Professional Documents
Culture Documents
1 Introduction to Metatheory 2
1.1 Metatheory: What? . . . . . . . . . . . . . . . . . . . . . . . . 2
3 Soundness 16
3.1 Soundness - The Basic idea . . . . . . . . . . . . . . . . . . . 16
3.2 Soundness of the Simpler System SD& . . . . . . . . . . . . . 17
3.3 Soundness for SD& - in the form presented in the textbook . . 27
3.4 A complication - extra assumptions . . . . . . . . . . . . . . . 37
3.5 Aside: Some Useful Facts . . . . . . . . . . . . . . . . . . . . . 43
3.6 A Slightly Trickier Case: Conditional Introduction . . . . . . . 45
3.7 Negation Introduction . . . . . . . . . . . . . . . . . . . . . . 47
1
1. Introduction to Metatheory
1.1. Metatheory: What?
2
The concepts of these two types of good argument
are entirely different.
But in fact, as well learn in the next two weeks these two
concepts count exactly the same arguments from SL as
good ones.
3
There are two big results and one medium-sized result on
the agenda for the coming three weeks.
In symbols, S S
The second big result will be the topic for the following
two weeks:
In symbols: S S
4
2. Adequate sets of connectives
5
Now you might ask: is there a truth-function or truth-
table that we cant write as a sentence in the language
SL?
6
The proof of this turns out to be straightforward: there
is a mechanical procedure that takes a truth-table cor-
responding to a truth function and gives you a sentence
in SL containing just &, and , which has the given
truth-table.
Example:
A B ?
T T F
T F T
F T F
F F T
You want to produce a sentence of SL that has exactly
that truth-table.
7
Lets break down the truth-table line by line. Heres the
rst line:
A B ?
T T F
You can think of this line as this outcome for the two
possibilities A and B: A happens, and B happens.
8
Now lets look at the second line:
A B ?
T F T
You can think of this line as this outcome for the two
possibilities A and B: A happens, and B doesnt hap-
pen.
9
So lets look at the truth table, indicating for each line
the associated sentence that would be true if that line is
the one that actually happens.
Associated Sentence A B ?
A&B T T F
A& B T F T T-table says this one might occur
A&B F T F
A& B F F T T-table says this one might occur
(A& B) (A& B)
10
In table form:
A B (A& B) (A& B)
T T F
T F T
F T F
F F T
11
This procedure can be carried out for any truth table.
Say you are given this truth table:
A B C ?
T T T F
T T F F
T F T T
T F F F
F T T F
F T F T
F F T F
F F F T
12
Then write the associated sentences: the conjunction of
sentence letters (when the sentence letter gets T on that
row) and negated sentence letters (when the sentence let-
ter gets F on that row):
A B C ?
T T T F
T T F F
A& B&C T F T T
T F F F
F T T F
A&B& C F T F T
F F T F
A& B& C F F F T
13
Further observations:
14
In fact, you can give a single connective that is adequate
for all truth-functions.
You can see the truth-table for the two universal connec-
tives in problems p. 243 6.2 E # 5 and # 6.
15
3. Soundness
3.1. Soundness - The Basic idea
There are two things you need to note about the proof of
the soundness theorem.
16
3.2. Soundness of the Simpler System SD&
From the proof for this system we can get a sense of what
the inductive argument overall looks like. The full sound-
ness theorem diers from this one only in that you have
more rules to check in the inductive step.
17
Claim: SD& S S. (i.e. SD& is sound.)
18
OK, now lets think about what it means to say
{P1, P2, . . . , Pn} SD& S.
...
...
...
19
Lets look at the sentences in the lines between the premises
and the conclusion S, and ask: what truth-value do these
sentences get?
20
Here is a key point: If we start out with a string of Ts,
and at some point we end up with an F, there must be a
first line at which an F appears:
21
To repeat: if you have a derivation in SD& whose premises
are assigned true and whose conclusion is assigned false,
then there must be a rst line of the derivation at which
the sentence on that line is assigned false.
22
To be clear on the point, Ill review how we got here.
23
How do we show that? Well, fortunately, SD& involves
just two rules, and they are both simple. We have two
possibilities, putting a magnifying glass on the lines where
the F rst appears:
We could have:
24
Or we could have:
In both of these cases, you can just draw the truth tables
to show that this cant happen (I will leave it as an exer-
cise for you to draw them):
25
And that, in a nutshell, is the inductive part of the sound-
ness proof. The proof for the full system SD just involves
checking more rules.
But the basic idea remains the same: Show that no sin-
gle inference in SD can take you from true premises to a
false conclusion.
26
3.3. Soundness for SD& - in the form presented in the textbook
Now I want to make the proof a bit closer to the full proof
for SD.
I still want to stick with the simpler system SD& for this
example, but Ill set up the argument as an explicit in-
ductive argument.
27
Instead of thinking of ourselves as saying:
28
OK: Say we are given a set of sentences, and SD& S
using a proof of k lines. We want to show that S.
1 S Assumption
S.
29
Now for the induction step. Consider k > 0. As the
induction hypothesis, we assume that:
30
We just have to check two cases: on the last line, S can
be justied using & E or S can be justied using &I.
Sentences from
...
i S1
... ...
j S2
... ...
k S1&S2 i, j &I
31
Note that both S1 and S2 are derived from by deriva-
tions with fewer than k lines.
This means that they fall within the scope of the induc-
tion hypothesis, which gives us:
S1 and S2.
32
Now we reason this way: Say we have a truth-assignment
W that makes every sentence in true.
33
Then we have one of these two situations:
Sentences from
...
i S1&S2
... ...
k S1 i, &E
Sentences from
...
i S1&S2
... ...
k S2 i, &E
34
This means that it falls within the scope of the induction
hypothesis, which gives us:
S1&S2.
So: S1 and S2
35
So thats it: Thats the complete soundness proof for
SL&. The full induction proof for SL is more compli-
cated, but the basic structure of the induction is the same.
But also, the rules of & I and & E are particularly sim-
ple as far as this proof is concerned because they dont
require any additional assumptions in the course of the
proof.
36
3.4. A complication - extra assumptions
37
Say we have a derivation that looks like this - omitting
all but the relevant details.
Sentences from
..
.
i S1
.. ..
. .
i1 S1a
.. ..
. .
j S2
.. ..
. .
j1 S2a
k S3
38
(Repeating the diagram to be able to refer to it without scrolling back.)
Sentences from
..
.
i S1
.. ..
. .
i1 S1a
.. ..
. .
j S2
.. ..
. .
j1 S2a
k S3
39
Its important to consider at each line the assump-
tions that are being presupposed at that line - which we
track in the derivation with the horizontal and vertical
lines - not the assumptions that have been made at some
point or other in the derivation.
40
In the textbook, they use this terminology to describe this
situation: S2 is an open assumption in whose scope S2a
lies.
41
Returning to our example:
Sentences from
..
.
i S1
.. ..
. .
i1 S1a
.. ..
. .
j S2
.. ..
. .
j1 S2a
k S3
42
3.5. Aside: Some Useful Facts
43
Some other handy facts are listed on p. 245 of the text-
book. Ill refer you there to look them up and get the
arguments. They are all straightforward reasoning with
truth-assignments and truth-tables.
Ill note one useful fact in particular for the next part of
the lecture; Ill call it Useful Fact 2:
44
3.6. A Slightly Trickier Case: Conditional Introduction
But this time, lets say that our system is SD, and that
Sk+1 is inferred from earlier lines by Conditional In-
troduction.
45
That will look like this:
Assumptions
..
.
i Si A/ I
.. ..
. .
j Sj
k+1 Si Sj (= Sk+1 ) i j, I
46
3.7. Negation Introduction
47
NB: This version is simpler than the one in the book,
in that I am skipping some steps that involve the idea of
the derivation is accessible at position k. That part
of the textbook proof is just bookkeeping - you need to
make sure that any inference you make only appeals to
assumptions its allowed to appeal to, and you want to
keep track of what the scope of given assumptions are.
Here in lecture I will treat this as obvious from the dia-
gram of the derivation.
48
That is, I am doing things more bluntly than the text-
book does to keep the main idea from being snowed under
with bookkeeping subtleties.
49
Our derivation has this shape:
j P
l R
m R
k + 1 P
50
Since l < k + 1 and m < k + 1, we can apply the induc-
tion hypothesis to obtain:
l R and m R, so:
51
Since k+1 {P} R and k+1 {P} R we know
that any truth-assignment making all of k+1 {P} true
makes R true and R true.
52