur in
an undischarged
assumption

 In other words, the name you are replacing

cannot appear in the premises, or in an
assumption of an open subproof.
PREDICATE LOGIC
 For this rule we can not choose which
instances of the constant to replace. We
must replace all of them. Otherwise we could
argue this way:
 1. ∀x(Fx & Lxx)
 2. Fa & Laa ∀E 1
 3. Laa &E 2
 4. ∀xLax ∀I 3
 5. ∃y∀xLyx ∃I 4

 E.g., from ‘everyone is fat and loves

themselves’ to ‘there is someone who loves
everyone’.
PREDICATE LOGIC
 Let’s now introduce the last quantifier rule –
EXISTENTIAL ELIMINATION.
 Formally, it looks like this:
 m. ∃χΦ
 n. Φ [χ / c*] *: where c does not
occur
 o. Ψ outside the subproof
 p. Ψ

 Scary! The basic idea: This rule lets me

‘name’ the unknown thing in the existential
statement, and then reason about it.
PREDICATE LOGIC
 Imagine, for instance, introducing the name
‘Jack the Ripper’ to refer to whoever
committed the famous London murders in
1888.
I don’t really know who Jack is – it’s just a
convenient label.
 I could then reason about Jack:
 All murderers have a motivation. Jack the Ripper
is a murderer, so he must have a motivation,
too.

 But all through this, ‘Jack’ is functioning as a

PLACEHOLDER name.
PREDICATE LOGIC
 When we do this sort of reasoning, we isolate
the reasoning inside a subproof, and we only
use the placeholder name inside the
subproof.
 You might think of it this way – it’s okay to call
the murderer Jack and reason using that name,
but it’s not okay to conclude that the murderer
really IS named Jack!
PREDICATE LOGIC
 So: we start our subproof by ASSUMING a
placeholder name for our unknown
individual.
 If we manage (within the subproof) to get to

a conclusion that does not use the

placeholder name, we can end the subproof
and repeat that conclusion back in the main
proof.
PREDICATE LOGIC
 Here is an example:
 1. ∃x(Mx & Lx)
 2. ∀x(Mx → Tx)
 3. Mj & Ljfor ∃E
 4. Mj &E3
 5. Mj → Tj ∀E 2
 6. Tj →E 4,5
 7. ∃xTx ∃I 6
 8. ∃xTx ∃E 3-7
PREDICATE LOGIC
 It’s very important that your assumption use
a NEW name (not just an arbitrary one). This
name should not appear previously in the
proof, anywhere.
 In addition, you may not use it outside the

proof – your subproof cannot end with a line

still containing the name.
PREDICATE LOGIC
 Why? Because of how our ∀I rule is phrased.
That rule lets us introduce a ∀ on any name
that is arbitrary – that is, is not in the
premises or in the assumption of an OPEN
subproof.
 If we let a ∃I proof end with the name still

‘active’, that would allow it to count as

arbitrary. But it shouldn’t be: ‘Jack the
Ripper’ isn’t arbitrary, it stands for the
person who committed the London murders.
PREDICATE LOGIC
 1. ∃x(Mx & Lx)
 2. ∀x(Mx → Tx)
 3. Mj & Ljfor ∃E
 4. Mj &E3
 5. Mj → Tj ∀E 2
 6. Tj →E 4,5
 7. Tj ∃E 3-6
 8. ∀xTx ∀I 7

 Note that allowing ‘j’ out of the subproof

leads to the invalid move in 8.
PREDICATE LOGIC
 It can be hard to keep the particular
restrictions on name use in the new rules
straight. Here’s a summary:
 Universal elimination (∀E) : The name involved
doesn’t need to be arbitrary or new, but you
must replace all free variables that result from
dropping the ∀ with the name.
 Universal introduction (∀I) : The name involved
in using this rule (that is, the name you are
replacing) must be arbitrary. You must replace
all instances of the name with the variable of the
introduced quantifier.
PREDICATE LOGIC
 It can be hard to keep the particular
restrictions on name use in the new rules
straight. Here’s a summary:
 Existential elimination (∃E) : The name you
introduce must be new, and cannot be used
again outside of the subproof. It must replace all
free variables that result from dropping the ∃.
 Existential introduction (∃I) : No restriction on
the name you replace; and you may replace any
or all instances of the name.
PREDICATE LOGIC
 We now have all the quantifier rules. But we
still have a bit more to do – we’re going to
expand QL with one more logical piece.
 Before we do, a brief word about derived

rules – Ichikawa gives a derived rule for

moving between ~∀xFx and ∃x~Fx, and
between ~∃xFx and ∀x~Fx.
 Theseare two pairs of logically equivalent
formulae.
PREDICATE LOGIC
 Again, we will skip the derived rules here –
do not use them on the HWs or tests.
 This will actually make things easier for you –

less rules to remember, and it forces me to

construct less complicated proof questions.
PREDICATE LOGIC
 We now have a complete proof system for
QL. However, logicians like to add a little
extra to QL, to make it even stronger.
 So our final step is to learn ‘QL plus’ –

otherwise known as quantified logic with

IDENTITY.
PREDICATE LOGIC
 Identity is a logical notion that lets us
express that two elements of the UD are
IDENTICAL – that is, are literally the same
thing.
 We already know that, like the real world,

objects can have multiple names in QL.

 Identity lets us express things like:
 Batman is Bruce Wayne.
PREDICATE LOGIC
 Note that the word ‘is’ in English is used in
different ways. One is the ‘is’ of
PREDICATION:
 Batman is strong.
 We translate this in QL as an atomic
sentence, Sb.
PREDICATE LOGIC
 But the ‘is’ in “Batman is Bruce Wayne” is
logically a bit different.
 Suppose we translated that treating ‘is

Bruce Wayne’ as a predicate: Wb.

 Now consider the following argument:
 BruceWayne is rich. Batman is Bruce Wayne.
Therefore, Batman is rich.
PREDICATE LOGIC
 If we try to translate this using only the ‘is’ of
predication, we get:
 1. Rw
 2. Wb
 3. Rb

 This treats ‘Bruce Wayne’ as a name and ‘is

Bruce Wayne’ as a predicate – and misses
the relationship between them.
 The translated argument ends up invalid!
PREDICATE LOGIC
 We might try adding a two-place predicate
for ‘is identical to’. So ‘Batman is Bruce
Wayne’ could be Ibw.
 But now our argument is:
 1. Rw
 2. Ibw
 3. Rb

 This still doesn’t help!

PREDICATE LOGIC
 Perhaps you’re thinking: we should introduce
special rules for the identity predicate, to
capture this sort of inference.
 But the trouble is – we can’t do that for a

normal predicate like Ixy.

 Predicates in QL are not given a constant

interpretation: Ixy might mean ‘is identical

to’ in one problem and ‘is interested in’ in
another.
PREDICATE LOGIC
 So we need to treat identity differently in QL,
by giving it its own, consistently interpreted
logical symbol: =.
 Now, we can translate ‘Batman is Bruce

Wayne’ as:
b =w

 Essentially, this tells us that the two names b

and w refer to the same member of the UD.
PREDICATE LOGIC
 This new symbol means that we’ll need to
tweak our QL grammar a bit, specifying that
for any two names α and β, α = β is an
atomic sentence (like Fb or Lab).
 It’s also customary to add a notational

shortcut, using α ≠ β instead of ~(α = β).

PREDICATE LOGIC
 Identity lets us symbolize some cool things
that we couldn’t properly symbolize before.
 For instance, claims about uniqueness:
 Only Mary is happy.

 Without identity, we could say something

like:
 Hm

 But this is consistent with other objects being

happy!
PREDICATE LOGIC
 With identity, we can say:
 Hm & ∀x(Hx → x=m)

 This says Mary is happy, and all the happy

objects are identical to Mary.
 In other words – if a substituted name makes Hx
come out true, that name must refer to Mary
rather than anyone else.
PREDICATE LOGIC
 Alternately, we could translate it this way:
 Hm & ~∃x(Hx & x≠m)

 Or even:
 ∀x(Hx↔x=m)
PREDICATE LOGIC
 Identity also allows us to make claims about
particular numbers of objects.
 For example: There are at least two happy

things:
 ∃x∃y((Hx & Hy) & x≠y)
 There are at least three happy things:
 ∃x∃y∃z(((((Hx & Hy) & Hz) & x≠y) & x≠z) & y≠z)
 Obviously this gets complex quick! (you
won’t ever need to go above 2 in this class)
PREDICATE LOGIC
 What if I want to say there are exactly two
happy things?
 ∃x∃y(((Hx & Hy) & x≠y) & ∀z (Hz→(z=x v z=y)))
 (You won’t need to reproduce this either!)
PREDICATE LOGIC
 Of course, now that we’ve extended our
language we’ll need two new rules for our
natural deduction system.
 IDENTITY INTRODUCTION is easy:

 m. α= α =I

 This just says that, at any time, I can write

down an identity statement flanked by two
instances of the same name. (Any name!)
PREDICATE LOGIC
 Of course, now that we’ve extended our
language we’ll need two new rules for our
natural deduction system.
 IDENTITY INTRODUCTION is easy:

 m. α= α =I

 Notice I don’t cite any lines! This is because

it is a logical truth that everything is self-
identical.
PREDICATE LOGIC
 IDENTITY ELIMINATION is only slightly harder:
 m. α = β

 n. Φ

 o. Φ[[α / β]] OR Φ [[β / α]] = E m,n

 Basically, if you have an identity statement

between two names, then you can copy any
line and replace any or all instances of one
name with the other.
PREDICATE LOGIC
 IDENTITY ELIMINATION is only slightly harder:
 m. α = β

 n. Φ

 o. Φ[[α / β]] OR Φ [[β / α]] = E m,n

 This reflects the logical fact that if a=b, then

anything that is true of a is true of b.
PREDICATE LOGIC
 For example, if Bruce Wayne is rich then
Batman is too:
 1. Rw
 2. w=b
 3. Rb =E 1,2
PREDICATE LOGIC
 Notice we can choose to only substitute
some of the occurrences of the name:
 1. w=b
 2. Lbb
 3. Lwb =E 1,2

 If Batman loves himself, then it follows that

Bruce Wayne loves Batman.
PREDICATE LOGIC
 There are some tricky philosophical problems
that come with the principle “if a=b, then
anything that is true of a is true of b”.
 For instance:
 The Joker believes that Batman is in the building.
 The Joker believes that Bruce Wayne is in the
building.
PREDICATE LOGIC
 On ‘QL plus’, this does follow. And in a sense
it is true.
 Suppose the Joker believes that Batman is in

the building. If I were to point to the

individual that that belief is about, I would
point to Bruce Wayne/Batman.
 In another sense, it is false: if I asked the

Joker “Is Bruce Wayne in the building?” he

might say no.
PREDICATE LOGIC
 Predicate logic skips over these difficulties. In
predicate logic, the claim “if a=b, then
anything that is true of a is true of b” always
holds.
PREDICATE LOGIC
 Let’s look at a few example proofs.
 Proof 1: we’ll show that identity is

SYMMETRIC – i.e., that ∀x∀y(x=y → y=x)

 This is a logical truth, so we should be able to

do it with no premises.
PREDICATE LOGIC
 It will be a little tricky though. Since we have
no premises, we will need to use a subproof.
What kind of subproof?
 If we could prove an arbitrary instance of the

conditional x=y → y=x, we could use ∀I.

 So let’s try to prove a=b → b=a.
PREDICATE LOGIC
 1. a=b want b=a
 2. b=b =I
 3. b=a =E 1,2
 4. a=b → b=a →I 1-3
 5. ∀y(a=y → y=a) ∀I 4
 6. ∀x∀y(x=y → y=x) ∀I 5

 Notice a couple of things about this proof.

PREDICATE LOGIC
 1. a=b want b=a
 2. b=b =I
 3. b=a =E 1,2
 4. a=b → b=a →I 1-3
 5. ∀y(a=y → y=a) ∀I 4
 6. ∀x∀y(x=y → y=x) ∀I 5

 First, we made use of the fact that you can

choose to only substitute some instances
with =E (in line 3, I only substituted in the
second slot).
PREDICATE LOGIC
 1. a=b want b=a
 2. b=b =I
 3. b=a =E 1,2
 4. a=b → b=a →I 1-3
 5. ∀y(a=y → y=a) ∀I 4
 6. ∀x∀y(x=y → y=x) ∀I 5

 Second, it’s fine to apply ∀I to line 4, because

although a and b occur in an assumption, it is
an assumption of a CLOSED subproof.
PREDICATE LOGIC
 Another proof: From ∃x(Fx & x=a), let’s prove
Fa.

 1. ∃x(Fx & x=a)

 2. Fb & b=a for ∃E
 3. Fb 2 &E
 4. b=a 2 &E
 5. Fa 3,4 =E
 6. Fa 2-5, ∃E

 Notice that I needed to use b here, since a

would not be a new name!
PREDICATE LOGIC
 That’s all the content for the course! Let’s
finish up by talking about the final exam.
 The final exam will be semi-comprehensive.

It will contain a few questions on:

 Multiple choice involving concepts from earlier in
the semester – validity, logical equivalence, etc.
(see list for midterm)
 Natural deduction in SL
 The majority of questions (about 2/3) will be
on QL.
PREDICATE LOGIC
 QL topics:
 Translations. These will include nested
quantifiers (2 deep, as in ∀x∃yLxy). They will also
include identity, including how to translate “at
least two” (but not any of the more complex
examples).
 Natural deduction problems in QL, plus identity.
 Multiple choice questions on conceptual topics:
 Differences between SL and QL (how QL is stronger,
e.g.)
 Quantifer scope

 Free/bound variables

 Models in QL

 Identity (e.g., why we need the = symbol)

PREDICATE LOGIC
 QL topics:
 Interpretation building. Two types of question:
 I give you an interpretation and ask you whether a QL
sentence is true on that interpretation.
 I give you one or more QL sentences and ask you to

make an interpretation where the sentences are true,

false, etc. I might ask you to use a model to show that
a sentence is not a tautology, or to show that two
sentences are not equivalent.
 For the second type of question you will need to give a

UD, a specification of names, and an extension for

predicates in set notation.
PREDICATE LOGIC
 The format and procedures for the final exam
will be the same as for the midterm.
 There will be more details posted to Moodle

soon about:
 Room assignments for the final exam
 Scheduling for the review session
 If you have other questions about the exam
procedures, please contact the grader.
 A set of practice questions will be available

on Moodle for revision purposes.