Lecture 26: Proof Strategies & Examples

1 8x(A(x) B(x))
2 ¬A(a) ^ ¬B(a)
...
n ?
n + 1 ¬(¬A(a) ^ ¬B(a)) ¬I 2–n
12 8x¬(¬A(x) ^ ¬B(x)) 8I n + 1
12 + 1 8x(A(x) B(x)) ! 8x¬(¬A(x) ^ ¬B(x)) !I 1–12
Now all we have to do is to get the contradiction in line n . . .
This might be the time to use 8 E on our main premise, which will give us a disjunction.
From there, E will get us further. . .
Here is the completed derivation:
1 8x(A(x) B(x))
2 ¬A(a) ^ ¬B(a)
3 A(a) B(a) 8E 1
4 A(a)
5 ¬A(a) ^E 2
6 ? ¬E 4, 5
7 B(a)
8 ¬B(a) ^E 2
9 ? ¬E 7, 8
10 ? E 3, 4–6, 7–9
11 ¬(¬A(a) ^ ¬B(a)) ¬I 2–10
12 8x¬(¬A(x) ^ ¬B(x)) 8I 11
13 8x(A(x) _ B(x)) ! 8x¬(¬A(x) ^ ¬B(x)) !I 1–12
25.4.2 De Morgan’s Law with existential quantifiers
We now turn to a version of De Morgan’s Law with existentially quantified formulas:

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–8
` (9x (A(x) _ B(x)) ! 9x ¬(¬A(x) ^ ¬B(x))) .
What do we need to change to adapt to this?
• Here the overall strategy remains the same, but with one crucial difference:
it is impossible to apply 9 I as your before last step!
• Why?
Because the name that we would introduce for 9 E and
which we will use to derive the two contradictions
would feature in the conclusion for 9 E
— but, then we cannot export that conclusion, because of the restriction of 9 E!
• So, instead, we have to perform the 9 I within the subproof for 9 E.
In this way we get rid of the the name in the conclusion.
And then we can just export the conclusion!
This proof is an example that illustrates the earlier claim that you should get used to the idea
of assuming that 9 E is your last step if you have an existential formula in the premises.
In doing so, you’ll avoid hitting a potential dead-end!
Finally, here’s the proof:

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–9
1 9x(A(x) B(x))
2 A(a) B(a)
3 ¬A(a) ^ B(a)
4 A(a)
5 ¬A(a) ^E 3
6 ? ¬E 4, 5
7 B(a)
8 ¬B(a) ^E 3
9 ? ¬E 7, 8
10 ? E 2, 4–6, 7–9
11 ¬(¬A(a) ^ ¬B(a)) ¬I 3–10
12 9x¬(¬A(x) ^ ¬B(x)) 9I 11
13 9x¬(¬A(x) ^ ¬B(x)) 9E 1, 2–12
14 9x(A(x) B(x)) ! 9x¬(¬A(x) ^ ¬B(x)) !I 1–13
It is instructive to compare this proof with the previous one!

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–10
25.5 Equivalence of uniqueness translations
When discussing symbolizations of “there is exactly one hero”, we claimed that
9x (H(x) ^ 8y (H(y ) ! x = y )) and 9x8y (H(y ) $ x = y )
are equivalent.
But we also stated that this needed to be proved. Well, now we are in a position to prove it!
To show the equivalence expressed as biconditional would be quite long.
So, let’s prove something equivalent, namely the two implications: (9x (H(x) ^ 8y (H(y ) ! x = y )) ! 9x8y (H(y ) $ x = y ))
and
(9x8y (H(y ) $ x = y ) ! 9x (H(x) ^ 8y (H(y ) ! x = y )))
We begin by proving the first conditional.
25.5.1 ` (9x (H(x) ^ 8y (H(y ) ! x = y )) ! 9x8y (H(y ) $ x = y ))
Strategy:
• Clearly, the last step in the proof will be ! I.
If I take my consequent as my conclusion, then 9 I is a candidate for my before last
step.
However, looking at the antecedent (which I will assume in my first step), I notice that
it is an existential formula.
• Therefore, it looks like that my before last step will be 9 E,
and that I will apply the 9 I rule inside that subproof.
• Notice that before introducing the existential quantifier, I need to introduce a universal
quantifier (with 8 I).
But that shouldn’t be too much trouble: I can get that name from 8 E in my an-
tecedent.
I just need to be careful to choose a different name from the one I chose for 9 E.

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–11
• The final bit of strategy: I will absolutely need to set up two subproofs for $ I to derive
my consequent.
Here’s the full proof:
1 9x(H(x) ^ 8y (H(y ) ! x = y ))
2 H(a) ^ 8y (H(y ) ! a = y )
3 H(c)
4 8y (H(y ) ! a = y ) ^E 2
5 H(c) ! a = c 8E 4
6 a = c !E 3, 5
7 a = c
8 H(a) ^E 2
9 H(c) =E 7, 8
10 H(c) $ a = c $I 3–6, 7–9
11 8y (H(y ) $ a = y ) 8I 10
12 9x8y (H(y ) $ x = y ) 9I 11
13 9x8y (H(y ) $ x = y ) 9E 1, 2–12
14 9x (H(x) ^ 8y (H(y ) ! x = y )) ! [· · · ] !I 1–13
Next we will prove the second conditional from Section 25.5.
25.5.2 ` (9x8y (H(y ) $ x = y ) ! 9x (H(x) ^ 8y (H(y ) ! x = y )))
Strategy:
• Clearly, my last step in the proof will be ! I.
Again, if I take my consequent as the conclusion, then 9 I is a candidate for my last
step.
Accordingly, I will assume the antecedent in my first step. However, then I notice that
again I have an existential formula.
• So given this, my before last step will likely be 9 E,
and I will apply the 9 I rule inside that subproof.

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–12
• In some ways, this proof is easier than the previous one.
I can use $ E, since now the biconditional is in my antecedent.
But, we still must be careful: if I choose the same name for 8 E as I did for 9 E,
I still have to somehow derive one of the horns of the bi-implication.
Luckily, I can always derive a = a from nothing, which motivates the aim of reducing
the amount of names in play.
• The final bit of strategy: I will need to also show that a universal holds.
So for that part, I will need to use a new name, since whatever name I choose for 9 E
will be in the assumption of an undischarged subproof.
Proof:
1 9x8y (H(y ) $ x = y )
2 8y (H(y ) $ a = y )
3 H(a) $ a = a 8E 2
4 a = a =I
5 H(a) $E 3, 4
6 H(c)
7 H(c) $ a = c 8E 2
8 a = c $E 6, 7
9 H(c) ! a = c !I 6–8
10 8y (H(y ) ! a = y ) 8I 9
11 H(a) ^ 8y (H(y ) ! a = y ) ^I 5, 10
12 9x(H(x) ^ 8y (H(y ) ! x = y )) 9I 11
13 9x(H(x) ^ 8y (H(y ) ! x = y )) 9E 1, 2–12
14 9x8y (H(y ) $ x = y ) ! [· · · ] !I 1–13
25.6 Proofs with numerical claims

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–13
25.6.1 How many pink elephants are in the room?
As you’ve probably noticed by now, one use we can put our FOL to do is to express claims
involving quantities, so it’s also important that our Fitch system be able to prove things
about claims involving quantities.
Let’s consider an example:
There’s a pink elephant in the room.
There’s at most one pink elephant in the room.
) There’s exactly one pink elephant in the room.
Translated into FOL:
9x P(x)
8x8y ((P(x) ^ P(y )) ! x = y )
) 9x (P(x) ^ 8y (P(y ) ! x = y ))
Indeed, our system would not be very good if we couldn’t prove, given those two premises,
that there is exactly one pink elephant in the room!
So let’s look at the inference again:
9x P(x), 8x8y ((P(x) ^ P(y )) ! x = y ) ` 9x (P(x) ^ 8y (P(y ) ! x = y ))
Strategy:
• My conclusion has an existential as the main connective/quantifier.
So, again, 9 I is a candidate for my last step.
• But in my premises, the first one is an existential formula.
So, this makes me fairly confident 9 E will be my last step and that I’ll just do 9 I inside
that subproof.
• One complication is that I will also need to do a 8 I inside this subproof to get one of
the conjuncts of the conclusion.
• I have two universals in my second premise.
So it seems that, if I play my wild cards right, I can choose one of these 8 E to be the
same name as the one I will choose for 9 E.

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–14
But, I must choose a different one to get something I can generalize with the 8 I rule.
So I’ll choose a different one for the second 8 E.
• The other complication is that I will need to do 8 I on a general conditional.
So I won’t introduce the name for 8 I directly from 8 E, but rather save up both my
8 E until I am inside the subproof for the general conditional.
Proof:
1 9x P(x)
2 8x8y ((P(x) ^ P(y )) ! x = y )
3 P(a)
4 P(c)
5 8y ((P(a) ^ P(y )) ! a = y ) 8E 2
6 (P(a) ^ P(c)) ! a = c 8E 5
7 P(a) ^ P(c) ^I 3, 4
8 a = c !E 6, 7
9 P(c) ! a = c !I 4–8
10 8y (P(y ) ! a = y ) 8I 9
11 P(a) ^ 8y (P(y ) ! a = y ) ^I 3, 10
12 9x(P(x) ^ 8y (P(y ) ! x = y )) 9I 11
13 9x(P(x) ^ 8y (P(y ) ! x = y )) 9E 1, 3–12
25.6.2 The Drinker’s Paradox
There’s one last proof I want to show you. It’s a hard one, but it’s beautiful.
` 9x (P(x) ! 8y P(y ))
• We’ve mentioned that the Drinker’s Paradox is a first-order validity.
It is usually formulated as:
“There is someone in the bar such that, if they are drinking, then everyone is drinking”.

© Dirk Schlimm, McGill University, PHIL 210, 2025. • Do not distribute this document. • 25–15
To be a first-order truth means that the sentence 9x (P(x) ! 8y P(y )) is true in all
non-empty domains!
• Essentially, it is a first-order truth because it depends on the Law of Excluded Middle.
a) If everyone in the bar is drinking, then the implication is obviously true.
b) If not everyone is drinking, then there must be someone in the bar who’s not
drinking (quantifier interchange). And this person is the witness who will make
the antecedent of the conditional false, so the whole conditional true.
(It’s true that if they were drinking then maybe everyone would be drinking.)
That’s why the trick is usually framed as: “Pick the person least likely to be drinking”.
• We could build a proof along the lines just sketched:
this would involve simulating LEM as well as proving the quantifier equivalence.
This is quite intuitive, but very long!
• Instead, I’ll show you what is possibly the shortest proof of the Drinker’s Paradox,
using IP as the main strategy.
Unfortunately, it comes at the price of not being immediately obvious why it works. . .
Proof:
1 ¬9x(P(x) ! 8yP(y ))
2 P(a)
3 ¬P(b)
4 P(b)
5 ? ¬E 3, 4
6 8yP(y ) X 5
7 P(b) ! 8yP(y ) !I 4–6
8 9x(P(x) ! 8yP(y )) 9I 7
9 ? ¬E 1, 8
10 P(b) IP 3–9
11 8yP(y ) 8I 10
12 P(a) ! 8yP(y ) !I 2–11
13 9x(P(x) ! 8yP(y )) 9I 12
14 ? ¬E 1, 13
15 9x(P(x) ! 8yP(y )) IP 1–14