Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
(x
1
, . . . , x
n
) is false ,
A
(x
1
, x
2
) = ( x
1
is true or x
1
is false ) and ( x
2
is true or x
2
is false ),
P
(x
1
, x
2
) = both x
1
and x
2
are true ;
Logical disjunction: its the symbol and we have
for n ,= 2 A
(x
1
, . . . , x
n
) is false ,
A
(x
1
, x
2
) = ( x
1
is true or x
1
is false ) and ( x
2
is true or x
2
is false ),
P
(x
1
, x
2
) = at least one between x
1
and x
2
is true ;
Logical implication: its the symbol and we have
for n ,= 2 A
(x
1
, . . . , x
n
) is false ,
A
(x
1
, x
2
) = ( x
1
is true or x
1
is false ) and ( x
2
is true or x
2
is false ),
P
(x
1
, x
2
) = x
1
is false or x
2
is true ;
Double logical implication: its the symbol and we have
for n ,= 2 A
(x
1
, . . . , x
n
) is false ,
A dierent approach to logic 11
A
(x
1
, x
2
) = ( x
1
is true or x
1
is false ) and ( x
2
is true or x
2
is false ),
P
(x
1
, x
2
) = P
(x
1
, x
2
) and P
(x
2
, x
1
) ;
Logical negation: its the symbol and we have
for n > 1 A
(x
1
, . . . , x
n
) is false ,
A
(x
1
) is true,
P
(x
1
) = x
1
is false ;
Universal quantier: its the symbol and we have
for n > 1 A
(x
1
, . . . , x
n
) is false ,
A
(x
1
) = x
1
is a set and for each x in x
1
(x is true or x is false),
P
(x
1
) = for each x in x
1
(x is true) .
Existential quantier: its the symbol and we have
for n > 1 A
(x
1
, . . . , x
n
) is false ,
A
(x
1
) = x
1
is a set and for each x in x
1
(x is true or x is false),
P
(x
1
) = there exists x in x
1
such that (x is true) .
Membership predicate: its the symbol and we have
for n ,= 2 A
(x
1
, . . . , x
n
) is false ,
A
(x
1
, x
2
) = x
2
is a set,
P
(x
1
, x
2
) = x
1
is a member of x
2
;
Equality predicate: its the symbol = and we have
for n ,= 2 A
=
(x
1
, . . . , x
n
) is false ,
A
=
(x
1
, x
2
) is true,
P
=
(x
1
, x
2
) = x
1
is equal to x
2
.
We can think and use also other operators, for instance operations between sets such
as union or intersection can be represented through an operator, etc. .
In the standard approach to logic, quantiers are not treated like the other logical
connectives, but in this system we mean to separate the operation of applying a quanti-
er from the operation whereby we build the set to which the quantier is applied, and
therefore the quantier is treated as the other logical operators (altogether, the universal
quantier is simply an extension of logical conjunction, the existential quantier is simply
an extension of logical disjunction).
With regard to the operation of building a set, we need a specic symbol to indicate
that we are doing this, this symbol is the symbol which we will consider as a unique
symbol.
Besides the set builder symbol, we need parentheses and commas to avoid ambiguity
in the reading of our expressions; to this end we use the following symbols: left parenthe-
sis (, right parenthesis ), comma , and colon :. We can indicate this further set of
12 M. Avon
symbols with :.
To avoid ambiguity in reading our expressions we require that the sets 1, (, T and
: are disjoint. Its also requested that a symbol does not correspond to any chain of
more symbols of the language. More generally, given
1
, . . . ,
n
and
1
, . . . ,
m
symbols
of our language, and using the symbol | to indicate the concatenation of characters and
strings, we assume that equality of the two chains
1
| . . . |
n
and
1
| . . . |
m
is achieved
when and only when m = n and for each i = 1 . . . n
i
=
i
.
While the set : will be always the same, the sets 1, (, T may change according to
what is the language that we describe. To describe our language it is required to know
the sets 1, (, T and the function # which associates a meaning to every element of (. In
other words, our language is identied by the 4-tuple (1, T, (, #). Since the meaning of
an operator is not a mathematical object, operators must be seen as symbols which are
tightly coupled with their meaning.
Before we can describe the process of constructing expressions we still need to in-
troduce some notation. In fact in that process well use the notion of context and the
notion of state. Context and states have a similar form, and here we want to dene their
common form.
We dene T = 1, . . . , m[ m is a positive integer.
Suppose x is a function whose domain dom(x) belongs to T. Suppose C T is such
that C dom(x). Then we dene x
/C
as a function whose domain is C and such that
for each j C x
/C
(j) = x(j) .
Suppose x and are two functions with the same domain D, and D T. Then we
say that (x, ) is a state-like pair.
Given a state-like pair k = (x, ) the domain of x will be also called the domain of k.
Therefore dom(k) = dom(x) = dom().
Furthermore dom(k) T and given C T such that C dom(k) we can dene
k
/C
= (x
/C
,
/C
). Clearly k
/C
is a state-like pair.
We dene !(k) = k
/C
[ C T, C dom(k).
Given another state-like pair h we say that h _ k if and only if h !(k) .
Suppose h !(k), then there exists C T such that C dom(k), h = k
/C
=
(x
/C
,
/C
). Therefore dom(h) = C and k
/dom(h)
= k
/C
= h.
Suppose h !(k) and g !(h). This means there exist C T such that C
dom(k), h = k
/C
, and there exist D T such that D dom(h), g = h
/D
. So
D dom(h) = C dom(k), g = (k
/C
)
/D
= (x
/C
,
/C
)
/D
= (x
/D
,
/D
) = k
/D
. There-
fore g !(k).
A dierent approach to logic 13
Suppose k = (x, ) is a state-like pair whose domain is D. Suppose (y, ) is an ordered
pair. Then we can dene the addition of (y, ) to k.
Suppose D = 1, . . . , m, then we dene D
= 1, . . . , m+1. We dene x
as a function
whose domain is D
() = x(), and x
(m+ 1) = y. We
dene
() = (),
). Obviously (k + (y, ))
/{1,...,m}
= k,
so k !(k + (y, )).
If D = then clearly D
= 1. We dene x
such that
x
(1) = y. We dene
such that
(1) = . Then we
dene k + (y, ) = (x
). Obviously (k + (y, ))
/
= = k, so k !(k + (y, )).
In both cases k + (y, ) is a state-like pair, and k !(k + (y, )).
We also dene = (, ), so is a state-like pair.
In the next lemma we prove that, when a state-like pair is obtained as k +(y, ), then
k, y, and are univocally determined.
Lemma 2.1. Suppose k
1
= (x
1
,
1
) is a state-like pair whose domain is D
1
, and (y
1
,
1
)
is an ordered pair. Suppose k
2
= (x
2
,
2
) is a state-like pair whose domain is D
2
, and
(y
2
,
2
) is an ordered pair. Finally suppose k
1
+ (y
1
,
1
) = k
2
+ (y
2
,
2
). Under these
assumptions we can prove that k
1
= k
2
, y
1
= y
2
,
1
=
2
.
Proof.
We dene h = k
1
+ (y
1
,
1
) = k
2
+ (y
2
,
2
). Since h = k
1
+ (y
1
,
1
) we can have two
possibilities:
D
1
= , D
1
= 1 and there exist two functions x
1
and
1
whose domain is D
1
such that h = (x
1
,
1
) ;
there exists a positive integer m
1
such that D
1
= 1, . . . , m
1
, D
1
= 1, . . . , m
1
+1
and there exist two functions x
1
and
1
whose domain is D
1
such that h = (x
1
,
1
).
Similarly, since h = k
2
+ (y
2
,
2
) we can have two possibilities:
D
2
= , D
2
= 1 and there exist two functions x
2
and
2
whose domain is D
2
such that h = (x
2
,
2
) ;
there exists a positive integer m
2
such that D
2
= 1, . . . , m
2
, D
2
= 1, . . . , m
2
+1
and there exist two functions x
2
and
2
whose domain is D
2
such that h = (x
2
,
2
).
It follows that (x
1
,
1
) = h = (x
2
,
2
), so x
1
= x
2
and
1
=
2
, and D
1
= D
2
.
Suppose D
1
= . This implies that D
2
= D
1
= 1, thus D
2
= .
In this case k
1
= = k
2
, y
1
= x
1
(1) = x
2
(1) = y
2
,
1
=
1
(1) =
2
(1) =
2
.
Suppose there exists a positive integer m
1
such that D
1
= 1, . . . , m
1
. This implies
that D
2
= D
1
= 1, . . . , m
1
+ 1, thus D
2
= 1, . . . , m
1
.
In this case for each = 1 . . . m
1
x
1
() = x
1
() = x
2
() = x
2
(),
1
() =
1
() =
2
() =
2
() . So k
1
= (x
1
,
1
) = (x
2
,
2
) = k
2
; and moreover y
1
= x
1
(m
1
+ 1) =
x
2
(m
1
+ 1) = y
2
,
1
=
1
(m
1
+ 1) =
2
(m
1
+ 1) =
2
.
14 M. Avon
Other useful results are the following.
Lemma 2.2. Suppose h = (x, ), k = (z, ) are state-like pairs such that h !(k)
and for each i, j dom(k) i ,= j z
i
,= z
j
. Then, for each i dom(k), j dom(h)
z
i
= x
j
i
=
j
.
Proof. Let i dom(k), j dom(h) and z
i
= x
j
. Clearly j dom(k), x
j
= z
j
, thus
z
i
= z
j
, i = j,
j
=
j
=
i
.
Lemma 2.3. Suppose h = (x, ) is a state-like pair, (y, ) is an ordered pair and dene
k = h + (y, ). Suppose g !(k) is such that g ,= k. Then g !(h).
Proof.
Let D = dom(h).
Suppose m is a positive integer and D = 1, . . . , m. Then k = (x
) has a domain
1, . . . , m+1. Moreover there exists C T such that C 1, . . . , m+1 and g = k
/C
.
Since g ,= k we must have C 1, . . . , m. We have
g = k
/C
= (x
/C
,
/C
) = ((x
/D
)
/C
, (
/D
)
/C
) = (x
/C
,
/C
) = h
/C
.
Now suppose D = . Then k = (x
(k)
#(k, t, ).
For each k K(n) we dene M(n, k) =
tE
s
(n,k)
M(k, t) .
We nally dene M(n) =
kK(n)
M(n, k).
We have seen that some entities may have been dened before step n, and in this case
we are not to dene them at step n, however at step n we need to check the denition
that has been given is consistent with what we would expect.
We are now are ready to begin the actual denition process, so we perform the simple
initial step of our inductive process.
We dene K(1) = , () = , E(1, ) = (.
For each t E(1, ) we dene #(, t, ) = #(t), V
b
(t) = , V
f
(t) = .
The inductive step is much more complex. Suppose all our denitions have been given
at step n and lets proceed with step n + 1. In this inductive step well need several as-
sumptions which will be identied with a title like Assumption 2.1.x. Each assumption
is a statement that must be valid at step 1, we suppose is valid at step n and needs to
be proved true at step n + 1 at the end of our denition process.
The rst assumption we need is the following.
Assumption 2.1.1. For each k K(n) such that k ,= and for each (k) there exist
a positive integer m, a function x: 1, . . . , m 1, a function : 1, . . . , m E(n), a
function s: 1, . . . , m M(n) such that
for each i, j 1 . . . m (i ,= j x
i
,= x
j
)
k = (x, )
= (x, s)
A dierent approach to logic 17
This assumption ensures that for each k K(n) such that k ,= k is a state-like pair,
and for each (k) is a state-like pair.
Given k = (x, ) K(n) we dene var(k) as the image of the function x. In other
words if k = then x = , so var(k) = , otherwise x has a domain 1, . . . , m and
var(k) = x
i
[i = 1 . . . m.
We can go on with the inductive step and dene
K(n)
+
= h + (y, )[ h K(n), E
s
(n, h), y (1 var(h)) ,
K(n + 1) = K(n) K(n)
+
.
Let k K(n)
+
. Then there exist h K(n), E
s
(n, h), y (1 var(h)) such that
k = h + (y, ). By lemma 2.1 we know that h, , y are univocally determined.
We can assume that (k) is dened for k K(n), and we need to dene this for
k K(n + 1) K(n). If k K(n + 1) K(n) then clearly k K(n)
+
and so there
exist h K(n), E
s
(n, h), y (1 var(h)) such that k = h + (y, ); and h, , y are
univocally determined. So we can dene
(k) = + (y, s)[ (h), s #(h, , ) .
We need to prove that this denition of (k) holds for all k K(n)
+
. To prove this
we need a second assumption.
Assumption 2.1.2. For each k K(n)
(k = )
((n > 1) g K(n 1), z 1 var(g), E
s
(n 1, g) :
k = g + (z, ) (k) = + (z, s)[ (g), s #(g, , ))
Thanks to this assumption we can prove the following lemma.
Lemma 2.1.3. For each k K(n)
+
, h K(n), E
s
(n, h), y (1 var(h)) such that
k = h + (y, ) we have
(k) = + (y, s)[ (h), s #(h, , ) .
Proof. If k / K(n) this is true by denition. If k K(n) we can apply the former lemma.
Since k ,= we have n > 1 and there exist g K(n 1), z 1 var(g), E
s
(n 1, g)
such that k = g + (z, ) (k) = + (z, s)[ (g), s #(g, , ) .
18 M. Avon
Since k = h + (y, ) we have g = h, z = y, = , and therefore
(k) = + (y, s)[ (h), s #(h, , ) .
Another consequence of lemma 2.1 is the following: for each k K(n)
+
and +(y, s)
in (k), , y and s are univocally determined.
To ensure the unique readability of our expressions we need the following assumption
(which is clearly satised for n = 1).
Assumption 2.1.4. For each t E(n)
t[(t)] ,= ( ;
if t[(t)] = ) then d(t, (t)) = 1, else d(t, (t)) = 0 ;
for each 1, . . . , (t) if (t[] = :) (t[] = ,) (t[] = )) then d(t, ) 1.
It is time to dene E(n +1, k), for each k in K(n +1). Then for each t in E(n +1, k)
and in (k) we need to dene #(k, t, ), and also we need to dene V
b
(t) and V
f
(t).
We have to warn that the denition of #(k, t, ), V
b
(t) and V
f
(t) is not an easy matter.
In fact, E(n+1, k) will be dened as the union of dierent sets. Consider for instance
the situation where k K(n). One of these sets is E(n, k), the old set of expressions
bound to context k. But of course there are also new sets. If an expression t belongs
just to E(n, k), and not to the new sets, then we dont need to reason about #(k, t, ),
because simply it has already been dened.
However, if t belongs both to E(n, k) and to one or more of the new sets, well have a
proposed denition of #(k, t, ) for each of the new sets, and well have to check that
this proposed denition is equal to the real denition.
If t belongs to just one new set and not to E(n, k) then well simply dene #(k, t, ) with
the proposed denition of #(k, t, ) for the new set.
If t belongs to more than one new set, and not to E(n, k), well need to check that the
proposed denitions of #(k, t, ) for each new set are equal to each other, and then well
be authorized to set #(k, t, ) with one of these proposed denitions.
When k / K(n) the discussion is simpler: it cannot be t E(n, k), so we just have to
consider the other situations. For the denition of V
b
(t) and V
f
(t) the reasoning is similar
but slightly dierent.
At this point we can proceed to formally dene the new sets of expressions bound to
context k, and for expressions in each of them we dene the proposed values of #(k, t, ),
A dierent approach to logic 19
V
b
(t) and V
f
(t).
For each k = h + (y, ) K(n)
+
we dene
E
a
(n + 1, k) = t[t E(n, h) y / V
b
(t).
For each t E
a
(n + 1, k), = + (y, s) (k) we dene the proposed values of
#(k, t, ), V
b
(t) and V
f
(t):
#(k, t, )
(n+1,k,a)
= #(h, t, );
V
f
(t)
(n+1,k,a)
= V
f
(t); V
b
(t)
(n+1,k,a)
= V
b
(t).
For each k = h + (y, ) K(n)
+
we dene
E
b
(n + 1, k) = y.
For each t E
b
(n + 1, k), = + (y, s) (k) we dene:
#(k, t, )
(n+1,k,b)
= s;
V
f
(t)
(n+1,k,b)
= y; V
b
(t)
(n+1,k,b)
= .
As a premise to the following denition of E
c
(n + 1, k), we specify that, given a pos-
itive integer m and a set D, we call D
m
the set D D where D appears m times
(when m = 1 of course D
1
= D), and a function whose domain is a subset of D
m
is called
a function with m arguments.
For each k K(n) we dene E
c
(n + 1, k) as the set of the strings ()(
1
, . . . ,
m
)
such that:
m is a positive integer;
,
1
, . . . ,
m
E(n, k);
for each (k) #(k, , ) is a function with m arguments and
(#(k,
1
, ), . . . , #(k,
m
, )) is a member of its domain.
This means that for each t E
c
(n + 1, k) there exist a positive integer m and
,
1
, . . . ,
m
E(n) such that t = ()(
1
, . . . ,
m
). In the following lemma well show
that m, ,
1
, . . . ,
m
are uniquely determined. Within this complex denition this proof
of unique readability may be considered as a technical detail, and can be skipped at rst
reading. The proof will often exploit lemma 2.6 and assumption 2.1.4, they will not be
quoted each time they are used.
Lemma 2.1.5. Let t E
c
(n + 1, k) and suppose
there exist a positive integer m and ,
1
, . . . ,
m
E(n): t = ()(
1
, . . . ,
m
).
20 M. Avon
there exist a positive integer p and ,
1
, . . . ,
p
E(n): t = ()(
1
, . . . ,
p
).
Then p = m, = and for each i 1, . . . , m
i
=
i
.
Proof.
If we know m we can provide an explicit representation of t. In fact if m = 2 then
t = ()(
1
,
2
), if m = 3 then t = ()(
1
,
2
,
3
) and so on. In this explicit representation
we can see explicit occurrences of the symbols , and ). There are explicit occurrences
of , only when m > 1. We indicate with q the position of the rst explicit occurrence
of ), and the second explicit occurrence of ) is clearly in position (t). If m > 1 we
indicate with q
1
, . . . , q
m1
the positions of the explicit occurrences of ,.
In the same way, if we know p we can provide another explicit representation of t.
In fact if p = 2 then t = ()(
1
,
2
), if p = 3 then t = ()(
1
,
2
,
3
) and so on. In this
explicit representation we can see explicit occurrences of the symbols , and ). There
are explicit occurrences of , only when p > 1. We indicate with r the position of the rst
explicit occurrence of ), and the second explicit occurrence of ) is clearly in position
(t). If p > 1 we indicate with r
1
, . . . , r
p1
the positions of the explicit occurrences of ,.
We have d(t, q 1) = d(t, 1 +()) = d(t, 1 + 1) +d(, ()) = 1 +d(, ()).
If t[q 1] = [()] = ) then d(t, q) = d(t, q 1) 1 = d(, ()) = 1.
Else t[q 1] = [()] / (, ), so d(t, q) = d(t, q 1) = 1 +d(, ()) = 1.
Suppose q < r. Obviously q > 1, q 1 1, q 1 r 2 = (); [q 1] = t[q] = ).
So d(t, q) = d(t, 1 + (q 1)) = d(t, 2) +d(, q 1) = 1 +d(, q 1) 2.
This is a contradiction, because we have proved d(t, q) = 1. Thus q r.
In the same way we can prove that r q, so we have r = q.
Clearly () = r2 = q 2 = (), and for each = 1 . . . () [] = t[+1] = [].
In other words = .
Of course we have also d(t, r) = d(t, q) = 1, d(t, r + 2) = d(t, r) 1 + 1 = 1,
d(t, q + 2) = d(t, q) 1 + 1 = 1.
We still need to show that p = m and for each i 1, . . . , m
i
=
i
.
First we examine the case where m = 1. We want to show that p = 1.
Suppose p > 1. In this situation we have
d(t, r
1
1) = d(t, r + 1 + (r
1
1 (r + 1))) = d(t, r + 1 +(
1
)) =
= d(t, r + 2) +d(
1
, (
1
)) = 1 +d(
1
, (
1
)).
If t[r
1
1] =
1
[(
1
)] = ) then d(t, r
1
) = d(t, r
1
1) 1 = d(
1
, (
1
)) = 1.
Else t[r
1
1] =
1
[(
1
)] / (, ) so d(t, r
1
) = d(t, r
1
1) = 1 +d(
1
, (
1
)) = 1.
A dierent approach to logic 21
Moreover we have to consider that
(
1
) = (t) 1 (q + 1) = (t) q 2,
r
1
(t) 1,
r
1
(q + 1) (t) 1 (q + 1) = (t) q 2 = (
1
),
r
1
q + 2,
r
1
(q + 1) 1,
1
[r
1
(q + 1)] = t[r
1
] = ,,
1 = d(t, r
1
) = d(t, q + 2) +d(
1
, r
1
(q + 1)) = 1 +d(
1
, r
1
(q + 1)).
This causes d(
1
, r
1
(q + 1)) = 0, but by assumption 2.1.4 we must have
d(
1
, r
1
(q + 1)) 1. So it must be p = 1.
Of course
(
1
) = (t) 1 (r + 1) = (t) r 2 = (t) q 2 = (
1
).
For each = 1 . . . (
1
)
1
[] = t[q +1+] = t[r +1+] =
1
[]. Therefore
1
=
1
.
Now lets discuss the case where m > 1.
First we want to prove that for each i = 1 . . . m1 p > i, d(t, q
i
) = 1, r
i
= q
i
,
i
=
i
.
Lets show that p > 1, d(t, q
1
) = 1, r
1
= q
1
,
1
=
1
.
If p = 1 of course m = 1, so p > 1 holds.
We have that
d(t, q
1
1) = d(t, q + 1 +(
1
)) = d(t, q + 1 + 1) +d(
1
, (
1
)) = 1 +d(
1
, (
1
)).
If t[q
1
1] =
1
[(
1
)] = ) then d(t, q
1
) = d(t, q
1
1) 1 = d(
1
, (
1
)) = 1 .
Else t[q
1
1] =
1
[(
1
)] / (, ) so d(t, q
1
) = d(t, q
1
1) = 1 +d(
1
, (
1
)) = 1.
Suppose q
1
< r
1
, we have
(
1
) = r
1
1 (r + 1) = r
1
r 2,
q
1
r 1 < r
1
r 1,
q
1
r 1 (
1
),
q
1
> q + 1,
q
1
> r + 1,
q
1
r 1 1,
and then
1 = d(t, q
1
) = d(t, r + 1 + (q
1
r 1)) = d(t, r + 2) +d(
1
, q
1
r 1) =
= 1 +d(
1
, q
1
r 1).
So d(
1
, q
1
r 1) = 0. But since
1
[q
1
r 1] = t[q
1
] = ,, by assumption 2.1.4 we
must have d(
1
, q
1
r 1) 1, so we have a contradiction.
22 M. Avon
Hence q
1
r
1
and in the same way we can show that r
1
q
1
, therefore r
1
= q
1
.
At this point we observe that
(
1
) = q
1
1 (q + 1) = q
1
q 2 = r
1
r 2 = (
1
).
Moreover, for each = 1 . . . (
1
)
1
[] = t[q + 1 +] = t[r + 1 +] =
1
[].
Therefore
1
=
1
.
We have proved that p > 1, d(t, q
1
) = 1, r
1
= q
1
,
1
=
1
, and if m = 2 we have also
shown that for each i = 1 . . . m1 p > i, d(t, q
i
) = 1, r
i
= q
i
,
i
=
i
.
Now suppose m > 2, let i = 1 . . . m 2, suppose we have proved p > i, d(t, q
i
) = 1,
r
i
= q
i
,
i
=
i
, we want to show that p > i +1, d(t, q
i+1
) = 1, r
i+1
= q
i+1
,
i+1
=
i+1
.
First of all
d(t, q
i+1
1) = d(t, q
i
+(
i+1
)) = d(t, q
i
+ 1) +d(
i+1
, (
i+1
)) =
= 1 +d(
i+1
, (
i+1
)).
If t[q
i+1
1] =
i+1
[(
i+1
)] = ) then
d(t, q
i+1
) = d(t, q
i+1
1) 1 = d(
i+1
, (
i+1
)) = 1.
Else t[q
i+1
1] =
i+1
[(
i+1
)] / (, ) so
d(t, q
i+1
) = d(t, q
i+1
1) = 1 +d(
i+1
, (
i+1
)) = 1.
Suppose p = i + 1. We have i m2, i + 2 m, t[q
i+1
] = ,. And we have also
(
p
) = (t) 1 r
i
,
q
i+1
(t) 1,
q
i+1
r
i
(t) 1 r
i
= (
p
),
q
i+1
r
i
= q
i+1
q
i
1,
p
[q
i+1
r
i
] = t[q
i+1
] = ,,
and
1 = d(t, q
i+1
) = d(t, r
i
+ (q
i+1
r
i
)) = d(t, r
i
+ 1) +d(
p
, q
i+1
r
i
) =
= 1 +d(
p
, q
i+1
r
i
).
So d(
p
, q
i+1
r
i
) = 0 and this contradicts assumption 2.1.4. Therefore p > i + 1.
Now suppose q
i+1
< r
i+1
. In this case
(
i+1
) = r
i+1
1 r
i
,
q
i+1
r
i+1
1,
q
i+1
r
i
r
i+1
1 r
i
= (
i+1
),
q
i+1
r
i
= q
i+1
q
i
1,
i+1
[q
i+1
r
i
] = t[q
i+1
] = ,,
A dierent approach to logic 23
and
1 = d(t, q
i+1
) = d(t, r
i
+ (q
i+1
r
i
)) = d(t, r
i
+ 1) +d(
i+1
, q
i+1
r
i
) =
= 1 +d(
i+1
, q
i+1
r
i
).
So d(
i+1
, q
i+1
r
i
) = 0 and this contradicts assumption 2.1.4. Therefore q
i+1
r
i+1
.
In the same way we can prove that q
i+1
r
i+1
, hence r
i+1
= q
i+1
is proved.
Moreover
(
i+1
) = q
i+1
1 q
i
= r
i+1
1 r
i
= (
i+1
),
and for each = 1 . . . (
i+1
)
i+1
[] = t[r
i
+] = t[q
i
+] =
i+1
[].
We have proved that for each i = 1 . . . m1 p > i, d(t, q
i
) = 1, r
i
= q
i
,
i
=
i
.
So p m, and in the same way we could prove m p, therefore p = m.
We have seen that r
m1
= q
m1
, it follows
(
m
) = (t) 1 q
m1
= (t) 1 r
m1
= (
m
),
and for each = 1 . . . (
m
)
m
[] = t[r
m1
+] = t[q
m1
+] =
m
[],
therefore
m
=
m
.
So also in the case m > 1 it is shown that p = m and for each i = 1 . . . m
i
=
i
.
For each t = ()(
1
, . . . ,
m
) E
c
(n + 1, k) we dene
#(k, t, )
(n+1,k,c)
= #(k, , )(#(k,
1
, ), . . . , #(k,
m
, )),
V
f
(t)
(n+1,k,c)
= V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t)
(n+1,k,c)
= V
b
() V
b
(
1
) V
b
(
m
).
For each k K(n) we dene E
d
(n + 1, k) as the set of the strings (f)(
1
, . . . ,
m
)
such that:
f belongs to T
m is a positive integer;
1
, . . . ,
m
E(n, k);
for each (k) A
f
(#(k,
1
, ), . . . , #(k,
m
, )) is true.
For instance, this means that if the logical conjunction symbol belongs to T,
1
,
2
belong to E(n, k) and for each (k) both #(k,
1
, ) and #(k,
2
, ) are true or
false, then ()(
1
,
2
) belongs to E
d
(n + 1, k).
24 M. Avon
This implies that for each t E
d
(n +1, k) there are f in T, a positive integer m and
1
, . . . ,
m
E(n) such that t = (f)(t
1
, . . . , t
m
). We will now show that f, m,
1
, . . . ,
m
are uniquely determined. Within this complex denition this proof of unique readability
may be considered as a technical detail, and can be skipped at rst reading. The proof
will often exploit lemma 2.6 and assumption 2.1.4, they will not be quoted each time
they are used.
Lemma 2.1.6. Let t E
d
(n + 1, k) and suppose
there exist f T, a positive integer m and
1
, . . . ,
m
E(n): t = (f)(
1
, . . . ,
m
).
there exist g T, a positive integer p and
1
, . . . ,
p
E(n): t = (g)(
1
, . . . ,
p
).
Then g = f, p = m and for each i 1, . . . , m
i
=
i
.
Proof.
If we know m we can provide an explicit representation of t. In fact if m = 2 then
t = (f)(
1
,
2
), if m = 3 then t = (f)(
1
,
2
,
3
) and so on. In this explicit representation
we can see explicit occurrences of the symbols , and ). There are explicit occurrences
of , only when m > 1. The explicit occurrences of ) are clearly in positions 3 and (t).
If m > 1 we indicate with q
1
, . . . , q
m1
the positions of the explicit occurrences of ,.
In the same way, if we know p we can provide another explicit representation of t.
In fact if p = 2 then t = (g)(
1
,
2
), if p = 3 then t = (g)(
1
,
2
,
3
) and so on. In this
explicit representation we can see explicit occurrences of the symbols , and ). There
are explicit occurrences of , only when p > 1. The explicit occurrences of ) are clearly
in positions 3 and (t). If p > 1 we indicate with r
1
, . . . , r
p1
the positions of the explicit
occurrences of ,.
It is immediate to see that g = t[2] = f.
We still need to show that p = m and for each i 1, . . . , m
i
=
i
.
First we examine the case where m = 1. We want to show that p = 1.
Suppose p > 1. In this situation we have
d(t, r
1
1) = d(t, 4 + (r
1
1 4)) = d(t, 4 +(
1
)) =
= d(t, 4 + 1) +d(
1
, (
1
)) = 1 +d(
1
, (
1
)).
If t[r
1
1] =
1
[(
1
)] = ) then d(t, r
1
) = d(t, r
1
1) 1 = d(
1
, (
1
)) = 1.
Else t[r
1
1] =
1
[(
1
)] / (, ) so d(t, r
1
) = d(t, r
1
1) = 1 +d(
1
, (
1
)) = 1.
A dierent approach to logic 25
Moreover we have to consider that
(
1
) = (t) 1 4 = (t) 5,
r
1
(t) 1,
r
1
4 (t) 1 4 = (t) 5 = (
1
),
r
1
4 + 1,
r
1
4 1,
1
[r
1
4] = t[r
1
] = ,,
1 = d(t, r
1
) = d(t, 4 + 1) +d(
1
, r
1
4) = 1 +d(
1
, r
1
4).
This causes d(
1
, r
1
4) = 0, but by assumption 2.1.4 we must have
d(
1
, r
1
4) 1. So it must be p = 1.
Of course
(
1
) = (t) 1 4 = (
1
).
For each = 1 . . . (
1
)
1
[] = t[4 +] =
1
[]. Therefore
1
=
1
.
Now lets discuss the case where m > 1.
First we want to prove that for each i = 1 . . . m1 p > i, d(t, q
i
) = 1, r
i
= q
i
,
i
=
i
.
Lets show that p > 1, d(t, q
1
) = 1, r
1
= q
1
,
1
=
1
.
If p = 1 of course m = 1, so p > 1 holds.
We have that
d(t, q
1
1) = d(t, 4 +(
1
)) = d(t, 4 + 1) +d(
1
, (
1
)) = 1 +d(
1
, (
1
)).
If t[q
1
1] =
1
[(
1
)] = ) then d(t, q
1
) = d(t, q
1
1) 1 = d(
1
, (
1
)) = 1 .
Else t[q
1
1] =
1
[(
1
)] / (, ) so d(t, q
1
) = d(t, q
1
1) = 1 +d(
1
, (
1
)) = 1.
Suppose q
1
< r
1
, we have
(
1
) = r
1
1 4 = r
1
5,
q
1
4 < r
1
4,
q
1
4 (
1
),
q
1
> 4,
q
1
4 1,
and then
1 = d(t, q
1
) = d(t, 4 + (q
1
4)) = d(t, 4 + 1) +d(
1
, q
1
4) =
= 1 +d(
1
, q
1
4).
So d(
1
, q
1
4) = 0. But since
1
[q
1
4] = t[q
1
] = ,, by assumption 2.1.4 we must
have d(
1
, q
1
4) 1, so we have a contradiction.
Hence q
1
r
1
and in the same way we can show that r
1
q
1
, therefore r
1
= q
1
.
26 M. Avon
At this point we observe that (
1
) = q
1
1 4 = r
1
1 4 = (
1
).
Moreover, for each = 1 . . . (
1
)
1
[] = t[4 +] =
1
[].
Therefore
1
=
1
.
We have proved that p > 1, d(t, q
1
) = 1, r
1
= q
1
,
1
=
1
, and if m = 2 we have also
shown that for each i = 1 . . . m1 p > i, d(t, q
i
) = 1, r
i
= q
i
,
i
=
i
.
Now suppose m > 2, let i = 1 . . . m 2, suppose we have proved p > i, d(t, q
i
) = 1,
r
i
= q
i
,
i
=
i
, we want to show that p > i +1, d(t, q
i+1
) = 1, r
i+1
= q
i+1
,
i+1
=
i+1
.
First of all
d(t, q
i+1
1) = d(t, q
i
+(
i+1
)) = d(t, q
i
+ 1) +d(
i+1
, (
i+1
)) =
= 1 +d(
i+1
, (
i+1
)).
If t[q
i+1
1] =
i+1
[(
i+1
)] = ) then
d(t, q
i+1
) = d(t, q
i+1
1) 1 = d(
i+1
, (
i+1
)) = 1.
Else t[q
i+1
1] =
i+1
[(
i+1
)] / (, ) so
d(t, q
i+1
) = d(t, q
i+1
1) = 1 +d(
i+1
, (
i+1
)) = 1.
Suppose p = i + 1. We have i m2, i + 2 m, t[q
i+1
] = ,. And we have also
(
p
) = (t) 1 r
i
,
q
i+1
(t) 1,
q
i+1
r
i
(t) 1 r
i
= (
p
),
q
i+1
r
i
= q
i+1
q
i
1,
p
[q
i+1
r
i
] = t[q
i+1
] = ,,
and
1 = d(t, q
i+1
) = d(t, r
i
+ (q
i+1
r
i
)) = d(t, r
i
+ 1) +d(
p
, q
i+1
r
i
) =
= 1 +d(
p
, q
i+1
r
i
).
So d(
p
, q
i+1
r
i
) = 0 and this contradicts assumption 2.1.4. Therefore p > i + 1.
Now suppose q
i+1
< r
i+1
. In this case
(
i+1
) = r
i+1
1 r
i
,
q
i+1
r
i+1
1,
q
i+1
r
i
r
i+1
1 r
i
= (
i+1
),
q
i+1
r
i
= q
i+1
q
i
1,
i+1
[q
i+1
r
i
] = t[q
i+1
] = ,,
and
1 = d(t, q
i+1
) = d(t, r
i
+ (q
i+1
r
i
)) = d(t, r
i
+ 1) +d(
i+1
, q
i+1
r
i
) =
= 1 +d(
i+1
, q
i+1
r
i
).
A dierent approach to logic 27
So d(
i+1
, q
i+1
r
i
) = 0 and this contradicts assumption 2.1.4. Therefore q
i+1
r
i+1
.
In the same way we can prove that q
i+1
r
i+1
, hence r
i+1
= q
i+1
is proved.
Moreover
(
i+1
) = q
i+1
1 q
i
= r
i+1
1 r
i
= (
i+1
),
and for each = 1 . . . (
i+1
)
i+1
[] = t[r
i
+] = t[q
i
+] =
i+1
[].
We have proved that for each i = 1 . . . m1 p > i, d(t, q
i
) = 1, r
i
= q
i
,
i
=
i
.
So p m, and in the same way we could prove m p, therefore p = m.
We have seen that r
m1
= q
m1
, it follows
(
m
) = (t) 1 q
m1
= (t) 1 r
m1
= (
m
),
and for each = 1 . . . (
m
)
m
[] = t[r
m1
+] = t[q
m1
+] =
m
[],
therefore
m
=
m
.
So also in the case m > 1 it is shown that p = m and for each i = 1 . . . m
i
=
i
.
For each t = (f)(
1
, . . . ,
m
) E
d
(n + 1, k) we dene
#(k, t, )
(n+1,k,d)
= P
f
(#(k,
1
, ), . . . , #(k,
m
, )),
V
f
(t)
(n+1,k,d)
= V
f
(
1
) V
f
(
m
),
V
b
(t)
(n+1,k,d)
= V
b
(
1
) V
b
(
m
).
Let k K(n), m a positive integer, x a function whose domain is 1, . . . , m such
that for each i = 1 . . . m x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n), and
nally let E(n). We write
c(n, k, m, x, , )
to indicate the following condition (where k
1
= k + (x
1
,
1
), and if m > 1 for each
i = 1 . . . m1 k
i+1
= k
i
+ (x
i+1
,
i+1
)):
1
E
s
(n, k) ;
if m > 1 then for each i = 1 . . . m1 k
i
K(n)
i+1
E
s
(n, k
i
);
k
m
K(n) E(n, k
m
).
For each k K(n) we dene E
e
(n + 1, k) as the set of the strings
(x
1
:
1
, . . . , x
m
:
m
, )
such that:
m is a positive integer;
28 M. Avon
x is a function whose domain is 1, . . . , m such that for each i = 1 . . . m x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
;
is a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n);
E(n);
c(n, k, m, x, , ).
This implies that for each t E
e
(n+1, k) there exist a positive integer m, a function
x whose domain is 1, . . . , m such that for each i = 1 . . . m x
i
1, a function whose
domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n), and E(n) such that
t = (x
1
:
1
, . . . , x
m
:
m
, ). We will now show that m, x, , are uniquely deter-
mined. Within this complex denition this proof of unique readability may be considered
as a technical detail, and can be skipped at rst reading. The proof will often exploit
lemma 2.6 and assumption 2.1.4, they will not be quoted each time they are used.
Lemma 2.1.7. Let t E
e
(n + 1, k) and suppose
there exist a positive integer m, a function x whose domain is 1, . . . , m such that
for each i = 1 . . . m x
i
1, a function whose domain is 1, . . . , m such that for
each i = 1 . . . m
i
E(n), and E(n) such that t = (x
1
:
1
, . . . , x
m
:
m
, );
there exist a positive integer p, a function y whose domain is 1, . . . , p such that
for each i = 1 . . . p y
i
1, a function whose domain is 1, . . . , p such that for
each i = 1 . . . p
i
E(n), and E(n) such that t = (y
1
:
1
, . . . , y
p
:
p
, );
Then p = m, y = x, = and = .
Proof.
If we know m we can provide an explicit representation of t. In fact if m = 2 then
t = (x
1
:
1
, x
2
:
2
, ), if m = 3 then t = (x
1
:
1
, x
2
:
2
, x
3
:
3
, ), and so on.
In this explicit representation of t we can see explict occurrences of the symbols , and
:. We indicate with q
1
, . . . , q
m
the positions of the explicit occurrences of : and with
r
1
. . . r
m
the positions of the explicit occurrences of ,.
In the same way, if we know p we can provide another explicit representation of t. In
fact if p = 2 then t = (y
1
:
1
, y
2
:
2
, ), if p = 3 then t = (y
1
:
1
, y
2
:
2
, y
3
:
3
, ),
and so on. In this explicit representation of t we can see explict occurrences of the symbols
, and :. We indicate with q
1
, . . . , q
m
the positions of the explicit occurrences of : and
with r
1
. . . r
m
the positions of the explicit occurrences of ,.
We want to show that for each i = 1 . . . m
p i, y
i
= x
i
, q
i
= q
i
, d(t, r
i
) = 1, r
i
= r
i
,
i
=
i
.
The rst step is to show that y
1
= x
1
, q
1
= q
1
, d(t, r
1
) = 1, r
1
= r
1
,
1
=
1
.
A dierent approach to logic 29
Of course y
1
= t[3] = x
1
, q
1
= 4 = q
1
. Moreover
d(t, r
1
1) = d(t, q
1
+ (r
1
1 q
1
)) = d(t, q
1
+(
1
)) = d(t, q
1
+ 1) +d(
1
, (
1
)) =
= 1 +d(
1
, (
1
)).
If t[r
1
1] =
1
[(
1
)] = ) then d(t, r
1
) = d(t, r
1
1) 1 = d(
1
, (
1
)) = 1.
Else t[r
1
1] =
1
[(
1
)] / (, ) so d(t, r
1
) = d(t, r
1
1) = 1 +d(
1
, (
1
)) = 1.
Now suppose r
1
< r
1
. This would mean that
(
1
) = r
1
1 q
1
,
r
1
q
1
r
1
1 q
1
= (
1
),
r
1
q
1
= r
1
q
1
1,
1
[r
1
q
1
] = t[r
1
] = ,,
and
1 = d(t, r
1
) = d(t, q
1
+ (r
1
q
1
)) = d(t, q
1
+ 1) +d(
1
, r
1
q
1
) = 1 +d(
1
, r
1
q
1
).
So d(
1
, r
1
q
1
) = 0 and this contradicts assumption 2.1.4. Hence r
1
r
1
, and in the
same way we can show that r
1
r
1
, therefore r
1
= r
1
.
At this point we observe that (
1
) = r
1
1 q
1
= (
1
).
Moreover, for each = 1 . . . (
1
)
1
[] = t[q
1
+] = t[q
1
+] =
1
[], hence
1
=
1
.
We have proved that y
1
= x
1
, q
1
= q
1
, d(t, r
1
) = 1, r
1
= r
1
,
1
=
1
. As a consequence,
if m = 1 we have proved that for each i = 1 . . . m
p i, y
i
= x
i
, q
i
= q
i
, d(t, r
i
) = 1, r
i
= r
i
,
i
=
i
.
Consider the case where m > 1. Let i = 1 . . . m1, we suppose
p i, y
i
= x
i
, q
i
= q
i
, d(t, r
i
) = 1, r
i
= r
i
,
i
=
i
,
and want to show that
p i + 1, y
i+1
= x
i+1
, q
i+1
= q
i+1
, d(t, r
i+1
) = 1, r
i+1
= r
i+1
,
i+1
=
i+1
.
We can immediately show that d(t, r
i+1
) = 1. In fact d(t, q
i+1
+ 1) = d(t, r
i
) = 1,
d(t, r
i+1
1) = d(t, q
i+1
+ (r
i+1
1 q
i+1
)) = d(t, q
i+1
+(
i+1
)) =
= d(t, q
i+1
+ 1) +d(
i+1
, (
i+1
)) = 1 +d(
i+1
, (
i+1
)).
If t[r
i+1
1] =
i+1
[(
i+1
)] = ) then
d(t, r
i+1
) = d(t, r
i+1
1) 1 = d(
i+1
, (
i+1
)) = 1.
Else t[r
i+1
1] =
i+1
[(
i+1
)] / (, ) so
d(t, r
i+1
) = d(t, r
i+1
1) = 1 +d(
i+1
, (
i+1
)) = 1.
Suppose p = i. In this case
30 M. Avon
() = (t) 1 r
i
,
r
i+1
r
i
(t) 1 r
i
= (),
r
i+1
r
i
= r
i+1
r
i
1,
[r
i+1
r
i
] = t[r
i+1
] = ,,
and
1 = d(t, r
i+1
) = d(t, r
i
+ (r
i+1
r
i
)) = d(t, r
i
+ 1) +d(, r
i+1
r
i
) =
= 1 +d(, r
i+1
r
i
).
So d(, r
i+1
r
i
) = 0, and this contradicts assumption 2.1.4. Therefore p i + 1.
It follows immediately that y
i+1
= t[r
i
+ 1] = t[r
i
+ 1] = x
i+1
and q
i+1
= r
i
+ 2 = q
i+1
.
Now we suppose r
i+1
< r
i+1
. This would mean that
(
i+1
) = r
i+1
1 q
i+1
,
r
i+1
q
i+1
r
i+1
1 q
i+1
= (
i+1
),
r
i+1
q
i+1
= r
i+1
q
i+1
1,
i+1
[r
i+1
q
i+1
] = t[r
i+1
] = ,,
and
1 = d(t, r
i+1
) = d(t, q
i+1
+ (r
i+1
q
i+1
)) = d(t, q
i+1
+ 1) +d(
i+1
, r
i+1
q
i+1
) =
= 1 +d(
i+1
, r
i+1
q
i+1
).
So d(
i+1
, r
i+1
q
i+1
) = 0 and this contradicts assumption 2.1.4. Hence r
i+1
r
i+1
.
In the same way we can show that r
i+1
r
i+1
, therefore r
i+1
= r
i+1
.
At this point we observe that (
i+1
) = r
i+1
1 q
i+1
= (
i+1
).
Furthermore, for each = 1 . . . (
i+1
)
i+1
[] = t[q
i+1
+ ] = t[q
i+1
+ ] =
i+1
[],
hence
i+1
=
i+1
.
It is shown that for each i = 1 . . . m
p i, y
i
= x
i
, q
i
= q
i
, d(t, r
i
) = 1, r
i
= r
i
,
i
=
i
.
So p m. In the same way we could prove that m p, so p = m. At this stage we
have shown that y = x and = , we just need a nal step, which is proving that = .
This clearly holds because of
() = (t) 1 r
p
= (t) 1 r
m
= (),
and for each = 1 . . . ()
[] = t[r
p
+] = t[r
m
+] = [].
A dierent approach to logic 31
For every t = (x
1
:
1
, . . . , x
m
:
m
, ) E
e
(n + 1, k) we dene
#(k, t, )
(n+1,k,e)
= #(k
m
, ,
m
)[
m
(k
m
), _
m
,
where k
1
= k +(x
1
,
1
), and if m > 1 for each i = 1 . . . m1 k
i+1
= k
i
+(x
i+1
,
i+1
).
Notice that the set #(k
m
, ,
m
)[
m
(k
m
), _
m
is specied using a standard
mathematical notation. We could specify it using a notation closer to the one of our for-
mulas, in this case it could have been written as (
m
(k
m
) : _
m
, #(k
m
, ,
m
)).
It might still be a bit unclear what is the intended meaning of the expression
(x
1
:
1
, . . . , x
m
:
m
, ).
This is the same meaning that the expression
[ x
1
1
, . . . , x
m
m
a
(n + 1, k) = E
a
(n + 1, k), E
b
(n + 1, k) = E
b
(n + 1, k),
else we dene E
a
(n + 1, k) = , E
b
(n + 1, k) = .
If k K(n) we dene E
c
(n + 1, k) = E
c
(n + 1, k), E
d
(n + 1, k) =
E
d
(n + 1, k), E
e
(n + 1, k) = E
e
(n + 1, k),
else E
(n, k) = , E
c
(n + 1, k) = , E
d
(n + 1, k) = , E
e
(n + 1, k) = .
Finally we dene
E(n+1, k) = E
(n, k)E
a
(n+1, k)E
b
(n+1, k)E
c
(n+1, k)E
d
(n+1, k)E
e
(n+1, k).
For every k K(n+1), t E(n+1, k) and (k) we need that #(k, t, ) is dened.
But we also need that the denition is such that for each k K(n+1), w a, b, c, d, e,
t E
w
(n + 1, k) and (k) #(k, t, )
(n+1,k,w)
= #(k, t, ).
32 M. Avon
Given k K(n + 1), t E(n + 1, k) and (k) there are three possibilities.
1. t is in E
(n, k): then #(k, t, ) is already dened; if t is in one or more of the sets
E
w
(n +1, k) then for each w we need to verify that #(k, t, ) = #(k, t, )
(n+1,k,w)
.
2. t is not in E
w
(n +1, k): then we just dene
#(k, t, ) = #(k, t, )
(n+1,k,w)
.
3. t is not in E
w
(n + 1, k): in this case
we choose w such that t E
w
(n + 1, k) and dene #(k, t, ) = #(k, t, )
(n+1,k, w)
.
We also need to verify that for each w such that t E
w
(n+1, k) #(k, t, )
(n+1,k,w)
=
#(k, t, )
(n+1,k, w)
.
By point 1. we are required to verify that for each k K(n + 1), w a, b, c, d, e,
t E
(n, k) E
w
(n + 1, k) and (k) #(k, t, ) = #(k, t, )
(n+1,k,w)
.
By point 3. we are required to verify that for each k K(n+1), w
1
, w
2
a, b, c, d, e :
w
1
,= w
2
, t E
w
1
(n + 1, k) E
w
2
(n + 1, k) and (k)
#(k, t, )
(n+1,k,w
1
)
= #(k, t, )
(n+1,k,w
2
)
.
Its easy to see that if these properties are veried then we can state that for each
k K(n+1), w a, b, c, d, e, t E
w
(n+1, k), (k) #(k, t, )
(n+1,k,w)
= #(k, t, ).
With respect to the denitions of V
b
(t) and V
f
(t) we can make a similar argument.
For every k K(n+1) and t E(n+1, k) we need that V
b
(t) and V
f
(t) are dened. But
we also need that the denition is such that for each k K(n +1), w a, b, c, d, e and
t E
w
(n + 1, k) V
b
(t)
(n+1,k,w)
= V
b
(t) and V
f
(t)
(n+1,k,w)
= V
f
(t).
Given t E(n + 1) there are three possibilities.
1. t is in E(n): then V
b
(t) and V
f
(t) are already dened; if t is in one or more of the
sets E
w
(n+1, k) then for each k and w we need to verify that V
b
(t) = V
b
(t)
(n+1,k,w)
and V
f
(t) = V
f
(t)
(n+1,k,w)
.
2. t is not in E(n) and t is in just one of the sets E
w
(n + 1, k): then we just dene
V
b
(t) = V
b
(t)
(n+1,k,w)
and V
f
(t) = V
f
(t)
(n+1,k,w)
.
3. t is not in E(n) and there are more than one k K(n + 1) and w a, b, c, d, e
such that t is in E
w
(n +1, k). In this case we arbitrarily choose
k and w such that
t is in E
w
(n + 1,
k) and dene V
b
(t) = V
b
(t)
(n+1,
k, w)
; V
f
(t) = V
f
(t)
(n+1,
k, w)
.
Here we need to verify that for each k, w such that t is in E
w
(n + 1, k)
V
b
(t)
(n+1,k,w)
= V
b
(t)
(n+1,
k, w)
, V
f
(t)
(n+1,k,w)
= V
f
(t)
(n+1,
k, w)
.
By point 1. we are required to verify that for each k K(n + 1), w a, b, c, d, e,
t E(n) E
w
(n + 1, k) V
b
(t) = V
b
(t)
(n+1,k,w)
and V
f
(t) = V
f
(t)
(n+1,k,w)
.
A dierent approach to logic 33
By point 3. we are required to verify that for each k
1
, k
2
K(n + 1), w
1
, w
2
a, b, c, d, e, t E
w
1
(n + 1, k
1
) E
w
2
(n + 1, k
2
) such that t / E(n) we have
V
b
(t)
(n+1,k
1
,w
1
)
= V
b
(t)
(n+1,k
2
,w
2
)
,
V
f
(t)
(n+1,k
1
,w
1
)
= V
f
(t)
(n+1,k
2
,w
2
)
.
Its easy to see that if these properties are veried then we can state that for each
k K(n + 1), w a, b, c, d, e and t E
w
(n + 1, k) V
b
(t)
(n+1,k,w)
= V
b
(t) and
V
f
(t)
(n+1,k,w)
= V
f
(t).
We now have to perform the required verications. These verications require a further
set of assumptions. Well list those assumptions, and also signicant consequences to them
and other results that will in turn be used in our verication process.
Assumption 2.1.8. if n > 1 then K(n 1) K(n).
Assumption 2.1.9. Let k = (x, ), h = (y, ) K(n) such that for each i dom(k),
j dom(h) x
i
= y
j
i
=
j
. Let t E(n, k) E(n, h). Let = (x, s) (k),
= (y, r) (h) such that for each i dom(), j dom() x
i
= y
j
s
i
= r
j
. Then
#(k, t, ) = #(h, t, ).
The next assumption has a central role in our verication process.
34 M. Avon
Assumption 2.1.10. For each k K(n), t E(n, k) one and only one of these ve
alternative situations is veried:
a.
t (, (k) #(k, t, ) = #(t), V
f
(t) = , V
b
(t) = .
b.
n > 1,
if we set k = (x, ) then i dom(k) : (t = x
i
, = (x, s) (k) #(k, t, ) = s
i
),
V
f
(t) = t, V
b
(t) = .
c.
n > 1,
h K(n 1) : h _ k, m positive integer , ,
1
, . . .
m
E(n 1, h) :
t = ()(
1
, . . . ,
m
), t E(n, h),
(h) ( #(h, , ) is a function with m arguments,
(#(h,
1
, ), . . . , #(h,
m
, )) is a member of the domain of #(h, , ),
#(h, t, ) = #(h, , )(#(h,
1
, ), . . . , #(h,
m
, )) ),
V
f
(t) = V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
() V
b
(
1
) V
b
(
m
),
(k), (h) : _ it results #(k, t, ) = #(h, t, ).
d.
n > 1,
h K(n 1) : h _ k, f T, m positive integer ,
1
, . . .
m
E(n 1, h) :
t = (f)(
1
, . . . ,
m
), t E(n, h),
(h) ( A
f
(#(h,
1
, ), . . . , #(h,
m
, )),
#(h, t, ) = P
f
(#(h,
1
, ), . . . , #(h,
m
, )) ),
V
f
(t) = V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
(
1
) V
b
(
m
),
(k), (h) : _ it results #(k, t, ) = #(h, t, ).
A dierent approach to logic 35
e.
n > 1,
there exist
h K(n 1) : h _ k,
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(h), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n 1),
E(n 1)
such that
c(n 1, h, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ), t E(n, h),
for each (h) #(h, t, ) = #(h
m
, ,
m
)[
m
(h
m
), _
(where h
1
= h + (x
1
,
1
), and if m > 1 for each i = 1 . . . m1
h
i+1
= h
i
+ (x
i+1
,
i+1
) ),
if m = 1 V
f
(t) = V
f
(
1
) (V
f
() x
1
), V
b
(t) = x
1
V
b
(
1
) V
b
(),
if m > 1
V
f
(t) = V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
),
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
(),
(k), (h) : _ it results #(k, t, ) = #(h, t, ).
Assumption 2.1.11. Let n > 1, k K(n), h !(k) : h ,= k. Then h K(n 1), for
each (k) if we dene =
/dom(h)
then (h).
Assumption 2.1.12. If n > 1 then for each g K(n 1) E(n 1, g) E(n, g).
Lemma 2.1.13. Suppose h, k K(n), y 1 var(h), E
s
(n, h), k = h + (y, ).
Moreover let (h), (k) such that _ . Then there is s #(h, , ) such that
= + (y, s).
Proof.
We can apply our assumption 2.1.2 and get
(n > 1) g K(n 1), z 1 var(g), E
s
(n 1, g) :
k = g + (z, ) (k) = + (z, s)[ (g), s #(g, , )
So h + (y, ) = k = g + (z, ) and by lemma 2.1 h = g, y = z, = .
36 M. Avon
Therefore (k) =
+ (y, s)[
(h), s #(h, ,
).
Hence there exist
(h), s #(h, ,
) such that =
+ (y, s).
Now dom() = dom(h) = dom(
!() we have
=
/dom()
=
/dom(
)
=
.
Therefore there is s #(h, , ) such that = + (y, s).
Lemma 2.1.14. Suppose h, k K(n), y 1 var(h), E
s
(n, h), k = h + (y, ).
Moreover let (h) and = +(y, s) with s #(h, , ). Then (k), and clearly
_ .
Proof.
We can apply our assumption 2.1.2 and get
(n > 1) g K(n 1), z 1 var(g), E
s
(n 1, g) :
k = g + (z, ) (k) = + (z, s)[ (g), s #(g, , )
So h + (y, ) = k = g + (z, ) and by lemma 2.1 h = g, y = z, = .
Therefore (k) =
+ (y, s)[
(h), s #(h, ,
).
It follows immediately that (k), and clearly _ .
Lemma 2.1.15. Let g = (y, ), h = (z, ) K(n); m a positive integer; x a func-
tion whose domain is 1, . . . , m such that for each i = 1 . . . m x
i
(1 var(g))
(1 var(h)), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
; a function whose do-
main is 1, . . . , m such that for each i = 1 . . . m
i
E(n); E(n). Let also
c(n, g, m, x, , ), c(n, h, m, x, , ).
Moreover we suppose that for each i dom(g), j dom(h), y
i
= z
j
i
=
j
. Let
also = (y, r) (g), = (z, u) (h) be such that for each i dom(), j dom(),
y
i
= z
j
r
i
= u
j
. If as usual we dene
g
1
= g + (x
1
,
1
), and if m > 1 for each i = 1 . . . m1 g
i+1
= g
i
+ (x
i+1
,
i+1
),
h
1
= h + (x
1
,
1
), and if m > 1 for each i = 1 . . . m1 h
i+1
= h
i
+ (x
i+1
,
i+1
),
then we have
#(h
m
, ,
m
)[
m
(h
m
), _
m
= #(g
m
, ,
m
)[
m
(g
m
), _
m
.
Proof.
By assumption 2.1.1 h = or there exists a positive integer p such that dom(h) =
1, . . . , p. In the case where h = we dene p = 0. At this point we can notice that for
each i = 1 . . . m dom(h
i
) = 1, . . . , p + i. In fact dom(h
1
) = 1, . . . , p + 1. If m > 1
we need an inductive step. Let i = 1 . . . m 1, suppose dom(h
i
) = 1, . . . , p + i. Then
dom(h
i+1
) = 1, . . . , p +i + 1.
Let u #(h
m
, ,
m
)[
m
(h
m
), _
m
, we want to show that
u #(g
m
, ,
m
)[
m
(g
m
), _
m
.
A dierent approach to logic 37
There exists
m
(h
m
) such that _
m
and u = #(h
m
, ,
m
).
If m > 1 then for each i = 1 . . . m1, since dom(
m
) = dom(h
m
) = 1, . . . , p +m ,
we can dene
i
= (
m
)
/dom(h
i
)
.
We also dene h
0
= h,
0
= .
We can prove that for each i = 1 . . . m
i
(h
i
) and there is s
i
#(h
i1
,
i
,
i1
)
such that
i
=
i1
+ (x
i
, s
i
).
Well prove this by induction on i. Let us perform the initial step.
If m = 1 then
1
=
m
(h
1
). Otherwise
1
= (
m
)
/dom(h
1
)
. Since h
1
!(h
m
) we
can apply assumption 2.1.11 and determine that
1
(h
1
).
At this point we need to show that
0
!(
1
). We have that _
m
.
If m = 1 this means precisely that
0
!(
1
).
Otherwise there exists C T such that C dom(
m
) and = (
m
)
/C
. We have
C = dom() = dom(h) dom(h
1
) = dom(
1
). Suppose
m
= (z
m
, s
m
), then
(
1
)
/C
= ((
m
)
/dom(h
1
)
)
/C
= ((z
m
)
/dom(h
1
)
, (s
m
)
/dom(h
1
)
)
/C
=
= (((z
m
)
/dom(h
1
)
)
/C
, ((s
m
)
/dom(h
1
)
)
/C
) = ((z
m
)
/C
, (s
m
)
/C
) = (
m
)
/C
= .
Therefore
0
!(
1
).
We observe that h, h
1
K(n), x
1
1 var(h),
1
E
s
(n, h), h
1
= h + (x
1
,
1
),
and also
0
(h),
1
(h
1
),
0
_
1
as already seen. By lemma 2.1.13 there is
s
1
#(h
0
,
1
,
0
) such that
1
=
0
+ (x
1
, s
1
).
If m > 1 we need an inductive step. Let i = 1 . . . m 1. We suppose
i
(h
i
) and
there is s
i
#(h
i1
,
i
,
i1
) such that
i
=
i1
+ (x
i
, s
i
).
If i + 1 = m then
i+1
=
m
(h
i+1
). Otherwise
i+1
= (
m
)
/dom(h
i+1
)
. Since
h
i+1
!(h
m
) we can apply assumption 2.1.11 and determine that
i+1
(h
i+1
).
At this point we need to show that
i
!(
i+1
). Consider that dom(
i+1
) =
dom(h
i+1
) = 1, . . . , p +i + 1. We have
(
i+1
)
/{1,...,p+i}
= ((
m
)
/dom(h
i+1
)
)
/{1,...,p+i}
=
= ((z
m
)
/dom(h
i+1
)
, (s
m
)
/dom(h
i+1
)
)
/{1,...,p+i}
=
= (((z
m
)
/dom(h
i+1
)
)
/{1,...,p+i}
, ((s
m
)
/dom(h
i+1
)
)
/{1,...,p+i}
) =
= ((z
m
)
/{1,...,p+i}
, (s
m
)
/{1,...,p+i}
) = (
m
)
/{1,...,p+i}
=
i
.
This proves
i
!(
i+1
).
We then observe that h
i
, h
i+1
K(n), x
i+1
1 var(h
i
),
i+1
E
s
(n, h
i
), h
i+1
=
h
i
+ (x
i+1
,
i+1
), and also
i
(h
i
),
i+1
(h
i+1
),
i
_
i+1
as already seen. By
lemma 2.1.13 there is s
i+1
#(h
i
,
i+1
,
i
) such that
i+1
=
i
+ (x
i+1
, s
i+1
).
We have proved that for each i = 1 . . . m
i
(h
i
) and there is s
i
#(h
i1
,
i
,
i1
)
such that
i
=
i1
+ (x
i
, s
i
).
38 M. Avon
We now dene
1
= + (x
1
, s
1
), and, if m > 1, for each i = 1 . . . m 1
i+1
=
i
+ (x
i+1
, s
i+1
).
By assumption 2.1.1 g = or there exists a positive integer q such that dom(g) =
1, . . . , q. In the case where g = we dene q = 0.
For each i = 1 . . . m we dene y
i
,
i
, r
i
as functions whose domain is 1, . . . , q + i such
that
for each j = 1 . . . q y
i
(j) = y(j),
i
(j) = (j), r
i
(j) = r(j);
for each j = 1 . . . i y
i
(q +j) = x
j
,
i
(q +j) =
j
, r
i
(q +j) = s
j
.
For each i = 1 . . . m we also dene z
i
,
i
, u
i
as functions whose domain is 1, . . . , p+i
such that
for each j = 1 . . . p z
i
(j) = z(j),
i
(j) = (j), u
i
(j) = u(j);
for each j = 1 . . . i z
i
(p +j) = x
j
,
i
(p +j) =
j
, u
i
(p +j) = s
j
.
We now prove that for each i = 1 . . . m
g
i
= (y
i
,
i
), h
i
= (z
i
,
i
),
i
= (y
i
, r
i
),
i
= (z
i
, u
i
).
We see that
g
1
= g + (x
1
,
1
) = (y, ) + (x
1
,
1
) = (y
1
,
1
),
h
1
= h + (x
1
,
1
) = (z, ) + (x
1
,
1
) = (z
1
,
1
),
1
= + (x
1
, s
1
) = (y, r) + (x
1
, s
1
) = (y
1
, r
1
),
1
= + (x
1
, s
1
) = (z, u) + (x
1
, s
1
) = (z
1
, u
1
).
If m > 1 our proof needs an inductive step. In this case, given i = 1 . . . m1, we see
that
g
i+1
= g
i
+ (x
i+1
,
i+1
) = (y
i
,
i
) + (x
i+1
,
i+1
) = (y
i+1
,
i+1
),
h
i+1
= h
i
+ (x
i+1
,
i+1
) = (z
i
,
i
) + (x
i+1
,
i+1
) = (z
i+1
,
i+1
),
i+1
=
i
+ (x
i+1
, s
i+1
) = (y
i
, r
i
) + (x
i+1
, s
i+1
) = (y
i+1
, r
i+1
),
i+1
=
i
+ (x
i+1
, s
i+1
) = (z
i
, u
i
) + (x
i+1
, s
i+1
) = (z
i+1
, u
i+1
).
We also can prove that for each = 1 . . . m
for each i dom(g
), j dom(h
) (y
)
i
= (z
)
j
(
)
i
= (
)
j
;
for each i dom(
), j dom(
) (y
)
i
= (z
)
j
(r
)
i
= (u
)
j
.
In fact, let = 1 . . . m. We notice that
dom(g
) = 1, . . . , q + = dom(
),
dom(h
) = 1, . . . , p + = dom(
).
Let i 1, . . . , q +, j 1, . . . , p +.
If q > 0, i q, p > 0, j p and y
i
= (y
)
i
= (z
)
j
= z
j
then
(
)
i
=
i
=
j
= (
)
j
and (r
)
i
= r
i
= u
j
= (u
)
j
.
A dierent approach to logic 39
If q > 0, i q, j > p then (y
)
i
= y
i
var(g), (z
)
j
= x
jp
1 var(g) so
(y
)
i
,= (z
)
j
.
If i > q, p > 0, j p then (y
)
i
= x
iq
1 var(h), (z
)
j
= z
j
var(h) so
(y
)
i
,= (z
)
j
.
If i > q, j > p and x
iq
= (y
)
i
= (z
)
j
= x
jp
then i q = j p, so
(
)
i
=
iq
=
jp
= (
)
j
and (r
)
i
= s
iq
= s
jp
= (u
)
j
.
Well now show that for each i = 1 . . . m
i
(g
i
).
We begin by showing that
1
(g
1
). We intend to use assumption 2.1.9 to show
that s
1
#(g,
1
, ).
We consider that g, h K(n), for each i dom(g), j dom(h), y
i
= z
j
i
=
j
,
1
E(n, g) E(n, h), (g), (h), for each i dom(), j dom(), y
i
= z
j
r
i
= u
j
. By assumption 2.1.9 #(g,
1
, ) = #(h,
1
, ), so s
1
#(g,
1
, ).
We can now use lemma 2.1.14 to show that
1
(g
1
). In fact g, g
1
K(n),
x
1
1 var(g),
1
E
s
(n, g), g
1
= g + (x
1
,
1
), (g),
1
= + (x
1
, s
1
),
s
1
#(g,
1
, ). So by 2.1.14 we get
1
(g
1
).
If m > 1 we need an inductive step. Let = 1 . . . m1, we suppose that
(g
)
and want to show that
+1
(g
+1
). First we intend to use assumption 2.1.9 to show
that s
+1
#(g
,
+1
,
).
We consider that g
, h
), j dom(h
), (y
)
i
=
(z
)
j
(
)
i
= (
)
j
,
+1
E(n, g
) E(n, h
),
(g
),
(h
), for
each i dom(
), j dom(
) (y
)
i
= (z
)
j
(r
)
i
= (u
)
j
. By assumption 2.1.9
#(g
,
+1
,
) = #(h
,
+1
,
), so s
+1
#(g
,
+1
,
).
We can now use lemma 2.1.14 to show that
+1
(g
+1
). In fact
g
, g
+1
K(n), x
+1
1 var(g
),
+1
E
s
(n, g
), g
+1
= g
+ (x
+1
,
+1
),
(g
),
+1
=
+ (x
+1
, s
+1
), s
+1
#(g
,
+1
,
). So by 2.1.14 we get
+1
(g
+1
).
We can conclude that
m
(g
m
). By 2.1.9 we can derive that
#(g
m
, ,
m
) = #(h
m
, ,
m
). In fact g
m
, h
m
K(n), for each i dom(g
m
),
j dom(h
m
), (y
m
)
i
= (z
m
)
j
(
m
)
i
= (
m
)
j
, E(n, g
m
) E(n, h
m
),
m
(g
m
),
m
(h
m
), for each i dom(
m
), j dom(
m
) (y
m
)
i
= (z
m
)
j
(r
m
)
i
= (u
m
)
j
. Therefore #(g
m
, ,
m
) = #(h
m
, ,
m
).
It follows that u = #(g
m
, ,
m
), and since
m
(g
m
), _
m
also hold, we have
proved that
u #(g
m
, ,
m
)[
m
(g
m
), _
m
.
With a perfectly analogous proof we can show the converse implication i.e. that if
u #(g
m
, ,
m
)[
m
(g
m
), _
m
then
u #(h
m
, ,
m
)[
m
(h
m
), _
m
.
40 M. Avon
Lemma 2.1.16. Let h = (x, ) K(n), E
s
(n, h), y 1 var(h) and
k = h + (y, ). Let = (x, s) (h), r #(h, , ) and = + (y, r). Then k is a
state-like pair (x
, s
). Moreover
for each i dom(k), j dom(h) x
i
= x
j
i
=
j
.
for each i dom(), j dom() x
i
= x
j
s
i
= s
j
.
Proof. If dom(h) = then dom() = and the statements are trivially satised.
Otherwise there exists a positive integer m such that dom() = dom(h) = 1, . . . , m,
and dom() = dom(k) = 1, . . . , m+ 1.
Let i dom(k), j dom(h). If i = m + 1 then x
i
= y / var(h), so x
i
,= x
j
. Else
x
i
= x
i
, so x
i
= x
j
implies x
i
= x
j
, which implies i = j and
i
=
i
=
j
.
Let i dom(), j dom(). If i = m + 1 then x
i
= y / var(h), so x
i
,= x
j
. Else
x
i
= x
i
, so x
i
= x
j
implies x
i
= x
j
, which implies i = j and s
i
= s
i
= s
j
.
We now start with the verications required to dene #(k, t, ). There we have to
verify that
for each k K(n + 1), w a, b, c, d, e, t E
(n, k) E
w
(n + 1, k) and (k)
#(k, t, ) = #(k, t, )
(n+1,k,w)
;
for each k K(n + 1), w
1
, w
2
a, b, c, d, e : w
1
,= w
2
, t E
w
1
(n + 1, k)
E
w
2
(n + 1, k) and (k) #(k, t, )
(n+1,k,w
1
)
= #(k, t, )
(n+1,k,w
2
)
.
We begin by verifying the rst statement.
Suppose t E
(n, k) E
a
(n +1, k), and so t E(n, k) E
a
(n + 1, k). As a conse-
quence of t E
a
(n+1, k) we have that k K(n)
+
, so there exist h K(n), E
s
(n, h),
y 1 var(h) such that k = h +(y, ). We also have t E(n, h). Given (k) there
exist (h), s #(h, , ) such that = +(y, s), and #(k, t, )
(n+1,k,a)
= #(h, t, ).
We want to apply assumption 2.1.9. We observe that h = (x, ), k = (x
) K(n),
for each i dom(k), j dom(h) x
i
= x
j
i
=
j
. Moreover t E(n, k) E(n, h),
= (x, r) (h), = (x
, r
i
= x
j
r
i
= r
j
.
At this point by assumption 2.1.9 we have #(k, t, ) = #(h, t, ) = #(k, t, )
(n+1,k,a)
.
Suppose t E
(n, k) E
b
(n +1, k), and so t E(n, k) E
b
(n + 1, k). As a conse-
quence of t E
b
(n+1, k) we have that k K(n)
+
, so there exist h K(n), E
s
(n, h),
y 1 var(h) such that k = h + (y, ). We also have t = y. Given (k) there exist
(h), s #(h, , ) such that = + (y, s), and #(k, t, )
(n+1,k,b)
= s.
Suppose h = (x, ) and = (x, r), then we can also set k = (x
), = (x
, r
). By
assumption 2.1.10 i dom(k) such that t = x
i
, #(k, t, ) = r
i
. There exists an integer
m 0 such that dom(k) = 1, . . . , m + 1. Since y = t = x
i
it must be i = m + 1, so
#(k, t, ) = r
m+1
= s = #(k, t, )
(n+1,k,b)
.
A dierent approach to logic 41
Lets examine the situation in which t E
(n, k) E
c
(n +1, k), and then t belongs
to E(n, k) E
c
(n + 1, k). As a consequence of t E
c
(n + 1, k) there exist ,
1
, . . . ,
m
in E(n, k) such that t = ()(
1
, . . . ,
m
) and
#(k, t, )
(n+1,k,c)
= #(k, , )(#(k,
1
, ), . . . , #(k,
m
, )).
Since t E(n, k) we can apply assumption 2.1.10 and obtain that n > 1, there exists
h K(n1): h _ k, ,
1
, . . .
m
E(n1, h), t E(n, h), for each (h) #(h, , )
is a function with m arguments, (#(h,
1
, ), . . . , #(h,
m
, )) is a member of the domain
of #(h, , ), #(h, t, ) = #(h, , )(#(h,
1
, ), . . . , #(h,
m
, )).
Moreover, given (k) and (h): _ it results #(k, t, ) = #(h, t, ).
Given (k) we have dom(h) T and dom(h) dom(k) = dom(), so we can
dene =
/dom(h)
. If h = k then = (h). Otherwise by assumption 2.1.11 we still
get (h). Therefore
#(k, t, ) = #(h, t, ) = #(h, , )(#(h,
1
, ), . . . , #(h,
m
, )).
We want to apply assumption 2.1.9. We observe that k = (x, ), h = (y, ) K(n),
h !(k), for each i, j dom(k) i ,= j x
i
,= x
j
. Then (by lemma 2.2) for each
i dom(k), j dom(h) x
i
= y
j
i
=
j
. Moreover = (x, s) (k),
= (y, r) (h), !(), for each i, j dom() i ,= j x
i
,= x
j
. Then for
each i dom(), j dom() x
i
= y
j
s
i
= r
j
. Since E(n, k) E(n, h) and for
each = 1 . . . m
(n, k) E
d
(n +1, k), and then t belongs
to E(n, k)E
d
(n+1, k). As a consequence of t E
d
(n+1, k) there exist f T,
1
, . . . ,
m
in E(n, k) such that t = (f)(
1
, . . . ,
m
) and
#(k, t, )
(n+1,k,d)
= P
f
(#(k,
1
, ), . . . , #(k,
m
, )).
Since t E(n, k) we can apply assumption 2.1.10 and obtain that n > 1, there
exists h K(n 1): h _ k,
1
, . . .
m
E(n 1, h), t E(n, h), for each (h)
A
f
(#(h,
1
, ), . . . , #(h,
m
, )), #(h, t, ) = P
f
(#(h,
1
, ), . . . , #(h,
m
, )). Moreover,
given (k) and (h): _ it results #(k, t, ) = #(h, t, ).
Given (k) we have dom(h) T and dom(h) dom(k) = dom(), so we can
dene =
/dom(h)
. If h = k then = (h). Otherwise by assumption 2.1.11 we still
get (h). Therefore
#(k, t, ) = #(h, t, ) = P
f
(#(h,
1
, ), . . . , #(h,
m
, )).
We want to apply assumption 2.1.9. We observe that k = (x, ), h = (y, ) K(n),
h !(k), for each i, j dom(k) i ,= j x
i
,= x
j
. Then (by lemma 2.2) for each
i dom(k), j dom(h) x
i
= y
j
i
=
j
. Moreover = (x, s) (k),
= (y, r) (h), !(), for each i, j dom() i ,= j x
i
,= x
j
. Then for each
42 M. Avon
i dom(), j dom() x
i
= y
j
s
i
= r
j
. Since for each = 1 . . . m
(n, k) E
e
(n +1, k), and then t belongs to E(n, k) E
e
(n + 1, k). As a consequence
to t E
e
(n + 1, k) there exist
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, k, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ).
For a xed (k) we have
#(k, t, )
(n+1,k,e)
= #(k
m
, ,
m
)[
m
(k
m
), _
m
,
where k
1
= k +(x
1
,
1
), and if m > 1 for each i = 1 . . . m1 k
i+1
= k
i
+(x
i+1
,
i+1
).
Since t E(n, k) we can apply assumption 2.1.10 and obtain that n > 1, there exists
h K(n 1) such that
h _ k,
for each i = 1 . . . m
i
E(n 1),
E(n 1),
c(n 1, h, m, x, , ),
t E(n, h),
for each (k), (h) : _ it results #(k, t, ) = #(h, t, ),
for each (h) #(h, t, ) = #(h
m
, ,
m
)[
m
(h
m
), _
m
,
where h
1
= h+(x
1
,
1
), and if m > 1 for each i = 1 . . . m1 h
i+1
= h
i
+(x
i+1
,
i+1
).
Given (k) we have dom(h) T and dom(h) dom(k) = dom(), so we can
dene =
/dom(h)
. If h = k then = (h). Otherwise by assumption 2.1.11 we still
get (h). Therefore
#(k, t, ) = #(h, t, ) = #(h
m
, ,
m
)[
m
(h
m
), _
m
.
We clearly need to show that
#(h
m
, ,
m
)[
m
(h
m
), _
m
= #(k
m
, ,
m
)[
m
(k
m
), _
m
.
To show this we need lemma 2.1.15. To apply that lemma we observe that
h = (y, ), k = (z, ) K(n), since c(n 1, h, m, x, , ) holds then also
A dierent approach to logic 43
c(n, h, m, x, , ) holds, and c(n, k, m, x, , ) holds too. We have h !(k), so by
lemma 2.2 we obtain that for each i dom(k), j dom(h) z
i
= y
j
i
=
j
.
Moreover = (z, s) (k), = (y, r) (h), and since !() by lemma 2.2 we also
obtain that for each i dom(), j dom() z
i
= y
j
s
i
= r
j
. At this point we can
apply lemma 2.1.15 from which it follows
#(h
m
, ,
m
)[
m
(h
m
), _
m
= #(k
m
, ,
m
)[
m
(k
m
), _
m
,
and this clearly causes #(k, t, )
(n+1,k,e)
= #(k, t, ).
w
1
(n +1, k) E
w
2
(n +1, k)
and (k) #(k, t, )
(n+1,k,w
1
)
= #(k, t, )
(n+1,k,w
2
)
.
Fortunately for us, for many values of w
1
, w
2
its easy to see that
E
w
1
(n + 1, k) E
w
2
(n + 1, k) = .
In fact, consider all the cases in which w
1
, w
2
b, c, d, e and w
1
,= w
2
. It is immediate
to see that E
w
1
(n + 1, k) E
w
2
(n + 1, k) = (actually, assumption 2.1.10 is needed to
prove E
c
(n + 1, k) E
d
(n + 1, k) = ).
We miss to consider the cases where w
1
= a and w
2
b, c, d, e.
We can easily prove that E
a
(n + 1, k) E
b
(n + 1, k) = .
Suppose t E
a
(n+1, k)E
b
(n+1, k). This means that t E
a
(n+1, k) and k K(n)
+
,
so there exist h K(n), E
s
(n, h), y 1 var(h) such that k = h + (y, ). We also
have t E(n, h), and since t E
b
(n + 1, k) we have t = y. Since t E(n, h) we can
apply assumption 2.1.10, situations a,c,d,e can not occur, so situation b must occur. This
means that y = t var(h) and so we have reached a contradiction.
Therefore we just need to examine three cases: t E
a
(n + 1, k) E
c
(n + 1, k),
t E
a
(n + 1, k) E
d
(n + 1, k), t E
a
(n + 1, k) E
e
(n + 1, k).
We start with the case where t E
a
(n +1, k) E
c
(n +1, k), and so
t E
a
(n + 1, k) E
c
(n + 1, k).
As a consequence of t E
c
(n + 1, k) there exist ,
1
, . . . ,
m
in E(n, k) such that
t = ()(
1
, . . . ,
m
) and for each (k)
#(k, t, )
(n+1,k,c)
= #(k, , )(#(k,
1
, ), . . . , #(k,
m
, )).
As a consequence of t E
a
(n+1, k) we have that k K(n)
+
, so there exist h K(n),
E
s
(n, h), y 1 var(h) such that k = h + (y, ). We also have t E(n, h).
Given (k) there exist (h), s #(h, , ) such that = + (y, s), and
#(k, t, )
(n+1,k,a)
= #(h, t, ).
Since t E(n, h) we can apply assumption 2.1.10 and obtain that n > 1, there exists
g K(n 1): g _ h, ,
1
, . . .
m
E(n 1, g), t E(n, g), for each (g) #(g, , )
44 M. Avon
is a function with m arguments, (#(g,
1
, ), . . . , #(g,
m
, )) is a member of the domain
of #(g, , ), #(g, t, ) = #(g, , )(#(g,
1
, ), . . . , #(g,
m
, )).
Moreover, given (h) and (g): _ it results #(h, t, ) = #(g, t, ).
We have seen that given (k) there exist (h), s #(h, , ) such that
= + (y, s), and #(k, t, )
(n+1,k,a)
= #(h, t, ).
We have dom(g) T and dom(g) dom(h) = dom(), so we can dene =
/dom(g)
.
If g = h then = (g). Otherwise by assumption 2.1.11 we still get (g).
Therefore
#(k, t, )
(n+1,k,a)
= #(h, t, ) = #(g, t, ) = #(g, , )(#(g,
1
, ), . . . , #(g,
m
, )).
We have g = (x, ), k = (z, ) K(n), with g !(k). By lemma 2.2 we obtain
that for each i dom(k), j dom(g) z
i
= x
j
i
=
j
. Moreover = (z, s) (k),
= (x, r) (g), !() and so by lemma 2.2 we obtain that for each i dom(),
j dom() z
i
= x
j
s
i
= r
j
. Furthermore we can see that ,
1
, . . . ,
m
E(n, k)
E(n, g). By assumption 2.1.9 we can state
#(g, , )(#(g,
1
, ), . . . , #(g,
m
, )) = #(k, , )(#(k,
1
, ), . . . , #(k,
m
, )),
and therefore
#(k, t, )
(n+1,k,a)
= #(k, t, )
(n+1,k,c)
.
Consider now the case where t E
a
(n +1, k) E
d
(n +1, k), and so
t E
a
(n + 1, k) E
d
(n + 1, k).
As a consequence of t E
d
(n + 1, k) there exist f T,
1
, . . . ,
m
in E(n, k) such
that t = (f)(
1
, . . . ,
m
) and for each (k)
#(k, t, )
(n+1,k,d)
= P
f
(#(k,
1
, ), . . . , #(k,
m
, )).
As a consequence of t E
a
(n+1, k) we have that k K(n)
+
, so there exist h K(n),
E
s
(n, h), y 1 var(h) such that k = h + (y, ). We also have t E(n, h).
Given (k) there exist (h), s #(h, , ) such that = + (y, s), and
#(k, t, )
(n+1,k,a)
= #(h, t, ).
Since t E(n, h) we can apply assumption 2.1.10 and obtain that n > 1, there
exists g K(n 1): g _ h,
1
, . . .
m
E(n 1, g), t E(n, g), for each (g)
A
f
(#(g,
1
, ), . . . , #(g,
m
, )), #(g, t, ) = P
f
(#(g,
1
, ), . . . , #(g,
m
, )). Moreover,
given (h) and (g): _ it results #(h, t, ) = #(g, t, ).
We have seen that given (k) there exist (h), s #(h, , ) such that
= + (y, s), and #(k, t, )
(n+1,k,a)
= #(h, t, ).
We have dom(g) T and dom(g) dom(h) = dom(), so we can dene =
/dom(g)
.
If g = h then = (g). Otherwise by assumption 2.1.11 we still get (g).
Therefore
#(k, t, )
(n+1,k,a)
= #(h, t, ) = #(g, t, ) = P
f
(#(g,
1
, ), . . . , #(g,
m
, )).
A dierent approach to logic 45
We have g = (x, ), k = (z, ) K(n), with g !(k). By lemma 2.2 we obtain
that for each i dom(k), j dom(g) z
i
= x
j
i
=
j
. Moreover = (z, s) (k),
= (x, r) (g), !() and so by lemma 2.2 we obtain that for each i dom(),
j dom() z
i
= x
j
s
i
= r
j
. Furthermore we can see that
1
, . . . ,
m
E(n, k)
E(n, g). By assumption 2.1.9 we can state
P
f
(#(g,
1
, ), . . . , #(g,
m
, )) = P
f
(#(k,
1
, ), . . . , #(k,
m
, )),
and therefore
#(k, t, )
(n+1,k,a)
= #(k, t, )
(n+1,k,d)
.
Finally we examine the case where t E
a
(n +1, k) E
e
(n +1, k), and so
t E
a
(n + 1, k) E
e
(n + 1, k).
As a consequence to t E
e
(n + 1, k) there exist
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, k, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ).
For a xed (k) we have
#(k, t, )
(n+1,k,e)
= #(k
m
, ,
m
)[
m
(k
m
), _
m
,
where k
1
= k +(x
1
,
1
), and if m > 1 for each i = 1 . . . m1 k
i+1
= k
i
+(x
i+1
,
i+1
).
As a consequence of t E
a
(n+1, k) we have that k K(n)
+
, so there exist h K(n),
E
s
(n, h), y 1 var(h) such that k = h + (y, ). We also have t E(n, h).
Given (k) there exist (h), s #(h, , ) such that = + (y, s), and
#(k, t, )
(n+1,k,a)
= #(h, t, ).
Since t E(n, h) we can apply assumption 2.1.10 and obtain that n > 1, there exists
g K(n 1) such that
g _ h,
for each i = 1 . . . m
i
E(n 1),
E(n 1),
c(n 1, g, m, x, , ),
t E(n, g),
for each (h), (g) : _ it results #(h, t, ) = #(g, t, ),
for each (g) #(g, t, ) = #(g
m
, ,
m
)[
m
(g
m
), _
m
,
46 M. Avon
where g
1
= g +(x
1
,
1
), and if m > 1 for each i = 1 . . . m1 g
i+1
= g
i
+(x
i+1
,
i+1
).
We have seen that given (k) there exist (h), s #(h, , ) such that
= + (y, s), and #(k, t, )
(n+1,k,a)
= #(h, t, ).
We have dom(g) T and dom(g) dom(h) = dom(), so we can dene =
/dom(g)
.
If g = h then = (g). Otherwise by assumption 2.1.11 we still get (g).
Therefore
#(k, t, )
(n+1,k,a)
= #(h, t, ) = #(g, t, ) = #(g
m
, ,
m
)[
m
(g
m
), _
m
.
We clearly need to show that
#(g
m
, ,
m
)[
m
(g
m
), _
m
= #(k
m
, ,
m
)[
m
(k
m
), _
m
.
To show this we need lemma 2.1.15. To apply that lemma we observe that
g = (y, ), k = (z, ) K(n), since c(n 1, g, m, x, , ) holds then also
c(n, g, m, x, , ) holds, and c(n, k, m, x, , ) holds too. We have g !(k), so by
lemma 2.2 we obtain that for each i dom(k), j dom(g) z
i
= y
j
i
=
j
. Moreover
= (z, s) (k), = (y, r) (g), and since !() by lemma 2.2 we also obtain
that for each i dom(), j dom() z
i
= y
j
s
i
= r
j
. At this point we can apply
lemma 2.1.15 from which it follows
#(g
m
, ,
m
)[
m
(g
m
), _
m
= #(k
m
, ,
m
)[
m
(k
m
), _
m
.
and this clearly causes #(k, t, )
(n+1,k,a)
= #(k, t, )
(n+1,k,e)
.
w
(n+1, k) V
b
(t) = V
b
(t)
(n+1,k,w)
and V
f
(t) = V
f
(t)
(n+1,k,w)
;
for each k
1
, k
2
K(n+1), w
1
, w
2
a, b, c, d, e, t E
w
1
(n+1, k
1
) E
w
2
(n+1, k
2
)
such that t / E(n) we have V
b
(t)
(n+1,k
1
,w
1
)
= V
b
(t)
(n+1,k
2
,w
2
)
and
V
f
(t)
(n+1,k
1
,w
1
)
= V
f
(t)
(n+1,k
2
,w
2
)
.
Suppose t E(n) E
a
(n +1, k). As a consequence of t E
a
(n + 1, k) we have that
k K(n)
+
, so there exist h K(n), E
s
(n, h), y 1var(h) such that k = h+(y, ).
We also have t E(n, h), V
b
(t)
(n+1,k,a)
= V
b
(t) and V
f
(t)
(n+1,k,a)
= V
f
(t).
Suppose t E(n) E
b
(n +1, k). As a consequence of t E
b
(n + 1, k) we have that
k K(n)
+
, so there exist h K(n), E
s
(n, h), y 1var(h) such that k = h+(y, ).
We also have t = y, V
f
(t)
(n+1,k,b)
= y, V
b
(t)
(n+1,k,b)
= .
There exists g K(n) such that t E(n, g). By assumption 2.1.10
V
f
(t) = t = y = V
f
(t)
(n+1,k,b)
, V
b
(t) = = V
b
(t)
(n+1,k,b)
.
A dierent approach to logic 47
Suppose t E(n) E
c
(n +1, k). As a consequence of t E
c
(n + 1, k) there exist
,
1
, . . . ,
m
in E(n, k) such that t = ()(
1
, . . . ,
m
),
V
f
(t)
(n+1,k,c)
= V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t)
(n+1,k,c)
= V
b
() V
b
(
1
) V
b
(
m
).
There exists K(n) such that t E(n, ). By assumption 2.1.10 we get n > 1,
h K(n 1): h _ , ,
1
, . . .
m
E(n 1, h), t E(n, h),
V
f
(t) = V
f
() V
f
(
1
) V
f
(
m
) = V
f
(t)
(n+1,k,c)
,
V
b
(t) = V
b
() V
b
(
1
) V
b
(
m
) = V
b
(t)
(n+1,k,c)
.
Suppose t E(n) E
d
(n +1, k). As a consequence of t E
d
(n + 1, k) there exist
f T,
1
, . . . ,
m
in E(n, k) such that t = (f)(
1
, . . . ,
m
)
V
f
(t)
(n+1,k,d)
= V
f
(
1
) V
f
(
m
),
V
b
(t)
(n+1,k,d)
= V
b
(
1
) V
b
(
m
).
There exists K(n) such that t E(n, ). By assumption 2.1.10 we get n > 1,
h K(n 1): h _ ,
1
, . . .
m
E(n 1, h), t E(n, h),
V
f
(t) = V
f
(
1
) V
f
(
m
) = V
f
(t)
(n+1,k,d)
,
V
b
(t) = V
b
(
1
) V
b
(
m
) = V
b
(t)
(n+1,k,d)
.
Suppose t E(n) E
e
(n +1, k). As a consequence to t E
e
(n + 1, k) there exist
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, k, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ).
If m = 1 we have also
V
f
(t)
(n+1,k,e)
= V
f
(
1
) (V
f
() x
1
),
V
b
(t)
(n+1,k,e)
= x
1
V
b
(
1
) V
b
().
If m > 1 we have
V
f
(t)
(n+1,k,e)
= V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
),
V
b
(t)
(n+1,k,e)
= x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
().
There exists K(n) such that t E(n, ). By assumption 2.1.10 we obtain that
n > 1, there exists h K(n 1) such that
48 M. Avon
h _ ,
for each i = 1 . . . m
i
E(n 1),
E(n 1),
c(n 1, h, m, x, , ),
t E(n, h).
Moreover, if m = 1
V
f
(t) = V
f
(
1
) (V
f
() x
1
) = V
f
(t)
(n+1,k,e)
,
V
b
(t) = x
1
V
b
(
1
) V
b
() = V
b
(t)
(n+1,k,e)
.
If m > 1
V
f
(t) = V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
) = V
f
(t)
(n+1,k,e)
;
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
() = V
b
(t)
(n+1,k,e)
.
w
1
(n + 1, k
1
) E
w
2
(n + 1, k
2
) such that t / E(n) we have
V
b
(t)
(n+1,k
1
,w
1
)
= V
b
(t)
(n+1,k
2
,w
2
)
, V
f
(t)
(n+1,k
1
,w
1
)
= V
f
(t)
(n+1,k
2
,w
2
)
.
First of all we observe that for each k K(n + 1), t E
a
(n + 1, k) we have that
k K(n)
+
, so there exist h K(n), E
s
(n, h), y 1var(h) such that k = h+(y, ).
We also have t E(n, h), this means that t E(n). This implies that we just need to
verify
for each k
1
, k
2
K(n + 1), w
1
, w
2
b, c, d, e, t E
w
1
(n + 1, k
1
) E
w
2
(n + 1, k
2
)
such that t / E(n) we have
V
b
(t)
(n+1,k
1
,w
1
)
= V
b
(t)
(n+1,k
2
,w
2
)
, V
f
(t)
(n+1,k
1
,w
1
)
= V
f
(t)
(n+1,k
2
,w
2
)
.
For each k
1
, k
2
K(n + 1), w
1
, w
2
b, c, d, e if w
1
,= w
2
then
E
w
1
(n + 1, k
1
) E
w
2
(n + 1, k
2
) = .
So we just need to verify
for each k
1
, k
2
K(n + 1), w b, c, d, e, t E
w
(n + 1, k
1
) E
w
(n + 1, k
2
) such
that t / E(n) we have
V
b
(t)
(n+1,k
1
,w)
= V
b
(t)
(n+1,k
2
,w)
, V
f
(t)
(n+1,k
1
,w)
= V
f
(t)
(n+1,k
2
,w)
.
Suppose t E
b
(n +1, k
1
) E
b
(n +1, k
2
).
From t E
b
(n + 1, k
1
) we obtain that k
1
K(n)
+
, so there exist h
1
K(n),
1
E
s
(n, h
1
), y
1
1 var(h
1
) such that k
1
= h
1
+ (y
1
,
1
). We also have t = y
1
,
V
f
(t)
(n+1,k
1
,b)
= y
1
, V
b
(t)
(n+1,k
1
,b)
= .
From t E
b
(n + 1, k
2
) we obtain that k
2
K(n)
+
, so there exist h
2
K(n),
2
E
s
(n, h
2
), y
2
1 var(h
2
) such that k
2
= h
2
+ (y
2
,
2
). We also have t = y
2
,
A dierent approach to logic 49
V
f
(t)
(n+1,k
2
,b)
= y
2
, V
b
(t)
(n+1,k
2
,b)
= .
Hence V
b
(t)
(n+1,k
1
,b)
= = V
b
(t)
(n+1,k
2
,b)
, V
f
(t)
(n+1,k
1
,b)
= t = V
f
(t)
(n+1,k
2
,b)
.
Suppose t E
c
(n +1, k
1
) E
c
(n +1, k
2
).
As a consequence of t E
c
(n + 1, k
1
) there exist ,
1
, . . . ,
m
in E(n, k
1
) such that
t = ()(
1
, . . . ,
m
),
V
f
(t)
(n+1,k
1
,c)
= V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t)
(n+1,k
1
,c)
= V
b
() V
b
(
1
) V
b
(
m
).
As a consequence of t E
c
(n + 1, k
2
) there exist ,
1
, . . . ,
p
in E(n, k
2
) such that
t = ()(
1
, . . . ,
p
),
V
f
(t)
(n+1,k
2
,c)
= V
f
() V
f
(
1
) V
f
(
p
),
V
b
(t)
(n+1,k
2
,c)
= V
b
() V
b
(
1
) V
b
(
p
).
So ()(
1
, . . . ,
m
) = t = ()(
1
, . . . ,
p
), it follows p = m, = ,
i
=
i
, hence
V
f
(t)
(n+1,k
1
,c)
= V
f
() V
f
(
1
) V
f
(
m
) = V
f
(t)
(n+1,k
2
,c)
,
V
b
(t)
(n+1,k
1
,c)
= V
b
() V
b
(
1
) V
b
(
m
) = V
b
(t)
(n+1,k
2
,c)
.
Suppose t E
d
(n +1, k
1
) E
d
(n +1, k
2
).
As a consequence of t E
d
(n+1, k
1
) there exist f T,
1
, . . . ,
m
in E(n, k
1
) such that
t = (f)(
1
, . . . ,
m
)
V
f
(t)
(n+1,k
1
,d)
= V
f
(
1
) V
f
(
m
),
V
b
(t)
(n+1,k
1
,d)
= V
b
(
1
) V
b
(
m
).
As a consequence of t E
d
(n + 1, k
2
) there exist g T,
1
, . . . ,
p
in E(n, k
2
) such
that t = (g)(
1
, . . . ,
p
)
V
f
(t)
(n+1,k
2
,d)
= V
f
(
1
) V
f
(
p
),
V
b
(t)
(n+1,k
2
,d)
= V
b
(
1
) V
b
(
p
).
So (f)(
1
, . . . ,
m
) = t = (g)(
1
, . . . ,
p
), if follows g = f, p = m,
i
=
i
, hence
V
f
(t)
(n+1,k
1
,d)
= V
f
(
1
) V
f
(
m
) = V
f
(t)
(n+1,k
2
,d)
,
V
b
(t)
(n+1,k
1
,d)
= V
b
(
1
) V
b
(
m
) = V
b
(t)
(n+1,k
2
,d)
.
Suppose t E
e
(n +1, k
1
) E
e
(n +1, k
2
).
As a consequence to t E
e
(n + 1, k
1
) there exist
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(k
1
), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, k
1
, m, x, , ),
50 M. Avon
t = (x
1
:
1
, . . . , x
m
:
m
, ).
If m = 1 we have also
V
f
(t)
(n+1,k
1
,e)
= V
f
(
1
) (V
f
() x
1
),
V
b
(t)
(n+1,k
1
,e)
= x
1
V
b
(
1
) V
b
().
If m > 1 we have
V
f
(t)
(n+1,k
1
,e)
= V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
),
V
b
(t)
(n+1,k
1
,e)
= x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
().
As a consequence to t E
e
(n + 1, k
2
) there exist
a positive integer p,
a function y whose domain is 1, . . . , p such that for each i = 1 . . . p
y
i
1 var(k
2
), and for each i, j = 1 . . . m i ,= j y
i
,= y
j
,
a function whose domain is 1, . . . , p such that for each i = 1 . . . p
i
E(n),
E(n)
such that
c(n, k
2
, p, y, , ),
t = (y
1
:
1
, . . . , y
p
:
p
, ).
If p = 1 we have also
V
f
(t)
(n+1,k
2
,e)
= V
f
(
1
) (V
f
() y
1
),
V
b
(t)
(n+1,k
2
,e)
= y
1
V
b
(
1
) V
b
().
If p > 1 we have
V
f
(t)
(n+1,k
2
,e)
= V
f
(
1
) (V
f
(
2
) y
1
) (V
f
(
p
) y
1
, . . . , y
p1
)
(V
f
() y
1
, . . . , y
p
),
V
b
(t)
(n+1,k
2
,e)
= y
1
, . . . , y
p
V
b
(
1
) V
b
(
p
) V
b
().
So (x
1
:
1
, . . . , x
m
:
m
, ) = t = (y
1
:
1
, . . . , y
p
:
p
, ), it follows p = m,
y = x, = , = . Hence if m = 1
V
f
(t)
(n+1,k
2
,e)
= V
f
(
1
) (V
f
() y
1
) = V
f
(
1
) (V
f
() x
1
) = V
f
(t)
(n+1,k
1
,e)
,
V
b
(t)
(n+1,k
2
,e)
= y
1
V
b
(
1
) V
b
() = x
1
V
b
(
1
) V
b
() = V
b
(t)
(n+1,k
1
,e)
.
A dierent approach to logic 51
If m > 1
V
f
(t)
(n+1,k
2
,e)
= V
f
(
1
) (V
f
(
2
) y
1
) (V
f
(
p
) y
1
, . . . , y
p1
)
(V
f
() y
1
, . . . , y
p
)
= V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
)
= V
f
(t)
(n+1,k
1
,e)
;
V
b
(t)
(n+1,k
2
,e)
= y
1
, . . . , y
p
V
b
(
1
) V
b
(
p
) V
b
()
= x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
()
= V
b
(t)
(n+1,k
1
,e)
.
In the last part of our denition we need to prove that all the assumptions we have
made at step n are true at step n + 1. The order in which well provide these proofs is
not the same in which we have listed the assumptions, but this of course is not a problem.
Proof of (assumption) 2.1.8 (at level n + 1).
We need to prove that K(n) K(n + 1), this is obvious by denition.
Proof of 2.1.12.
We need to prove that for each k K(n) E(n, k) E(n + 1, k).
For each k K(n) we have k K(n + 1) and
E(n + 1, k) = E
(n, k) E
a
(n + 1, k) E
b
(n + 1, k) E
c
(n + 1, k) E
d
(n + 1, k)
E
e
(n + 1, k)
= E(n, k) E
a
(n + 1, k) E
b
(n + 1, k) E
c
(n + 1, k) E
d
(n + 1, k)
E
e
(n + 1, k).
Proof of 2.1.4.
We need to prove that for each k K(n + 1), t E(n + 1, k)
t[(t)] ,= ( ;
if t[(t)] = ) then d(t, (t)) = 1, else d(t, (t)) = 0 ;
for each 1, . . . , (t) if (t[] = :) (t[] = ,) (t[] = )) then d(t, ) 1.
We recall that
E(n+1, k) = E
(n, k)E
a
(n+1, k)E
b
(n+1, k)E
c
(n+1, k)E
d
(n+1, k)E
e
(n+1, k).
Let t E
(n, k), this means that t E(n, k) E(n). In this case we just need to
apply assumption 2.1.4.
52 M. Avon
Let t E
a
(n +1, k). As a consequence of t E
a
(n + 1, k) we have that k K(n)
+
,
so there exist h K(n), E
s
(n, h), y 1 var(h) such that k = h + (y, ). We also
have t E(n, h), so we can apply assumption 2.1.4 to nish.
Let t E
b
(n +1, k). As a consequence of t E
b
(n + 1, k) we have that k K(n)
+
,
so there exist h K(n), E
s
(n, h), y 1 var(h) such that k = h + (y, ). We also
have t = y, so t has just one character, t[1] diers from (, :, ,, ) and d(t, (t)) = 0.
Let t E
c
(n +1, k). As a consequence of t E
c
(n + 1, k) there exist ,
1
, . . . ,
m
in E(n, k) such that t = ()(
1
, . . . ,
m
).
If we know m we can provide an explicit representation of t. In fact if m = 2 then
t = ()(
1
,
2
), if m = 3 then t = ()(
1
,
2
,
3
) and so on. In this explicit representation
we can see explicit occurrences of the symbols , and ). There are explicit occurrences
of , only when m > 1. We indicate with q the position of the rst explicit occurrence
of ), and the second explicit occurrence of ) is clearly in position (t). If m > 1 we
indicate with q
1
, . . . , q
m1
the positions of the explicit occurrences of ,.
We have d(t, q 1) = d(t, 1 +()) = d(t, 1 + 1) +d(, ()) = 1 +d(, ()).
If t[q 1] = [()] = ) then d(t, q) = d(t, q 1) 1 = d(, ()) = 1.
Else t[q 1] = [()] / (, ), so d(t, q) = d(t, q 1) = 1 +d(, ()) = 1.
If m > 1 we can prove that for each i = 1 . . . m1 d(t, q
i
) = 1.
First of all we agree that d(t, q + 2) = d(t, q) 1 + 1 = 1.
We have also that
d(t, q
1
1) = d(t, q + 1 +(
1
)) = d(t, q + 1 + 1) +d(
1
, (
1
)) = 1 +d(
1
, (
1
)).
If t[q
1
1] =
1
[(
1
)] = ) then d(t, q
1
) = d(t, q
1
1) 1 = d(
1
, (
1
)) = 1 .
Else t[q
1
1] =
1
[(
1
)] / (, ) so d(t, q
1
) = d(t, q
1
1) = 1 +d(
1
, (
1
)) = 1.
If m = 2 we have nished this step. Now suppose m > 2. Let i = 1 . . . m 2 and
suppose d(t, q
i
) = 1. Well show that d(t, q
i+1
) = 1 also holds.
In fact
d(t, q
i+1
1) = d(t, q
i
+(
i+1
)) = d(t, q
i
+ 1) +d(
i+1
, (
i+1
)) =
= 1 +d(
i+1
, (
i+1
)).
If t[q
i+1
1] =
i+1
[(
i+1
)] = ) then
d(t, q
i+1
) = d(t, q
i+1
1) 1 = d(
i+1
, (
i+1
)) = 1.
Else t[q
i+1
1] =
i+1
[(
i+1
)] / (, ) so
d(t, q
i+1
) = d(t, q
i+1
1) = 1 +d(
i+1
, (
i+1
)) = 1.
So it is shown that for each i = 1 . . . m1 d(t, q
i
) = 1.
We now want to show that d(t, (t)) = 1.
A dierent approach to logic 53
If m = 1 then
d(t, (t) 1) = d(t, q + 1 +(
1
)) = d(t, q + 2) +d(
1
, (
1
)) = d(t, q) +d(
1
, (
1
)) =
= 1 +d(
m
, (
m
)).
If m > 1 then
d(t, (t) 1) = d(t, q
m1
+(
m
)) = d(t, q
m1
+ 1) +d(
m
, (
m
)) = 1 +d(
m
, (
m
)).
If t[(t) 1] =
m
[(
m
)] = ) then d(t, (t)) = d(t, (t) 1) 1 = d(
m
, (
m
)) = 1.
Else t[(t)1] =
m
[(
m
)] / (, ) so d(t, (t)) = d(t, (t)1) = 1+d(
m
, (
m
)) = 1.
Lets now examine the facts we have to prove. It is true that t[(t)] ,= (. Its also true
that t[(t)] = ) and d(t, (t)) = 1.
Now let 1, . . . , (t) and ( t[] = : or t[] = , or t[] = ) ).
If q, q
1
, . . . , q
m1
, (t) we have already shown that d(t, ) = 1. Otherwise there
are these alternative possibilities:
a. ( > 1) ( < q),
b. (m = 1) ( > q + 1) ( < (t)),
c. (m > 1) ( > q + 1) ( < q
1
),
d. (m > 2) (i = 1 . . . m2 : ( > q
i
) ( < q
i+1
)),
e. (m > 1) ( > q
m1
) ( < (t)).
In the situation a. [ 1] = t[],
d(t, ) = d(t, 1 + ( 1)) = d(t, 2) +d(, 1) = 1 +d(, 1) 2.
In the situation b. we have
q + 1 < < (t),
0 < (q + 1) < (t) (q + 1),
1 (q + 1) (t) q 2 = (
1
),
1
[ (q + 1)] = t[],
d(t, ) = d(t, q + 1 + ( (q + 1))) = d(t, q + 2) +d(
1
, (q + 1)) =
= 1 +d(
1
, (q + 1)) 2.
In the situation c. we have
q + 1 < < q
1
,
0 < (q + 1) < q
1
(q + 1),
1 (q + 1) q
1
q 2 = (
1
),
1
[ (q + 1)] = t[],
d(t, ) = d(t, q + 1 + ( (q + 1))) = d(t, q + 2) +d(
1
, (q + 1)) =
= 1 +d(
1
, (q + 1)) 2.
54 M. Avon
In the situation d. we have
q
i
< < q
i+1
,
0 < q
i
< q
i+1
q
i
,
1 q
i
q
i+1
q
i
1 = (
i+1
),
i+1
[ q
i
] = t[],
d(t, ) = d(t, q
i
+ ( q
i
)) = d(t, q
i
+ 1) +d(
i+1
, q
i
) =
= 1 +d(
i+1
, q
i
) 2.
In the situation e. we have
q
m1
< < (t),
0 < q
m1
< (t) q
m1
,
1 q
m1
(t) q
m1
1 = (
m
),
m
[ q
m1
] = t[],
d(t, ) = d(t, q
m1
+ ( q
m1
)) = d(t, q
m1
+ 1) +d(
m
, q
m1
) =
= 1 +d(
m
, q
m1
) 2.
Let t E
d
(n +1, k). As a consequence of t E
d
(n + 1, k) there exist f T,
1
, . . . ,
m
in E(n, k) such that t = (f)(
1
, . . . ,
m
).
If we know m we can provide an explicit representation of t. In fact if m = 2 then
t = (f)(
1
,
2
), if m = 3 then t = (f)(
1
,
2
,
3
) and so on. In this explicit representation
we can see explicit occurrences of the symbols , and ). The occurrences of ) are clearly
in positions 3 and (t). There are explicit occurrences of , only when m > 1. If m > 1
we indicate with q
1
, . . . , q
m1
the positions of explicit occurrences of ,.
It is immediate to see that d(t, 3) = 1.
If m > 1 we can prove that for each i = 1 . . . m1 d(t, q
i
) = 1.
We have d(t, q
1
1) = d(t, 4 +(
1
)) = d(t, 4 + 1) +d(
1
, (
1
)) = 1 +d(
1
, (
1
)).
If t[q
1
1] =
1
[(
1
)] = ) then d(t, q
1
) = d(t, q
1
1) 1 = d(
1
, (
1
)) = 1.
Else t[q
1
1] =
1
[(
1
)] / (, ), so d(t, q
1
) = d(t, q
1
1) = 1 +d(
1
, (
1
)) = 1.
If m = 2 we have nished this step. Now suppose m > 2. Let i = 1 . . . m 2 and
suppose d(t, q
i
) = 1. Well show that d(t, q
i+1
) = 1 also holds.
In fact
d(t, q
i+1
1) = d(t, q
i
+(
i+1
)) = d(t, q
i
+ 1) +d(
i+1
, (
i+1
)) =
= 1 +d(
i+1
, (
i+1
)).
If t[q
i+1
1] =
i+1
[(
i+1
)] = ) then
d(t, q
i+1
) = d(t, q
i+1
1) 1 = d(
i+1
, (
i+1
)) = 1.
A dierent approach to logic 55
Else t[q
i+1
1] =
i+1
[(
i+1
)] / (, ) so
d(t, q
i+1
) = d(t, q
i+1
1) = 1 +d(
i+1
, (
i+1
)) = 1.
So it is shown that for each i = 1 . . . m1 d(t, q
i
) = 1.
We now want to show that d(t, (t)) = 1.
If m = 1 then
d(t, (t) 1) = d(t, 4 +(
1
)) = d(t, 4 + 1) +d(
1
, (
1
)) = 1 +d(
1
, (
1
)) =
= 1 +d(
m
, (
m
)).
If m > 1 then
d(t, (t) 1) = d(t, q
m1
+(
m
)) = d(t, q
m1
+ 1) +d(
m
, (
m
)) = 1 +d(
m
, (
m
)).
If t[(t) 1] =
m
[(
m
)] = ) then d(t, (t)) = d(t, (t) 1) 1 = d(
m
, (
m
)) = 1.
Else t[(t)1] =
m
[(
m
)] / (, ) so d(t, (t)) = d(t, (t)1) = 1+d(
m
, (
m
)) = 1.
Lets now examine the facts we have to prove. It is true that t[(t)] ,= (. Its also true
that t[(t)] = ) and d(t, (t)) = 1.
Now let 1, . . . , (t) and ( t[] = : or t[] = , or t[] = ) ).
If 3, q
1
, . . . , q
m1
, (t) we have already shown that d(t, ) = 1. Otherwise there
are these alternative possibilities:
a. (m = 1) ( > 4) ( < (t)),
b. (m > 1) ( > 4) ( < q
1
),
c. (m > 2) (i = 1 . . . m2 : ( > q
i
) ( < q
i+1
)),
d. (m > 1) ( > q
m1
) ( < (t)).
In the situation a. we have
4 < < (t),
0 < 4 < (t) 4,
1 4 (t) 4 1 = (
1
),
1
[ 4] = t[],
d(t, ) = d(t, 4 + ( 4)) = d(t, 4 + 1) +d(
1
, 4) =
= 1 +d(
1
, 4) 2.
In the situation b. we have
4 < < q
1
,
0 < 4 < q
1
4,
1 4 q
1
4 1 = (
1
),
1
[ 4] = t[],
56 M. Avon
d(t, ) = d(t, 4 + ( 4)) = d(t, 4 + 1) +d(
1
, 4) =
= 1 +d(
1
, 4) 2.
In the situation c. we have
q
i
< < q
i+1
,
0 < q
i
< q
i+1
q
i
,
1 q
i
q
i+1
q
i
1 = (
i+1
),
i+1
[ q
i
] = t[],
d(t, ) = d(t, q
i
+ ( q
i
)) = d(t, q
i
+ 1) +d(
i+1
, q
i
) =
= 1 +d(
i+1
, q
i
) 2.
In the situation d. we have
q
m1
< < (t),
0 < q
m1
< (t) q
m1
,
1 q
m1
(t) q
m1
1 = (
m
),
m
[ q
m1
] = t[],
d(t, ) = d(t, q
m1
+ ( q
m1
)) = d(t, q
m1
+ 1) +d(
m
, q
m1
) =
= 1 +d(
m
, q
m1
) 2.
Let t E
e
(n +1, k). As a consequence to t E
e
(n + 1, k) there exist
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that t = (x
1
:
1
, . . . , x
m
:
m
, ).
If we know m we can provide an explicit representation of t. In fact if m = 2 then
t = (x
1
:
1
, x
2
:
2
, ), if m = 3 then t = (x
1
:
1
, x
2
:
2
, x
3
:
3
, ), and so on.
In this explicit representation of t we can see explict occurrences of the symbols , and
:. We indicate with q
1
, . . . , q
m
the positions of the explicit occurrences of : and with
r
1
. . . r
m
the positions of the explicit occurrences of ,. The only explicit occurrence of )
has the position (t). We want to show that for each i = 1 . . . m d(t, q
i
) = 1, d(t, r
i
) = 1
and that d(t, (t)) = 1.
It is obvious that d(t, q
1
) = 1. Moreover
d(t, r
1
1) = d(t, q
1
+ (r
1
1 q
1
)) = d(t, q
1
+(
1
)) = d(t, q
1
+ 1) +d(
1
, (
1
)) =
= 1 +d(
1
, (
1
)).
If t[r
1
1] =
1
[(
1
)] = ) then d(t, r
1
) = d(t, r
1
1) 1 = d(
1
, (
1
)) = 1.
Else t[r
1
1] =
1
[(
1
)] / (, ) so d(t, r
1
) = d(t, r
1
1) = 1 +d(
1
, (
1
)) = 1.
A dierent approach to logic 57
If m = 1 we have shown that for each i = 1 . . . m d(t, q
i
) = 1, d(t, r
i
) = 1. Now
suppose m > 1, let i = 1 . . . m 1 and suppose d(t, q
i
) = 1, d(t, r
i
) = 1. We show that
d(t, q
i+1
) = 1, d(t, r
i+1
) = 1.
We have q
i+1
= r
i
+ 2 and it is immediate that d(t, q
i+1
) = 1. Moreover
d(t, r
i+1
1) = d(t, q
i+1
+ (r
i+1
1 q
i+1
)) = d(t, q
i+1
+(
i+1
)) =
= d(t, q
i+1
+ 1) +d(
i+1
, (
i+1
)) = 1 +d(
i+1
, (
i+1
)).
If t[r
i+1
1] =
i+1
[(
i+1
)] = ) then
d(t, r
i+1
) = d(t, r
i+1
1) 1 = d(
i+1
, (
i+1
)) = 1.
Else t[r
i+1
1] =
i+1
[(
i+1
)] / (, ) so
d(t, r
i+1
) = d(t, r
i+1
1) = 1 +d(
i+1
, (
i+1
)) = 1.
Furthermore
d(t, (t) 1) = d(t, r
m
+ ((t) 1 r
m
)) = d(t, r
m
+()) =
= d(t, r
m
+ 1) +d(, ()) = 1 +d(, ()).
If t[(t) 1] = [()] = ) then d(t, (t)) = d(t, (t) 1) 1 = d(, ()) = 1.
Else t[(t) 1] = [()] / (, ) so d(t, (t)) = d(t, (t) 1) = 1 +d(, ()) = 1.
Lets now examine the facts we have to prove. It is true that t[(t)] ,= (. Its also true
that t[(t)] = ) and d(t, (t)) = 1.
Now let 1, . . . , (t) and ( t[] = : or t[] = , or t[] = ) ).
If q
1
, . . . , q
m
, r
1
, . . . , r
m
, (t) we have already shown that d(t, ) = 1. Otherwise
there are these alternative possibilities:
a. i = 1 . . . m such that q
i
< < r
i
,
b. r
m
< < (t).
In the situation a. we have
q
i
< < r
i
,
0 < q
i
< r
i
q
i
,
1 q
i
r
i
q
i
1 = (
i
),
i
[ q
i
] = t[],
d(t, ) = d(t, q
i
+ ( q
i
)) = d(t, q
i
+ 1) +d(
i
, q
i
) =
= 1 +d(
i
, q
i
) 2.
In the situation b. we have
r
m
< < (t),
0 < r
m
< (t) r
m
,
1 r
m
(t) r
m
1 = (),
[ r
m
] = t[],
58 M. Avon
d(t, ) = d(t, r
m
+ ( r
m
)) = d(t, r
m
+ 1) +d(, r
m
) =
= 1 +d(, r
m
) 2.
Proof of 2.1.1.
We need to prove that for each k K(n + 1) such that k ,= and for each (k)
there exist a positive integer m, a function x: 1, . . . , m 1, a function
: 1, . . . , m E(n + 1), a function s: 1, . . . , m M(n + 1) such that
for each i, j 1 . . . m (i ,= j x
i
,= x
j
),
k = (x, ),
= (x, s).
We can observe the following facts.
E(n) =
kK(n)
E(n, k)
kK(n)
E(n + 1, k) E(n + 1),
for each k K(n)
E
s
(n, k) = t[t E(n, k), (k) #(k, t, ) is a set
t[t E(n + 1, k), (k) #(k, t, ) is a set = E
s
(n + 1, k),
M(n, k) =
tE
s
(n,k)
M(k, t)
tE
s
(n+1,k)
M(k, t) = M(n + 1, k),
M(n) =
kK(n)
M(n, k)
kK(n)
M(n + 1, k) M(n + 1).
Now let k K(n + 1) such that k ,= , (k).
If k K(n) by our assumption there exist a positive integer m,
a function x: 1, . . . , m 1, a function : 1, . . . , m E(n), a function
s: 1, . . . , m M(n) such that
for each i, j 1 . . . m (i ,= j x
i
,= x
j
),
k = (x, ),
= (x, s).
This completes the proof in the case k K(n).
Now suppose k / K(n), i.e. k K(n)
+
. Then there exist h K(n), E
s
(n, h),
y (1 var(h)) such that k = h + (y, ).
By lemma 2.1.3 there exist (h), s #(h, , ) such that = + (y, s).
We can observe that E(n) E(n + 1), s M(h, ) M(n, h) M(n)
M(n + 1).
A dierent approach to logic 59
If dom(h) = then h = , = and dom() = . So we can dene three functions
x, , u over the domain 1 by setting x(1) = y, (1) = , u(1) = s. It clearly results
k = (x, ), = (x, u), and in this case the proof is nished.
We still need to examine the case where dom(h) ,= and so h ,= . We can ap-
ply our assumtpion 2.1.1 to h and and determine that there exist a positive integer m,
a function x: 1, . . . , m 1, a function : 1, . . . , m E(n), a function
u: 1, . . . , m M(n) such that
for each i, j 1 . . . m (i ,= j x
i
,= x
j
),
h = (x, ),
= (x, u).
We dene three new functions x
, u
() = x(),
() = (), u
() = u(), x
(m + 1) = y,
(m+ 1) = , u
(m+ 1) = s.
Since k = h+(y, ) we have k = (x
, u
).
We also observe that y / var(h) so for each i = 1 . . . m y ,= x
i
. This completes the
proof.
Proof of 2.1.2.
We need to prove that for each k K(n + 1)
(k = )
(g K(n), z 1 var(g), E
s
(n, g) :
k = g + (z, ) (k) = + (z, s)[ (g), s #(g, , )).
If k K(n) we can apply assumption 2.1.2 and get
(k = )
((n > 1) g K(n 1), z 1 var(g), E
s
(n 1, g) :
k = g + (z, ) (k) = + (z, s)[ (g), s #(g, , )).
In the case k ,= we have g K(n), E
s
(n, g). Therefore when k K(n) our
result is veried.
Now suppose k / K(n), i.e. k K(n)
+
. There exist h K(n), E
s
(n, h),
y (1 var(h)) such that k = h + (y, ). By lemma 2.1.3 we have also
(k) = + (y, s)[ (h), s #(h, , ) ,
and this completes the proof.
Proof of 2.1.11.
We need to prove that for each k K(n + 1), h !(k): h ,= k we have h K(n)
and for each (k) if we dene =
/dom(h)
then (h).
60 M. Avon
If k K(n) since k ,= (and therefore n > 1) we can exploit assumption 2.1.11 and
say that h K(n 1) K(n) and for each (k) if we dene =
/dom(h)
then
(h).
Now suppose k / K(n), i.e. k K(n)
+
. There exist g K(n), E
s
(n, g),
y (1 var(g)) such that k = g + (y, ). By lemma 2.1.3 we have also
(k) = + (y, s)[ (g), s #(g, , ) .
By lemma 2.3 h !(g) and we can distinguish two cases: h = g and h ,= g.
If h = g then h K(n). Let (k) and we dene =
/dom(h)
. There exist
(g), s #(g, , ) such that = + (y, s). We have
=
/dom()
=
/dom(g)
=
/dom(h)
= ,
so (h).
If h ,= g then we can apply assumption 2.1.11 to g and h and obtain that h K(n1),
for each (g) if we dene =
/dom(h)
then (h). So h K(n). Moreover, let
(k) and dene =
/dom(h)
. There exist (g), s #(g, , ) such that
= + (y, s), so (with the assumptions that = (x
), = (x
))
=
/dom(h)
= ((x
)
/dom(h)
, (
)
/dom(h)
) =
= (((x
)
/dom()
)
/dom(h)
, ((
)
/dom()
)
/dom(h)
) =
= ((x
)
/dom(h)
, (
)
/dom(h)
) =
/dom(h)
(h).
A dierent approach to logic 61
Proof of 2.1.10.
We need to prove that for each k K(n + 1), t E(n + 1, k) one and only one of
these ve alternative situations is veried:
a.
t (, (k) #(k, t, ) = #(t), V
f
(t) = , V
b
(t) = .
b.
if we set k = (x, ) then i dom(k) : (t = x
i
, = (x, s) (k) #(k, t, ) = s
i
),
V
f
(t) = t, V
b
(t) = .
c.
h K(n) : h _ k, m positive integer , ,
1
, . . .
m
E(n, h) :
t = ()(
1
, . . . ,
m
), t E(n + 1, h),
(h) ( #(h, , ) is a function with m arguments,
(#(h,
1
, ), . . . , #(h,
m
, )) is a member of the domain of #(h, , ),
#(h, t, ) = #(h, , )(#(h,
1
, ), . . . , #(h,
m
, )) ),
V
f
(t) = V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
() V
b
(
1
) V
b
(
m
),
(k), (h) : _ it results #(k, t, ) = #(h, t, ).
d.
h K(n) : h _ k, f T, m positive integer ,
1
, . . .
m
E(n, h) :
t = (f)(
1
, . . . ,
m
), t E(n + 1, h),
(h) ( A
f
(#(h,
1
, ), . . . , #(h,
m
, )),
#(h, t, ) = P
f
(#(h,
1
, ), . . . , #(h,
m
, )) ),
V
f
(t) = V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
(
1
) V
b
(
m
),
(k), (h) : _ it results #(k, t, ) = #(h, t, ).
62 M. Avon
e.
there exist
h K(n) : h _ k,
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(h), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, h, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ), t E(n + 1, h),
for each (h) #(h, t, ) = #(h
m
, ,
m
)[
m
(h
m
), _
(where h
1
= h + (x
1
,
1
), and if m > 1 for each i = 1 . . . m1
h
i+1
= h
i
+ (x
i+1
,
i+1
) ),
if m = 1 V
f
(t) = V
f
(
1
) (V
f
() x
1
), V
b
(t) = x
1
V
b
(
1
) V
b
(),
if m > 1
V
f
(t) = V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
),
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
(),
(k), (h) : _ it results #(k, t, ) = #(h, t, ).
We recall that
E(n+1, k) = E
(n, k)E
a
(n+1, k)E
b
(n+1, k)E
c
(n+1, k)E
d
(n+1, k)E
e
(n+1, k).
So we need to prove that
for each t E(n, k) one of the ve alternative situations is veried;
for each w a, b, c, d, e and t E
w
(n+1, k) one of the ve alternative situations
is veried.
Suppose t E
(n, k), this means that t E(n, k) and k K(n). This case is easily
solved, in fact we apply assumption 2.1.10 and obtain that one of the ve situations holds
at level n, but this means the situation is also veried at level n + 1.
Let t E
a
(n +1, k), this means that t E
a
(n + 1, k) and k K(n)
+
. There exist
h K(n), E
s
(n, h), y (1 var(h)) such that k = h + (y, ). We have t E(n, h),
so we can apply assumption 2.1.10 to h and t. Assumption 2.1.10 says that one of ve
alternative situations (referred to h, n) is true; we need to show that the corresponding
situation, referred to k, n + 1 is also true.
A dierent approach to logic 63
Lets consider the situation in which
t (, (h) #(h, t, ) = #(t), V
f
(t) = , V
b
(t) = .
In this case for each = + (y, s) (k)
#(k, t, ) = #(k, t, )
(n+1,k,a)
= #(h, t, ) = #(t).
So one of the ve alternative situations at level n + 1 is satised.
Consider the situation where n > 1, if we set h = (x, ) then i dom(h) :
(t = x
i
, = (x, u) (h) #(h, t, ) = u
i
), V
f
(t) = t, V
b
(t) = .
If we set k = (x
i
= x
i
= t. Moreover given = (x
, u
i
.
Consider the situation where
n > 1,
g K(n 1) : g _ h, m positive integer , ,
1
, . . .
m
E(n 1, g) :
t = ()(
1
, . . . ,
m
), t E(n, g),
(g) ( #(g, , ) is a function with m arguments,
(#(g,
1
, ), . . . , #(g,
m
, )) is a member of the domain of #(g, , ),
#(g, t, ) = #(g, , )(#(g,
1
, ), . . . , #(g,
m
, )) ),
V
f
(t) = V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
() V
b
(
1
) V
b
(
m
),
(h), (g) : _ it results #(h, t, ) = #(g, t, ).
We have
g K(n) : g _ k, m positive integer , ,
1
, . . .
m
E(n, g) :
t = ()(
1
, . . . ,
m
), t E(n + 1, g),
(g) ( #(g, , ) is a function with m arguments,
(#(g,
1
, ), . . . , #(g,
m
, )) is a member of the domain of #(g, , ),
#(g, t, ) = #(g, , )(#(g,
1
, ), . . . , #(g,
m
, )) ),
V
f
(t) = V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
() V
b
(
1
) V
b
(
m
).
Moreover, given = + (y, s) (k), (g): _ by lemma 2.3 we have _
and so it results #(k, t, ) = #(k, t, )
(n+1,k,a)
= #(h, t, ) = #(g, t, ).
64 M. Avon
Consider the situation where
n > 1,
g K(n 1) : g _ h, f T, m positive integer ,
1
, . . .
m
E(n 1, g) :
t = (f)(
1
, . . . ,
m
), t E(n, g),
(g) ( A
f
(#(g,
1
, ), . . . , #(g,
m
, )),
#(g, t, ) = P
f
(#(g,
1
, ), . . . , #(g,
m
, )) ),
V
f
(t) = V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
(
1
) V
b
(
m
),
(h), (g) : _ it results #(h, t, ) = #(g, t, ).
We have
g K(n) : g _ k, f T, m positive integer ,
1
, . . .
m
E(n, g) :
t = (f)(
1
, . . . ,
m
), t E(n + 1, g),
(g) ( A
f
(#(g,
1
, ), . . . , #(g,
m
, )),
#(g, t, ) = P
f
(#(g,
1
, ), . . . , #(g,
m
, )) ),
V
f
(t) = V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
(
1
) V
b
(
m
).
Moreover, given = + (y, s) (k), (g): _ by lemma 2.3 we have _
and so it results #(k, t, ) = #(k, t, )
(n+1,k,a)
= #(h, t, ) = #(g, t, ).
A dierent approach to logic 65
Consider the situation where
n > 1,
there exist
g K(n 1) : g _ h,
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(g), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n 1),
E(n 1)
such that
c(n 1, g, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ), t E(n, g),
for each (g) #(g, t, ) = #(g
m
, ,
m
)[
m
(g
m
), _
(where g
1
= g + (x
1
,
1
), and if m > 1 for each i = 1 . . . m1
g
i+1
= g
i
+ (x
i+1
,
i+1
) ),
if m = 1 V
f
(t) = V
f
(
1
) (V
f
() x
1
), V
b
(t) = x
1
V
b
(
1
) V
b
(),
if m > 1
V
f
(t) = V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
),
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
(),
(h), (g) : _ it results #(h, t, ) = #(g, t, ).
66 M. Avon
We have
g K(n) : g _ k,
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(g), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, g, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ), t E(n + 1, g),
for each (g) #(g, t, ) = #(g
m
, ,
m
)[
m
(g
m
), _
(where g
1
= g + (x
1
,
1
), and if m > 1 for each i = 1 . . . m1
g
i+1
= g
i
+ (x
i+1
,
i+1
) ),
if m = 1 V
f
(t) = V
f
(
1
) (V
f
() x
1
), V
b
(t) = x
1
V
b
(
1
) V
b
(),
if m > 1
V
f
(t) = V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
),
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
().
Moreover, given = + (y, s) (k), (g): _ by lemma 2.3 we have _
and so it results #(k, t, ) = #(k, t, )
(n+1,k,a)
= #(h, t, ) = #(g, t, ).
Let t E
b
(n +1, k), this means that t E
b
(n + 1, k) and k K(n)
+
. There exist
h K(n), E
s
(n, h), y (1 var(h)) such that k = h + (y, ). Let k = (x
),
there exists a positive integer m such that dom(k) = 1, . . . , m, so t = y = x
m
. Let
= (x
, u
) (k), there exist (h), s #(h, , ) such that = +(y, s). We have
#(k, t, ) = #(k, t, )
(n+1,k,b)
= s = u
m
.
Moreover
V
f
(t) = V
f
(t)
(n+1,k,b)
= y = t; V
b
(t) = V
b
(t)
(n+1,k,b)
= .
Let t E
c
(n +1, k), this means that t E
c
(n+1, k) and k K(n). As a consequence
A dierent approach to logic 67
of t E
c
(n + 1, k):
m positive integer , ,
1
, . . .
m
E(n, k) :
t = ()(
1
, . . . ,
m
), t E(n + 1, k),
(k) ( #(k, , ) is a function with m arguments,
(#(k,
1
, ), . . . , #(k,
m
, )) is a member of the domain of #(k, , ),
#(k, t, ) = #(k, t, )
(n+1,k,c)
= #(k, , )(#(k,
1
, ), . . . , #(k,
m
, )) ),
V
f
(t) = V
f
(t)
(n+1,k,c)
= V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
(t)
(n+1,k,c)
= V
b
() V
b
(
1
) V
b
(
m
),
(k), (k) : _ it results = and obviously #(k, t, ) = #(k, t, ).
Let t E
d
(n +1, k), this means that t E
d
(n+1, k) and k K(n). As a consequence
of t E
d
(n + 1, k):
f T, m positive integer ,
1
, . . .
m
E(n, k) :
t = (f)(
1
, . . . ,
m
), t E(n + 1, k),
(k) ( A
f
(#(k,
1
, ), . . . , #(k,
m
, )),
#(k, t, ) = #(k, t, )
(n+1,k,d)
= P
f
(#(k,
1
, ), . . . , #(k,
m
, )) ),
V
f
(t) = V
f
(t)
(n+1,k,d)
= V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
(t)
(n+1,k,d)
= V
b
(
1
) V
b
(
m
),
(k), (k) : _ it results = and obviously #(k, t, ) = #(k, t, ).
68 M. Avon
Let t E
e
(n +1, k), this means that t E
e
(n+1, k) and k K(n). As a consequence
of t E
e
(n + 1, k):
there exist
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, k, m, x, , ),
t = (x
1
:
1
, . . . , x
m
:
m
, ), t E(n + 1, k),
for each (k)
#(k, t, ) = #(k, t, )
(n+1,k,e)
= #(k
m
, ,
m
)[
m
(k
m
), _
(where k
1
= k + (x
1
,
1
), and if m > 1 for each i = 1 . . . m1
k
i+1
= k
i
+ (x
i+1
,
i+1
) ),
if m = 1
V
f
(t) = V
f
(t)
(n+1,k,e)
= V
f
(
1
) (V
f
() x
1
),
V
b
(t) = V
b
(t)
(n+1,k,e)
= x
1
V
b
(
1
) V
b
(),
if m > 1
V
f
(t) = V
f
(t)
(n+1,k,e)
= V
f
(
1
) (V
f
(
2
) x
1
)
(V
f
(
m
) x
1
, . . . , x
m1
) (V
f
() x
1
, . . . , x
m
),
V
b
(t) = V
b
(t)
(n+1,k,e)
= x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
(),
(k), (k) : _ it results = and obviously #(k, t, ) = #(k, t, ).
Proof of 2.1.9.
Let k = (x, ), h = (y, ) K(n) such that for each i dom(k), j dom(h)
x
i
= y
j
i
=
j
. Let t E(n + 1, k) E(n + 1, h). Let = (x, s) (k),
= (y, r) (h) such that for each i dom(), j dom() x
i
= y
j
s
i
= r
j
.
We need to show that #(k, t, ) = #(h, t, ).
We have proved that assumption 2.1.10 is true at level n + 1, so
t E(n + 1, k) implies that one of ve alternative situations is veried,
t E(n + 1, h) implies that one of ve alternative situations is veried.
A dierent approach to logic 69
Suppose situation a. is the true situation caused by t E(n + 1, k). We have
t (, #(k, t, ) = #(t).
This causes situation a. is also the true situation due to t E(n + 1, h), so
#(h, t, ) = #(t) = #(k, t, ).
The same kind of reasoning applies for the other situations. We now analyze the case
where situation b. is the true situation caused by t E(n + 1, k). We have
i dom(k) : (t = x
i
, #(k, t, ) = s
i
)
j dom(h) : (t = y
j
, #(h, t, ) = r
j
)
Since x
i
= t = y
j
we have #(k, t, ) = s
i
= r
j
= #(h, t, ).
We turn to examine the case where situation c. is the true situation caused by
t E(n + 1, k). We have
K(n) : _ k, m positive integer , ,
1
, . . .
m
E(n, ) :
t = ()(
1
, . . . ,
m
), t E(n + 1, ),
() ( #(, , ) is a function with m arguments,
(#(,
1
, ), . . . , #(,
m
, )) is a member of the domain of #(, , ),
#(, t, ) = #(, , )(#(,
1
, ), . . . , #(,
m
, )) ),
V
f
(t) = V
f
() V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
() V
b
(
1
) V
b
(
m
),
() : _ it results #(k, t, ) = #(, t, ).
g K(n) : g _ h, p positive integer , ,
1
, . . .
p
E(n, g) :
t = ()(
1
, . . . ,
p
), t E(n + 1, g),
(g) ( #(g, , ) is a function with p arguments,
(#(g,
1
, ), . . . , #(g,
p
, )) is a member of the domain of #(g, , ),
#(g, t, ) = #(g, , )(#(g,
1
, ), . . . , #(g,
p
, )) ),
V
f
(t) = V
f
() V
f
(
1
) V
f
(
p
),
V
b
(t) = V
b
() V
b
(
1
) V
b
(
p
),
(g) : _ it results #(h, t, ) = #(g, t, ).
Of course p = m, = , i = 1 . . . m
i
=
i
.
Let =
/dom()
and =
/dom(g)
. By assumption 2.1.11 we get () and
(g). Therefore
#(k, t, ) = #(, t, ) = #(, , )(#(,
1
, ), . . . , #(,
m
, )),
#(h, t, ) = #(g, t, ) = #(g, , )(#(g,
1
, ), . . . , #(g,
p
, )).
70 M. Avon
We have = (x
/dom()
,
/dom()
), g = (y
/dom(g)
,
/dom(g)
). For each i dom(),
j dom(g) if (x
/dom()
)
i
= (y
/dom(g)
)
j
then x
i
= y
j
,
i
=
j
, (
/dom()
)
i
= (
/dom(g)
)
j
.
We have also = (x
/dom()
, s
/dom()
), = (y
/dom(g)
, r
/dom(g)
). For each i dom(),
j dom(g) if (x
/dom()
)
i
= (y
/dom(g)
)
j
then x
i
= y
j
, s
i
= r
j
, (s
/dom()
)
i
= (r
/dom(g)
)
j
.
Moreover ,
1
, . . .
m
E(n, ) E(n, g). By assumption 2.1.9 we get
#(k, t, ) = #(, , )(#(,
1
, ), . . . , #(,
m
, ))
= #(g, , )(#(g,
1
, ), . . . , #(g,
m
, )) = #(h, t, ).
Next we examine the case where situation d. is the true situation caused by
t E(n + 1, k). We have
K(n) : _ k, f T, m positive integer ,
1
, . . .
m
E(n, ) :
t = (f)(
1
, . . . ,
m
), t E(n + 1, ),
() ( A
f
(#(,
1
, ), . . . , #(,
m
, )),
#(, t, ) = P
f
(#(,
1
, ), . . . , #(,
m
, )) ),
V
f
(t) = V
f
(
1
) V
f
(
m
),
V
b
(t) = V
b
(
1
) V
b
(
m
),
() : _ it results #(k, t, ) = #(, t, ).
g K(n) : g _ h, T, p positive integer ,
1
, . . .
p
E(n, g) :
t = ()(
1
, . . . ,
p
), t E(n + 1, g),
(g) ( A
(#(g,
1
, ), . . . , #(g,
p
, )),
#(g, t, ) = P
(#(g,
1
, ), . . . , #(g,
p
, )) ),
V
f
(t) = V
f
(
1
) V
f
(
p
),
V
b
(t) = V
b
(
1
) V
b
(
p
),
(g) : _ it results #(h, t, ) = #(g, t, ).
Of course p = m, = f, i = 1 . . . m
i
=
i
.
Let =
/dom()
and =
/dom(g)
. By assumption 2.1.11 we get () and
(g). Therefore
#(k, t, ) = #(, t, ) = P
f
(#(,
1
, ), . . . , #(,
m
, )),
#(h, t, ) = #(g, t, ) = P
(#(g,
1
, ), . . . , #(g,
p
, )).
We have = (x
/dom()
,
/dom()
), g = (y
/dom(g)
,
/dom(g)
). For each i dom(),
j dom(g) if (x
/dom()
)
i
= (y
/dom(g)
)
j
then x
i
= y
j
,
i
=
j
, (
/dom()
)
i
= (
/dom(g)
)
j
.
We have also = (x
/dom()
, s
/dom()
), = (y
/dom(g)
, r
/dom(g)
). For each i dom(),
j dom(g) if (x
/dom()
)
i
= (y
/dom(g)
)
j
then x
i
= y
j
, s
i
= r
j
, (s
/dom()
)
i
= (r
/dom(g)
)
j
.
A dierent approach to logic 71
Moreover
1
, . . .
m
E(n, ) E(n, g). By assumption 2.1.9 we get
#(k, t, ) = P
f
(#(,
1
, ), . . . , #(,
m
, ))
= P
f
(#(g,
1
, ), . . . , #(g,
m
, )) = #(h, t, ).
We still need to examine the case where situation e. is the true situation caused by
t E(n + 1, k). We have the following.
There exist
K(n) : _ k,
a positive integer m,
a function z whose domain is 1, . . . , m such that for each i = 1 . . . m
z
i
1 var(), and for each i, j = 1 . . . m i ,= j z
i
,= z
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that
c(n, , m, z, , ),
t = (z
1
:
1
, . . . , z
m
:
m
, ), t E(n + 1, ),
for each () #(, t, ) = #(
m
, ,
m
)[
m
(
m
), _
(where
1
= + (z
1
,
1
), and if m > 1 for each i = 1 . . . m1
i+1
=
i
+ (z
i+1
,
i+1
) ),
if m = 1 V
f
(t) = V
f
(
1
) (V
f
() z
1
), V
b
(t) = z
1
V
b
(
1
) V
b
(),
if m > 1
V
f
(t) = V
f
(
1
) (V
f
(
2
) z
1
) (V
f
(
m
) z
1
, . . . , z
m1
)
(V
f
() z
1
, . . . , z
m
),
V
b
(t) = z
1
, . . . , z
m
V
b
(
1
) V
b
(
m
) V
b
(),
() : _ it results #(k, t, ) = #(, t, ).
72 M. Avon
There exist
g K(n) : g _ h,
a positive integer p,
a function u whose domain is 1, . . . , p such that for each i = 1 . . . p
u
i
1 var(g), and for each i, j = 1 . . . p i ,= j u
i
,= u
j
,
a function whose domain is 1, . . . , p such that for each i = 1 . . . p
i
E(n),
E(n)
such that
c(n, g, p, u, , ),
t = (u
1
:
1
, . . . , u
p
:
p
, ), t E(n + 1, g),
for each (g) #(g, t, ) = #(g
m
, ,
m
)[
m
(g
m
), _
(where g
1
= g + (u
1
,
1
), and if p > 1 for each i = 1 . . . p 1
g
i+1
= g
i
+ (u
i+1
,
i+1
) ),
if p = 1 V
f
(t) = V
f
(
1
) (V
f
() u
1
), V
b
(t) = u
1
V
b
(
1
) V
b
(),
if p > 1
V
f
(t) = V
f
(
1
) (V
f
(
2
) u
1
) (V
f
(
p
) u
1
, . . . , u
m1
)
(V
f
() u
1
, . . . , u
m
),
V
b
(t) = u
1
, . . . , u
m
V
b
(
1
) V
b
(
m
) V
b
(),
(g) : _ it results #(h, t, ) = #(g, t, ).
Clearly p = m, u = z, = , = .
Let =
/dom()
and =
/dom(g)
. By assumption 2.1.11 we get () and
(g). Therefore
#(k, t, ) = #(, t, ) = #(
m
, ,
m
)[
m
(
m
), _
m
,
#(h, t, ) = #(g, t, ) = #(g
m
, ,
m
)[
m
(g
m
), _
m
.
We have = (x
/dom()
,
/dom()
), g = (y
/dom(g)
,
/dom(g)
). For each i dom(),
j dom(g) if (x
/dom()
)
i
= (y
/dom(g)
)
j
then x
i
= y
j
,
i
=
j
, (
/dom()
)
i
= (
/dom(g)
)
j
.
We have also = (x
/dom()
, s
/dom()
), = (y
/dom(g)
, r
/dom(g)
). For each i dom(),
j dom(g) if (x
/dom()
)
i
= (y
/dom(g)
)
j
then x
i
= y
j
, s
i
= r
j
, (s
/dom()
)
i
= (r
/dom(g)
)
j
.
By lemma 2.1.15 we get
#(g
m
, ,
m
)[
m
(g
m
), _
m
= #(
m
, ,
m
)[
m
(
m
), _
m
,
and thereferore #(h, t, ) = #(k, t, ).
A dierent approach to logic 73
2.2. Consequences of the denition process. We have nished with denition 2.7.
We now prove a result that is closely related to the denition.
Lemma 2.8. For each positive integer n, k K(n) and t E(n, k)
V
f
(t) var(k) V
b
(t) 1 var(k).
Proof.
We use induction on n.
As for the initial step, we observe that for each t E(1, ) = (
V
f
(t) = = var(); V
b
(t) = 1 = 1 var() .
We now perform the inductive step. Let k K(n + 1) and t E(n + 1, k). We have
seen that
E(n+1, k) = E
(n, k)E
a
(n+1, k)E
b
(n+1, k)E
c
(n+1, k)E
d
(n+1, k)E
e
(n+1, k).
If t E
a
(n +1, k), this means that t E
a
(n + 1, k) and k K(n)
+
. There exist
h K(n), E
s
(n, h), y (1 var(h)) such that k = h + (y, ). We have also
t E(n, h), y / V
b
(t).
Therefore V
f
(t) var(h) var(k).
We have also V
b
(t) 1 var(h), and V
b
(t) / var(h) y = var(k), so
V
b
(t) 1 var(k) .
Let t E
b
(n +1, k), this means that t E
b
(n + 1, k) and k K(n)
+
. There exist
h K(n), E
s
(n, h), y (1 var(h)) such that k = h + (y, ). Moreover t = y,
V
f
(t) = V
f
(t)
(n+1,k,b)
= y var(k) .
V
b
(t) = V
b
(t)
(n+1,k,b)
= 1 var(k) .
Let t E
c
(n +1, k). As a consequence of t E
c
(n + 1, k) there exist ,
1
, . . . ,
m
in E(n, k) such that t = ()(
1
, . . . ,
m
), and
V
f
(t) = V
f
(t)
(n+1,k,c)
= V
f
() V
f
(
1
) V
f
(
m
) var(k) ,
V
b
(t) = V
b
(t)
(n+1,k,c)
= V
b
() V
b
(
1
) V
b
(
m
) 1 var(k) .
Let t E
d
(n +1, k). As a consequence of t E
d
(n + 1, k) there exist f T,
1
, . . . ,
m
in E(n, k) such that t = (f)(
1
, . . . ,
m
), and
V
f
(t) = V
f
(t)
(n+1,k,d)
= V
f
(
1
) V
f
(
m
) var(k) ,
V
b
(t) = V
b
(t)
(n+1,k,d)
= V
b
(
1
) V
b
(
m
) 1 var(k) .
Let t E
e
(n +1, k). As a consequence to t E
e
(n + 1, k) there exist
74 M. Avon
a positive integer m,
a function x whose domain is 1, . . . , m such that for each i = 1 . . . m
x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
,
a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E(n),
E(n)
such that t = (x
1
:
1
, . . . , x
m
:
m
, ) and c(n, k, m, x, , ).
Moreover if m = 1 we have
V
f
(t) = V
f
(
1
) (V
f
() x
1
),
V
b
(t) = x
1
V
b
(
1
) V
b
().
If m > 1 we have
V
f
(t) = V
f
(
1
) (V
f
(
2
) x
1
) (V
f
(
m
) x
1
, . . . , x
m1
)
(V
f
() x
1
, . . . , x
m
),
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
().
Lets consider the case where m = 1.
By the inductive hypothesis V
f
(
1
) var(k) and V
f
() var(k
1
) = var(k) x
1
.
It follows that
V
f
(t) = V
f
(
1
) (V
f
() x
1
) var(k) .
Moreover, the inductive hypothesis lets us state V
b
(
1
) 1 var(k),
V
b
() 1 var(k
1
) = 1 (var(k) x
1
) 1 var(k). Therefore
V
b
(t) = x
1
V
b
(
1
) V
b
() 1 var(k) .
We now turn to examine the case where m > 1.
By the inductive hypothesis V
f
(
1
) var(k) and for each i = 1 . . . m1
V
f
(
i+1
) var(k
i
) = var(k) x
1
, . . . , x
i
, so
V
f
(
i+1
) x
1
, . . . , x
i
var(k) .
Moreover
V
f
() var(k
m
) = var(k) x
1
, . . . , x
m
, so
V
f
() x
1
, . . . , x
m
var(k) .
It follows V
f
(t) var(k).
The inductive hypotesis also implies V
b
(
1
) 1 var(k) and for each i = 1 . . . m1
V
b
(
i+1
) 1 var(k
i
) 1 var(k) .
Moreover V
b
() 1 var(k
m
) 1 var(k) .
Therefore
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
() 1 var(k) .
A dierent approach to logic 75
This result ensures that V
b
(t) and V
f
(t) are always disjoint, so a variable cannot have
both bound and free occurrences in the same expression.
76 M. Avon
3. Introduction to the deductive methodology
In this chapter we will cover some fundamental principles that underlie our inferences.
An important target will be achieved with the proof of theorem 3.6, which is a simple
but signicant step to set up our deductive methodology.
Some preliminary denitions.
Let K =
n1
K(n).
For each k K let
E(k) =
n1:kK(n)
E(n, k) ,
E
s
(k) = t[t E(k), (k) #(k, t, ) is a set .
Let E =
kK
E(k); E is the set of all expressions in our language.
One expression t E(k) is a sentence with respect to k when for each (k)
#(k, t, ) is true or #(k, t, ) is false.
We dene S(k) = t[t E(k), t is a sentence with respect to k.
At the beginning of chapter 2 we have introduced the logical connectives. In our de-
ductions, expressions will make an extensive use of the logical connectives, so we assume
that all of these symbols: , , , , , , are in our set T. For each of these operators
f A
f
(x
1
, . . . , x
n
) and P
f
(x
1
, . . . , x
n
) are dened as specied at the beginning of chapter 2.
For each t E() we dene #(t) = #(, t, ).
On the way to theorem 3.6 we need some further preliminary work, beginning with
the following lemma.
Lemma 3.1. Let h K, E
s
(h), y (1 var(h)), k = h + (y, ). We have k K,
and if S(k) then
(y : , ) E(h);
()((y : , )) S(h), ()((y : , )) S(h);
(h) #(h, ()((y : , )), ) = P
(#(k, , )[ (k), _ );
(h) #(h, ()((y : , )), ) = P
(#(k, , )[ (k), _ ).
Proof.
Since E
s
(h) there is a positive integer n such that E
s
(n, h), h K(n). This
implies that k K(n)
+
K(n + 1) K.
A dierent approach to logic 77
Let S(k). There is a positive integer m such that E(m, k). We dene
p = maxn + 1, m, then we have
h K(p)
y (1 var(h))
E
s
(p, h)
k K(p), E(p, k)
This implies that (y : , ) E
e
(p + 1, h) E(p + 1, h) E(h).
Moreover for each (h)
#(h, (y : , ), ) = #(h, (y : , ), )
/(p+1,h,e)
=
= #(k, , )[ (k), _ .
We want to show that ()((y : , )) E(p + 2, h). To obtain this we just need to
show that for each (h) A
(#(h, (y : , ), )) holds.
Now A
(#(h, (y : , ), )) is equal to
#(h, (y : , ), ) is a set and for each u #(h, (y : , ), ) u is true or u is false.
Clearly #(h, (y : , ), ) is a set, furthermore for each u #(h, (y : , ), )
there is (k) such that _ and u = #(k, , ). Since S(k) u is true or u is
false. So A
(#(h, (y : , ), )) holds.
We have proved that ()((y : , )) E(p + 2, h). Similarly we can show that
()((y : , )) E(p + 2, h). In fact to show this we just need to prove that for each
(h) A
(#(h, (y : , ), )) = A
(#(h, (y : , ), )) .
For each (h)
#(h, ()((y : , )), ) = #(h, ()((y : , )), )
/(p+2,h,d)
=
= P
(#(h, (y : , ), )) =
= P
(#(k, , )[ (k), _ ) .
#(h, ()((y : , )), ) = #(h, ()((y : , )), )
/(p+2,h,d)
=
= P
(#(h, (y : , ), )) =
= P
(#(k, , )[ (k), _ ) .
Finally, as we have seen, for each (h)
#(h, ()((y : , )), ) = P
(#(k, , )[ (k), _ ) ,
and P
1
, . . . ,
m+1
E. We can assume to have dened H[x
1
:
1
, . . . , x
m
:
m
] and if this
holds to have dened also k[x
1
:
1
, . . . , x
m
:
m
] K, such that
var(k[x
1
:
1
, . . . , x
m
:
m
]) = x
1
, . . . , x
m
.
We dene
H[x
1
:
1
, . . . , x
m+1
:
m+1
] = H[x
1
:
1
, . . . , x
m
:
m
]
m+1
E
s
(k[x
1
:
1
, . . . , x
m
:
m
]) .
If H[x
1
:
1
, . . . , x
m+1
:
m+1
] then we dene
k[x
1
:
1
, . . . , x
m+1
:
m+1
] = k[x
1
:
1
, . . . , x
m
:
m
] + (x
m+1
,
m+1
) .
Clearly k[x
1
:
1
, . . . , x
m+1
:
m+1
] K and
var(k[x
1
:
1
, . . . , x
m+1
:
m+1
]) = x
1
, . . . , x
m+1
.
Remark 3.3. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. In these assumptions we can easily
see that for each i = 1 . . . m H[x
1
:
1
, . . . , x
i
:
i
] holds and so k[x
1
:
1
, . . . , x
i
:
i
] is
dened, k[x
1
:
1
, . . . , x
i
:
i
] K, var(k[x
1
:
1
, . . . , x
i
:
i
]) = x
1
, . . . , x
i
.
In fact this is clearly true for i = m. Given i = 2 . . . m, if we suppose this is true for
i, then we have H[x
1
:
1
, . . . , x
i1
:
i1
], and so the remaining facts also hold.
In these assumptions we can dene k
0
= and for each i = 1 . . . m
k
i
= k[x
1
:
1
, . . . , x
i
:
i
]. We have k
0
K, var(k
0
) = , for each i = 1 . . . m k
i
K,
var(k
i
) = x
1
, . . . , x
i
. Hereafter well often use this kind of simplied notation.
We can also easily see that for each i = 1 . . . m
i
E
s
(k
i1
) and k
i
= k
i1
+(x
i
,
i
),
and dom(k
i
) = 1, . . . , i.
Definition 3.4. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for
i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Let be a member of
S(k[x
1
:
1
, . . . , x
m
:
m
]). Dene
[x
m
:
m
, ] = ()((x
m
:
m
, )) .
By lemma 3.1 we have [x
m
:
m
, ] S(k
m1
).
If m > 1 for each i = 2 . . . m suppose we have dened [x
i
:
i
, . . . , x
m
:
m
, ] as a
member of S(k
i1
) and dene
[x
i1
:
i1
, . . . , x
m
:
m
, ] = ()((x
i1
:
i1
, [x
i
:
i
, . . . , x
m
:
m
, ])) .
A dierent approach to logic 79
By lemma 3.1 [x
i1
:
i1
, . . . , x
m
:
m
, ] S(k
i2
).
Lemma 3.5. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Let S(k[x
1
:
1
, . . . , x
m
:
m
]),
m > 1, j = 2 . . . m.
We have [x
j
:
j
, . . . , x
m
:
m
, ] S(k
j1
). We can show that for each i = 1 . . . j1
[x
i
:
i
, . . . , x
m
:
m
, ] = [x
i
:
i
, . . . , x
j1
:
j1
, [x
j
:
j
, . . . , x
m
:
m
, ]] .
Proof.
We show this by induction on i. First we prove the property for i = j 1.
[x
j1
:
j1
, . . . , x
m
:
m
, ] = ()((x
j1
:
j1
, [x
j
:
j
, . . . , x
m
:
m
, ])) =
= [x
j1
:
j1
, [x
j
:
j
, . . . , x
m
:
m
, ]] .
Now we assume j 1 2 and 2 i j 1. We assume the property is true for i
and want to show it holds also for i 1. We have
[x
i1
:
i1
, . . . , x
m
:
m
, ] = ()((x
i1
:
i1
, [x
i
:
i
, . . . , x
m
:
m
, ])) =
= ()((x
i1
:
i1
, [x
i
:
i
, . . . , x
j1
:
j1
, [x
j
:
j
, . . . , x
m
:
m
, ]])) =
= [x
i1
:
i1
, . . . , x
j1
:
j1
, [x
j
:
j
, . . . , x
m
:
m
, ]] .
Theorem 3.6. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Let S(k[x
1
:
1
, . . . , x
m
:
m
]).
Then
#([x
1
:
1
, . . . , x
m
:
m
, ])
P
(#(k[x
1
:
1
, . . . , x
m
:
m
], , )[ (k[x
1
:
1
, . . . , x
m
:
m
]))
Proof.
Well use the symbols k
0
, . . . , k
m
with the meaning specied in remark 3.3, so what
we need to show is:
#([x
1
:
1
, . . . , x
m
:
m
, ]) P
(#(k
m
, , )[ (k
m
)) .
To this end we need to show that for each i = m. . . 1 and for each (k
i1
)
#(k
i1
, [x
i
:
i
, . . . , x
m
:
m
, ], ) P
(#(k
m
, , )[ (k
m
), _ ) .
We prove this by induction on i, starting with the case where i = m. Here we need
to show that for each (k
m1
)
#(k
m1
, [x
m
:
m
, ], ) P
(#(k
m
, , )[ (k
m
), _ ) .
Actually
#(k
m1
, [x
m
:
m
, ], ) = #(k
m1
, ()((x
m
:
m
, )), ) =
= P
(#(k
m
, , )[ (k
m
), _ ) .
80 M. Avon
Now suppose m > 1, let i = 2 . . . m and suppose the property holds for i, we show it
also holds for i 1. We need to prove that for each (k
i2
)
#(k
i2
, [x
i1
:
i1
, . . . , x
m
:
m
, ], ) P
(#(k
m
, , )[ (k
m
), _ ) .
We have
#(k
i2
,[x
i1
:
i1
, . . . , x
m
:
m
, ], ) =
= #(k
i2
, ()((x
i1
:
i1
, [x
i
:
i
, . . . , x
m
:
m
, ])), ) =
= P
(#(k
i1
, [x
i
:
i
, . . . , x
m
:
m
, ], )[ (k
i1
), _ )
P
(P
(#(k
m
, , )[ (k
m
), _ )[ (k
i1
), _ ) .
So it comes to showing that
P
(P
(#(k
m
, , )[ (k
m
), _ )[ (k
i1
), _ )
P
(#(k
m
, , )[ (k
m
), _ ) .
Suppose P
(P
(#(k
m
, , )[ (k
m
), _ )[ (k
i1
), _ ).
This means that for each (k
i1
) such that _ and for each (k
m
) : _
#(k
m
, , ) holds.
Let (k
m
) : _ , we need to prove #(k
m
, , ).
We dene =
/dom(k
i1
)
. By assumption 2.1.11 (k
i1
). Moreover , !()
and dom() = dom(k
i2
) dom(k
i1
) = dom(). By lemma 2.5 we obtain _ .
Therefore #(k
m
, , ) holds.
Conversely suppose P
(#(k
m
, , )[ (k
m
), _ ), so that for each
(k
m
) : _ #(k
m
, , ) is true. Let (k
i1
) be such that _ and let
(k
m
) be such that _ . Since (k
m
) and _ we have #(k
m
, , ).
This completes the proof that for each (k
i2
)
#(k
i2
, [x
i1
:
i1
, . . . , x
m
:
m
, ], ) P
(#(k
m
, , )[ (k
m
), _ ) .
We have also nished the proof that for each i = m. . . 1 and for each (k
i1
)
#(k
i1
, [x
i
:
i
, . . . , x
m
:
m
, ], ) P
(#(k
m
, , )[ (k
m
), _ ) .
It follows that for each (k
0
)
#(k
0
, [x
i
:
i
, . . . , x
m
:
m
, ], ) P
(#(k
m
, , )[ (k
m
), _ ) .
and clearly this can be rewritten
#(, [x
i
:
i
, . . . , x
m
:
m
, ], ) P
(#(k
m
, , )[ (k
m
), _ ) ,
#([x
i
:
i
, . . . , x
m
:
m
, ]) P
(#(k
m
, , )[ (k
m
)) .
Well soon apply theorem 3.6 to show its importance. First we need to prove lemma 3.7,
which is in some way similar to 3.1 but involves the other logical connectives.
A dierent approach to logic 81
Lemma 3.7. Let h K,
1
,
2
S(h). Then
()(
1
,
2
), ()(
1
,
2
), ()(
1
,
2
), ()(
1
,
2
), ()(
1
) S(h);
for each (h) #(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, )) ;
for each (h) #(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, )) ;
for each (h) #(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, )) ;
for each (h) #(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, )) ;
for each (h) #(h, ()(
1
), ) = P
(#(h,
1
, )) .
Proof.
For each (h) #(h,
1
, ) is true or #(h,
1
, ) is false; #(h,
2
, ) is true or
#(h,
2
, ) is false.
We recall that for each (h) A
(#(h,
1
, ), #(h,
2
, )),
A
(#(h,
1
, ), #(h,
2
, )), A
(#(h,
1
, ), #(h,
2
, )), A
(#(h,
1
, ), #(h,
2
, ))
are all dened as
(#(h,
1
, ) is true or #(h,
1
, ) is false) and (#(h,
2
, ) is true or #(h,
2
, ) is false).
Therefore A
(#(h,
1
, ), #(h,
2
, )), A
(#(h,
1
, ), #(h,
2
, )),
A
(#(h,
1
, ), #(h,
2
, )), A
(#(h,
1
, ), #(h,
2
, )) are all true.
And for each (h) A
(#(h,
1
, )) is true.
There exists a positive integer n such that
1
,
2
E(n, h), so
()(
1
,
2
), ()(
1
,
2
), ()(
1
,
2
), ()(
1
,
2
), ()(
1
) E(h) .
Moreover for each (h)
#(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, ));
#(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, ));
#(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, ));
#(h, ()(
1
,
2
), ) = P
(#(h,
1
, ), #(h,
2
, ));
#(h, ()(
1
), ) = P
(#(h,
1
, )) .
so
#(h, ()(
1
,
2
), ) is true or false;
#(h, ()(
1
,
2
), ) is true or false;
#(h, ()(
1
,
2
), ) is true or false;
#(h, ()(
1
,
2
), ) is true or false;
#(h, ()(
1
), ) is true or false .
Therefore we get
()(
1
,
2
), ()(
1
,
2
), ()(
1
,
2
), ()(
1
,
2
), ()(
1
) S(h) .
The following lemma 3.8 is an example of how theorem 3.6 is applied.
82 M. Avon
Lemma 3.8. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let ,
1
,
2
S(k).
Under these assumptions we have ()(,
1
), ()(,
2
), ()(, ()(
1
,
2
)) S(k).
Moreover, if
#([x
1
:
1
, . . . , x
m
:
m
, ()(,
1
)]), #([x
1
:
1
, . . . , x
m
:
m
, ()(,
2
)])
then
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(
1
,
2
))]) .
Proof.
We need to show
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(
1
,
2
))]) ,
that is
P
(P
(P
(#(k, , ), P
(#(k,
1
, ), #(k,
2
, )))[ (k)) . (3.0.1)
But we have
#([x
1
:
1
, . . . , x
m
:
m
, ()(,
1
)]) ,
P
(#(k, ()(,
1
), )[ (k)) ,
P
(P
(#(k, , ), #(k,
1
, ))[ (k)) .
And we have
#([x
1
:
1
, . . . , x
m
:
m
, ()(,
2
)]) ,
P
(#(k, ()(,
2
), )[ (k)) ,
P
(P
(#(k, , ), #(k,
2
, ))[ (k)) .
So for each (k) if #(k, , ) holds true then both #(k,
1
, ) and #(k,
2
, )
hold. This implies 3.0.1 holds true in turn.
According to this lemma, if in our reasoning we have derived the sentences
[x
1
:
1
, . . . , x
m
:
m
, ()(,
1
)] and [x
1
:
1
, . . . , x
m
:
m
, ()(,
2
)], then we
can derive [x
1
:
1
, . . . , x
m
:
m
, ()(, ()(
1
,
2
))]. This is a rst example of how our
deductive methodology will work.
We terminate the chapter with other useful lemmas.
Lemma 3.9. Let c (. For each positive integer n and k K(n) we have
c E(n, k);
for each (k) #(k, c, ) = #(c).
A dierent approach to logic 83
Proof.
The proof is by induction on n.
For n = 1 we have k = so c E(1, ) = E(n, k) and for each (k) = , so
#(k, c, ) = #(, c, ) = #(c).
Let n be a positive integer and k K(n + 1) = K(n) K(n)
+
.
If k K(n) then c E(n, k) E(n + 1, k) and for each (k) #(k, c, ) = #(c).
Otherwise k K(n)
+
, so there exist h K(n), E
s
(n, h), y (1 var(h)) such
that k = h + (y, ). We have c E(n, h) and for each (h) #(h, c, ) = #(c).
It follows that c E(n + 1, k) and for each = + (y, s) (k)
#(k, c, ) = #(h, c, ) = #(c) .
Lemma 3.10. Let k K, m a positive integer, ,
1
, . . . ,
m
E(k). Suppose for each
(k) #(k, , ) is a function with m arguments and (#(k,
1
, ), . . . , #(k,
m
, )) is
a member of its domain. Then
()(
1
, . . . ,
m
) E(k);
for each (k)
#(k, ()(
1
, . . . ,
m
), ) = #(k, , )(#(k,
1
, ), . . . , #(k,
m
, ));
V
b
(()(
1
, . . . ,
m
)) = V
b
() V
b
(
1
) . . . V
b
(
m
);
V
f
(()(
1
, . . . ,
m
)) = V
f
() V
f
(
1
) . . . V
f
(
m
).
Proof.
There exists a positive integer n such that ,
1
, . . . ,
m
E(n, k). This implies that
()(
1
, . . . ,
m
) E(n + 1, k) and for each (k)
#(k, ()(
1
, . . . ,
m
), ) = #(k, , )(#(k,
1
, ), . . . , #(k,
m
, )) .
Clearly the following also hold:
V
b
(()(
1
, . . . ,
m
)) = V
b
() V
b
(
1
) . . . V
b
(
m
);
V
f
(()(
1
, . . . ,
m
)) = V
f
() V
f
(
1
) . . . V
f
(
m
).
Lemma 3.11. Let k K, f T, m a positive integer,
1
, . . . ,
m
E(k). Suppose for
each (k) A
f
(#(k,
1
, ), . . . , #(k,
m
, )) is true. Then
(f)(
1
, . . . ,
m
) E(k);
for each (k) #(k, (f)(
1
, . . . ,
m
), ) = P
f
(#(k,
1
, ), . . . , #(k,
m
, ));
V
b
((f)(
1
, . . . ,
m
)) = V
b
(
1
) . . . V
b
(
m
);
V
f
((f)(
1
, . . . ,
m
)) = V
f
(
1
) . . . V
f
(
m
).
84 M. Avon
Proof.
There exists a positive integer n such that
1
, . . . ,
m
E(n, k). This implies that
(f)(
1
, . . . ,
m
) E(n + 1, k) and for each (k)
#(k, (f)(
1
, . . . ,
m
), ) = P
f
(#(k,
1
, ), . . . , #(k,
m
, )) .
Clearly the following also hold:
V
b
((f)(
1
, . . . ,
m
)) = V
b
(
1
) . . . V
b
(
m
);
V
f
((f)(
1
, . . . ,
m
)) = V
f
(
1
) . . . V
f
(
m
).
Lemma 3.12. Suppose the equality predicate symbol = we dened at the beginning of
chapter 2 belongs to T. Suppose k K,
1
,
2
E(k). Then (=)(
1
,
2
) S(k).
Proof.
For each (k) A
=
(#(k,
1
, ), #(k,
2
, )) is true, so (=)(
1
,
2
) E(k).
Moreover for each (k)
#(k, (=)(
1
,
2
), ) = P
=
(#(k,
1
, ), #(k,
2
, )) =
= #(k,
1
, ) is equal to #(k,
2
, ),
so #(k, (=)(
1
,
2
), ) is true or false.
Therefore (=)(
1
,
2
) S(k).
Lemma 3.13. Suppose the membership predicate symbol we dened at the beginning of
chapter 2 belongs to T. Suppose k K, t, E(k) and for each (k) #(k, , ) is
a set. Then ()(t, ) S(k).
Proof.
For each (k) #(k, , ) is a set, so A
1
, . . . ,
m
E, assume H[x
1
:
1
, . . . , x
m
:
m
], dene k = k[x
1
:
1
, . . . , x
m
:
m
] and
as usual k
0
= and for each i = 1 . . . m k
i
= k[x
1
:
1
, . . . , x
i
:
i
].
Let i = 0 . . . m1 and let E(k
i
) such that for each j = i + 1 . . . m x
j
/ V
b
().
Then E(k) and for each (k) there exists (k
i
) such that _ and
#(k, , ) = #(k
i
, , ).
A dierent approach to logic 85
Proof.
We prove by induction on j that for each j = i . . . m E(k
j
) and for each (k
j
)
there exists (k
i
) such that _ and #(k
j
, , ) = #(k
i
, , ).
The initial step of the proof is obvious, so let j = i . . . m1 and assume E(k
j
) and
for each (k
j
) there exists (k
i
) such that _ and #(k
j
, , ) = #(k
i
, , ).
We have
j+1
E
s
(k
j
) and k
j+1
= k
j
+ (x
j+1
,
j+1
), x
j+1
/ V
b
() so we can apply
lemma 4.12 and obtain that E(k
j+1
) and for each = + (x
i+1
, s) (k
j+1
)
#(k
j+1
, , ) = #(k
j
, , ). Since (k
j
) there exists (k
i
) such that _ _
and #(k
j+1
, , ) = #(k
j
, , ) = #(k
i
, , ).
86 M. Avon
4. Substitution
First-order logic features the notion of substitution (see e.g. Endertons book [2]). Un-
der appropriate assumptions, we can apply substitution to a formula and obtain a new
formula
x
t
, by replacing the free occurrences of the variable x by the term t. In our ap-
proach well dene a similar notion, with the dierence that for us t is a generic expression.
We begin with some preliminary denitions and results, then substitution will be
dened through the complex denition process 4.16.
Definition 4.1. Let n be a positive integer, n > 1. Let k K(n), k ,= . Let p be a
positive integer, x
1
, . . . , x
p
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
p
E.
We dene k
0
= and for each i = 1 . . . p k
i
= k
i1
+ (x
i
,
i
).
We indicate with /(n; k; x
1
:
1
, . . . , x
p
:
p
) the condition in which
k = k
p
and for each i = 1 . . . p k
i1
K(n 1),
i
E
s
(n 1, k
i1
) .
Lemma 4.2. Let n be a positive integer, n > 1. Let k K(n), k ,= . Let p be a positive
integer, x
1
, . . . , x
p
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
p
E.
Suppose /(n; k; x
1
:
1
, . . . , x
p
:
p
) holds.
Then for each i = 1 . . . p dom(k
i
) = 1, . . . , i and if we dene k
i
= (u
i
,
i
) then for each
j = 1 . . . i (u
i
)
j
= x
j
and (
i
)
j
=
j
.
Proof.
We have k
1
= + (x
1
,
1
), so dom(k
1
) = 1, and (u
1
)
1
= x
1
, (
1
)
1
=
1
.
Given i = 1 . . . p 1 we assume dom(k
i
) = 1, . . . , i and for each j = 1 . . . i (u
i
)
j
=
x
j
and (
i
)
j
=
j
. We have k
i+1
= k
i
+ (x
i+1
,
i+1
), so dom(k
i+1
) = 1, . . . , i + 1.
Moreover, for each j = 1 . . . i (u
i+1
)
j
= (u
i
)
j
= x
j
and (
i+1
)
j
= (
i
)
j
=
j
. Finally
(u
i+1
)
i+1
= x
i+1
, (
i+1
)
i+1
=
i+1
.
Lemma 4.3. Let n be a positive integer, n > 1. Let k = (u, ) K(n), k ,= . Let p be a
positive integer, x
1
, . . . , x
p
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
p
E.
Suppose /(n; k; x
1
:
1
, . . . , x
p
:
p
) holds.
Let also q be a positive integer, y
1
, . . . , y
q
1, with y
i
,= y
j
for i ,= j. Let
1
, . . . ,
q
E.
Suppose /(n; k; y
1
:
1
, . . . , y
q
:
q
) holds.
Then q = p, for each i = 1 . . . p y
i
= x
i
,
i
=
i
.
Proof.
We have 1 . . . p = dom(k) = 1 . . . q and therefore q = p. Moreover for each
i = 1 . . . p y
i
= u
i
= x
i
,
i
=
i
=
i
.
A dierent approach to logic 87
Lemma 4.4. Let n be a positive integer, n > 1. Let k K(n), k ,= . Let p be a positive
integer with p < n, x
1
, . . . , x
p
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
p
E.
We dene k
0
= and for each i = 1 . . . p k
i
= k
i1
+ (x
i
,
i
).
Suppose /(n; k; x
1
:
1
, . . . , x
p
:
p
) holds.
Then for each i = 1 . . . p,
i
(k
i
) there exist (k
i1
), s #(k
i1
,
i
, ) such
that
i
= + (x
i
, s).
Proof.
We have k
i
= k
i1
+(x
i
,
i
), k
i1
K(n1),
i
E
s
(n1, k
i1
), x
i
1var(k
i1
).
Therefore k
i
K(n 1)
+
, and there exist (k
i1
), s #(k
i1
,
i
, ) such that
i
= + (x
i
, s).
Lemma 4.5. Let n be a positive integer, n > 1. Let k K(n), k ,= . Let p be a positive
integer with p < n, x
1
, . . . , x
p
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
p
E.
We dene k
0
= and for each i = 1 . . . p k
i
= k
i1
+ (x
i
,
i
).
Suppose /(n; k; x
1
:
1
, . . . , x
p
:
p
) holds.
Let (k), for each i = 0 . . . p dene
i
=
/dom(k
i
)
. Then for each i = 1 . . . p there
exists s
i
#(k
i1
,
i
,
i1
) such that
i
=
i1
+ (x
i
, s
i
).
Proof.
By lemma 2.1.11 we obtain that
i
(k
i
). Then there exist (k
i1
),
s
i
#(k
i1
,
i
, ) such that
i
= + (x
i
, s
i
).
Since !(
i
) we have
= (
i
)
/dom(k
i1
)
= (
/dom(k
i
)
)
/dom(k
i1
)
=
/dom(k
i1
)
=
i1
.
Lemma 4.6. For each positive integer n and k K(n) we have
k = or
( n > 1 and there exist
a positive integer p such that p < n,
x
1
, . . . , x
p
1 such that x
i
,= x
j
for i ,= j,
1
, . . . ,
p
E
such that /(n; k; x
1
:
1
, . . . , x
p
:
p
) ).
Proof.
We prove this by induction on n. The initial step is clearly satised because if k K(1)
then k = .
Then suppose the statement holds for n and lets see it holds also for n + 1.
88 M. Avon
So let k K(n + 1) and k ,= . By assumption 2.1.2
g K(n), z 1 var(g), E
s
(n, g) :
k = g + (z, ) (k) = + (z, s)[ (g), s #(g, , ) .
By the inductive hypothesis
g = or
( n > 1 and there exist
a positive integer p such that p < n,
x
1
, . . . , x
p
1 such that x
i
,= x
j
for i ,= j,
1
, . . . ,
p
E
such that /(n; g; x
1
:
1
, . . . , x
p
:
p
) ).
We rst consider the case where g = .
Here we dene p = 1 < n + 1, x
1
= z 1,
1
= E, and we want to show that
/(n + 1; k; x
1
:
1
).
We assume k
0
= and k
1
= + (x
1
,
1
).
This implies k
1
= k, k
0
K(n),
1
E
s
(n, k
0
), so /(n + 1; k; x
1
:
1
) holds.
We now turn to the case where g ,= and so
( n > 1 and there exist
a positive integer p such that p < n,
x
1
, . . . , x
p
1 such that x
i
,= x
j
for i ,= j,
1
, . . . ,
p
E
such that /(n; g; x
1
:
1
, . . . , x
p
:
p
) ).
In this case p + 1 is a positive integer and p + 1 < n + 1. We dene x
p+1
= z 1,
p+1
= E and need to show that /(n + 1; k; x
1
:
1
, . . . , x
p+1
:
p+1
).
We dene k
0
= and for each i = 1 . . . p + 1 k
i
= k
i1
+ (x
i
,
i
).
Since /(n; g; x
1
:
1
, . . . , x
p
:
p
) we have that g = k
p
and for each i = 1 . . . p
k
i1
K(n 1) K(n),
i
E
s
(n 1, k
i1
) E
s
(n, k
i1
).
To complete our proof we just need to show that k = k
p+1
, k
p
K(n) and
p+1
E
s
(n, k
p
).
We have k = g + (z, ) = k
p
+ (x
p+1
,
p+1
) = k
p+1
; k
p
= g K(n);
p+1
= E
s
(n, k
p
).
A dierent approach to logic 89
Lemma 4.7. Let n be a positive integer such that n 2, let k K(n) such that k ,= .
There exist a positive integer p such that p < n, x
1
, . . . , x
p
1 such that x
i
,= x
j
for
i ,= j,
1
, . . . ,
p
E such that /(n; k; x
1
:
1
, . . . , x
p
:
p
).
Let h K(n 1): h _ k, h ,= . Then there exists a positive integer q p such that
q < n 1, /(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
Moreover let m be a positive integer, E(n 1),
y a function whose domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var(h)
and for each , = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n 1).
Finally suppose that c(n1, h, m, y, , ), and dene h
1
= h+(y
1
,
1
), and if m > 1
for each j = 1 . . . m1 h
j+1
= h
j
+ (y
j+1
,
j+1
).
Then for each j = 1 . . . m /(n 1; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
).
Proof.
By lemma 4.6 we have
( n 1 > 1 and there exist
a positive integer q such that q < n 1,
y
1
, . . . , y
q
1 such that y
i
,= y
j
for i ,= j,
1
, . . . ,
q
E
such that /(n 1; h; y
1
:
1
, . . . , y
q
:
q
) ).
Let k = (u, ). There exists C T such that C dom(k) and h = k
/C
= (u
/C
,
/C
).
By lemma 4.2 we have dom(k) = 1, . . . , p and for each i = 1 . . . p u
i
= x
i
and
i
=
i
.
And by the same lemma C = dom(h) = 1, . . . , q and for each i = 1 . . . q
u
i
= (u
/C
)
i
= y
i
and
i
= (
/C
)
i
=
i
.
Therefore q p and for each i = 1 . . . q y
i
= x
i
and
i
=
i
.
We now turn to the second part of the lemma. We rst need to prove the truth of
/(n; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
).
To simplify we dene for each = 1 . . . q u
= x
= y
q
,
=
q
.
We dene
0
= , for each = 1 . . . q +j
=
1
+ (u
).
We have h =
q
, h
1
= h + (y
1
,
1
) =
q
+ (u
q+1
,
q+1
) =
q+1
.
if j > 1 then for each = 1 . . . j 1
h
+1
= h
+ (y
+1
,
+1
) =
q+
+ (u
q++1
,
q++1
) =
q++1
.
It follows h
j
=
q+j
.
Given = 1 . . . q +j we need to prove
1
K(n 1) and
E
s
(n 1,
1
).
90 M. Avon
For each = 1 . . . q
1
K(n 2) K(n 1),
E
s
(n 2,
1
).
We further have
q
= h K(n 1),
q+1
=
1
E
s
(n 1,
q
).
If j > 1 then for each = q + 2, . . . , q +j
1
= h
1q
K(n 1);
=
q
E
s
(n 1, h
q1
) = E
s
(n 1,
1
) .
With this we have proved /(n; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
).
Now by lemma 4.6, since h
j
K(n 1), h ,= , there exist
a positive integer r such that r < n 1,
v
1
, . . . , v
r
1 such that v
,= v
for ,= ,
1
, . . . ,
p
E
such that /(n 1; h
j
; v
1
:
1
, . . . , v
r
:
r
).
Clearly this implies /(n; h
j
; v
1
:
1
, . . . , v
r
:
r
). By lemma 4.3 we derive that r = q+j,
for each = 1 . . . q v
= x
, for each = q +1 . . . q +j v
= y
q
,
=
q
.
So we obtain that /(n 1; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
).
Lemma 4.8. Let n be a positive integer such that n 2, let k K(n) such that k ,= .
There exist a positive integer p such that p < n, x
1
, . . . , x
p
1 such that x
,= x
for
,= ,
1
, . . . ,
p
E such that /(n; k; x
1
:
1
, . . . , x
p
:
p
).
Let i = 1 . . . p, h K(n) be such that k
i
_ h.
There exist a positive integer q such that q < n, y
1
, . . . , y
q
1 such that y
,= y
for
,= ,
1
, . . . ,
q
E such that /(n; h; y
1
:
1
, . . . , y
q
:
q
).
Then i q and for each j = 1 . . . i y
j
= x
j
,
j
=
j
.
Proof.
Clearly /(n; k
i
; x
1
:
1
, . . . , x
i
:
i
). By lemma 4.2 dom(k
i
) = 1, . . . , i, and if we
dene k
i
= (u, ) then for each j = 1 . . . i u
j
= x
j
,
j
=
j
.
By lemma 4.2 dom(h) = 1, . . . , q, and if we dene h = (v, ) then for each j = 1 . . . q
v
j
= y
j
,
j
=
j
.
Since k
i
_ h there exists C T such that C 1, . . . , q, k
i
= h
/C
.
We have 1, . . . , i = dom(k
i
) = C 1, . . . , q so i q.
Moreover, (u, ) = k
i
= h
/C
= (v
/C
,
/C
), so u = v
/C
, =
/C
, and for each
j = 1 . . . i, y
j
= v
j
= u
j
= x
j
and
j
=
j
=
j
=
j
.
Lemma 4.9. Let p be a positive integer and = (u, r) be a state-like pair whose domain
is 1, . . . , p. We dene
0
= and for each j = 1 . . . p
j
=
j1
+ (u
j
, r
j
). Then it
results
p
= .
A dierent approach to logic 91
Proof.
For each j = 1 . . . p we dene
j
= (v
j
,
j
) and we prove dom(
j
) = 1, . . . , j and
for each i = 1 . . . j (v
j
)
i
= u
i
, (
j
)
i
= r
i
.
We have (v
1
,
1
) =
1
= + (u
1
, r
1
), therefore dom(
1
) = 1 and (v
1
)
1
= u
1
,
(
1
)
1
= r
1
.
Let j = 1 . . . p 1, suppose dom(
j
) = 1, . . . , j and for each i = 1 . . . j (v
j
)
i
= u
i
,
(
j
)
i
= r
i
.
Then (v
j+1
,
j+1
) =
j+1
=
j
+ (u
j+1
, r
j+1
), so dom(
j+1
) = 1, . . . , j + 1, for each
i = 1 . . . j (v
j+1
)
i
= (v
j
)
i
= u
i
, (
j+1
)
i
= (
j
)
i
= r
i
.
To nish, we have also (v
j+1
)
j+1
= u
j+1
and (
j+1
)
j+1
= r
j+1
.
Clearly we have proved dom(
p
) = 1 . . . p = dom(), and for each i = 1 . . . p
(v
p
)
i
= u
i
, (
p
)
i
= r
i
, so
p
= (v
p
,
p
) = (u, r) = .
Lemma 4.10. Let p be a positive integer and = (u, r) be a state-like pair whose domain
is 1, . . . , p. Let m be a positive integer and q be a non-negative integer. Let = (v, c)
be another state-like pair whose domain is 1, . . . , q +m.
We dene
1
= + (v
q+1
, c
q+1
) and if m > 1 for each j = 1 . . . m1
j+1
=
j
+ (v
q+j+1
, c
q+j+1
) .
In these assumptions for each j = 1 . . . m if we set
j
= (u
, r
) then
dom(
j
) = 1, . . . , p +j and for each = 1 . . . p +j
if p then u
= u
, r
= r
;
if > p then u
= v
q+p
, r
= c
q+p
.
Proof.
If we set
1
= (u
, r
) then dom(
1
) = 1, . . . , p + 1 and for each = 1 . . . p + 1
if p then u
= u
, r
= r
;
if > p then u
= u
p+1
= v
q+1
= v
q+p
, r
= r
p+1
= c
q+1
= c
q+p
.
Suppose m > 1, j = 1 . . . m 1. We dene
j
= (u
, r
) and assume
dom(
j
) = 1, . . . , p +j and for each = 1 . . . p +j
if p then u
= u
, r
= r
;
if > p then u
= v
q+p
, r
= c
q+p
.
We then dene
j+1
= (u
, r
). Clearly dom(
j+1
) = 1, . . . , p + j + 1 and for each
= 1 . . . p +j + 1
if p then u
= u
= u
, r
= r
= r
;
if p + 1 p +j then u
= u
= v
q+p
, r
= r
= c
q+p
.
if = p +j + 1 then u
= v
q+j+1
= v
q+p
, r
= c
q+j+1
= c
q+p
.
92 M. Avon
Lemma 4.11. Let h K, y 1 var(h), E
s
(h), k = h + (y, ). Then k K and
for each (k) there exist (h), s #(h, , ) such that = + (y, s).
for each (h), s #(h, , ) + (y, s) (k).
Proof.
There exists a positive integer n such that E
s
(n, h). Clearly also h K(n). So
k K(n)
+
and
(k) = + (y, s)[ (h), s #(h, , ) .
Lemma 4.12. Let h K, y 1 var(h), E
s
(h), k = h + (y, ). Let E(h) such
that y / V
b
(). Then E(k) and for each = +(y, s) (k) #(k, , ) = #(h, , ).
Proof.
There exists a positive integer n such that E
s
(n, h), E(n, h). Of course
h K(n), so k K(n)
+
and E
a
(n + 1, k) E(k).
For each = + (y, s) (k) #(k, , ) = #(h, , ).
Lemma 4.13. Let h K, y 1 var(h), E
s
(h), k = h+(y, ). Then y E(k) and
for each = + (y, s) (k) #(k, y, ) = s.
Proof.
There exists a positive integer n such that E
s
(n, h). Of course h K(n), so
k K(n)
+
.
It follows that y E
b
(n + 1, k) E(k). Moreover, for each = + (y, s) (k)
#(k, y, ) = s.
Definition 4.14. Let k K, m a positive integer, x a function whose domain is
1, . . . , m such that for each i = 1 . . . m x
i
1 var(k), and for each i, j = 1 . . . m
i ,= j x
i
,= x
j
, a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E, and nally let E. We write
c(k, m, x, , )
to indicate the following condition (where k
1
= k + (x
1
,
1
), and if m > 1 for each
i = 1 . . . m1 k
i+1
= k
i
+ (x
i+1
,
i+1
)):
1
E
s
(k) ;
if m > 1 then for each i = 1 . . . m1 k
i
K
i+1
E
s
(k
i
);
k
m
K E(k
m
).
Lemma 4.15. Suppose
k K;
A dierent approach to logic 93
m is a positive integer;
x is a function whose domain is 1, . . . , m such that for each i = 1 . . . m x
i
1 var(k), and for each i, j = 1 . . . m i ,= j x
i
,= x
j
;
is a function whose domain is 1, . . . , m such that for each i = 1 . . . m
i
E;
E;
k
1
= k + (x
1
,
1
), if m > 1 for each i = 1 . . . m1 k
i+1
= k
i
+ (x
i+1
,
i+1
);
c(k, m, x, , ).
Dene t = (x
1
:
1
, . . . , x
m
:
m
, ). Then
t E(k);
for each (k) #(k, t, ) = #(k
m
, ,
m
)[
m
(k
m
), _
m
;
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
().
Proof.
We have
1
E
s
(k) ;
if m > 1 then for each i = 1 . . . m1 k
i
K
i+1
E
s
(k
i
);
k
m
K E(k
m
).
There exist a positive integer n
1
such that
1
E
s
(n
1
, k). If m > 1 then for each
i = 1 . . . m 1 there exists a positive integer n
i+1
such that
i+1
E
s
(n
i+1
, k
i
). There
exists a positive integer n
m+1
such that E(n
m+1
, k
m
).
We dene n = maxn
1
, . . . , n
m+1
, then we have the following:
1
E
s
(n, k) ;
k K(n);
if m > 1 then for each i = 1 . . . m1 k
i
K(n)
i+1
E
s
(n, k
i
);
k
m
K(n) E(n, k
m
).
This implies c(n, k, m, x, , ) and consequently
t E(n + 1, k);
for each (k) #(k, t, ) = #(k
m
, ,
m
)[
m
(k
m
), _
m
;
V
b
(t) = x
1
, . . . , x
m
V
b
(
1
) V
b
(
m
) V
b
().
We are ready to start the big denition process in which we dene substitution. This
is an inductive denition process, so be aware that at step n we may nd that the notion
of kx
i
/t or
k
x
i
/t we are about to dene has already been dened in a former step.
Within the denition there are internal tasks in which we verify some expected condition.
Well use the symbol to mark the end of each of those tasks.
Definition 4.16. Let n be a positive integer such that n 2, let k K(n) such
that k ,= . There exist a positive integer p such that p < n, x
1
, . . . , x
p
1 such
94 M. Avon
that x
i
,= x
j
for i ,= j,
1
, . . . ,
p
E such that /(n; k; x
1
:
1
, . . . , x
p
:
p
). Clearly
p, x
1
, . . . , x
p
,
1
, . . . ,
p
are univocally determined.
Given i = 1 . . . p, t E(k
i1
) such that
for each
i1
(k
i1
) #(k
i1
, t,
i1
) #(k
i1
,
i
,
i1
),
for each j = 1 . . . p: j ,= i x
j
/ V
b
(t),
for each j = i + 1 . . . p V
b
(t) V
b
(
j
) = ;
what we want to do is the following.
If i = p: if kx
i
/t has been dened in a step before the current step n well verify
it is kx
i
/t = k
p1
, otherwise well explicitly dene kx
i
/t = k
p1
.
If i < p we want to verify the following
k
p1
x
i
/t is dened and belongs to K;
x
p
1 var(k
p1
x
i
/t);
(
p
)
k
p1
x
i
/t is dened and belongs to E
s
(k
p1
x
i
/t).
Then if kx
i
/t has been dened in a step before the current step n well verify it
is
kx
i
/t = k
p1
x
i
/t + (x
p
, (
p
)
k
p1
x
i
/t).
Otherwise well explicitly dene
kx
i
/t = k
p1
x
i
/t + (x
p
, (
p
)
k
p1
x
i
/t).
In both cases i = p and i < p well verify
dom(kx
i
/t) = 1, . . . , p 1;
kx
i
/t K;
var(kx
i
/t) = var(k) x
i
;
k
i1
_ kx
i
/t;
if we dene kx
i
/t = (u, ) then for each j = 1 . . . i 1 u
j
= x
j
, for each
j = i . . . p 1 u
j
= x
j+1
;
for each = (u, r) (kx
i
/t) we have that for each j = 1 . . . i 1 u
j
= x
j
,
for each j = i . . . p 1 u
j
= x
j+1
;
for each = (u, r) (kx
i
/t) if we dene
i1
=
/dom(k
i1
)
and dene
0
= and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
);
then
p
(k).
For each E(n, k) with V
b
(t) V
b
() =
Well dene
k
x
i
/t.
Well show that
k
x
i
/t E(kx
i
/t).
Well prove that for each = (u, r) (kx
i
/t), if we dene
i1
=
/dom(k
i1
)
and dene
0
= and for each j = 1 . . . p
A dierent approach to logic 95
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
);
then #(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ) .
Well prove that V
b
(
k
x
i
/t) V
b
() V
b
(t).
Well show that one of the following ve conditions holds
( and
k
x
i
/t = .
var(k), = x
i
k
x
i
/t = t, ,= x
i
k
x
i
/t = .
n > 1, there exist h K(n 1): h _ k, a positive integer m, ,
1
, . . . ,
m
E(n 1, h) such that = ()(
1
, . . . ,
m
), E(n, h),
for each (h) #(h, , ) is a function with m arguments,
(#(h,
1
, ), . . . , #(h,
m
, )) is a member of the domain of #(h, , ).
If h ,= by 4.7 we can derive there exists a positive integer q p such
that q < n 1, /(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have V
b
(t) V
b
()
V
b
(t) V
b
() = , we can dene
h
x
i
/t, and similarly we can dene
(
j
)
h
x
i
/t, and it results
k
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
n > 1, there exist h K(n 1): h _ k, f T, a positive integer m,
1
, . . . ,
m
E(n 1, h) such that = (f)(
1
, . . . ,
m
), E(n, h),
for each (h) A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
If h ,= by 4.7 we can derive there exists a positive integer q p such
that q < n 1, /(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have V
b
(t) V
b
(
j
)
V
b
(t) V
b
() = , we can dene (
j
)
h
x
i
/t, and it results
k
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
n > 1, there exist h K(n1): h _ k, a positive integer m, E(n1),
a function y whose domain is 1, . . . , m such that for each j = 1 . . . m
y
j
1 var(h) and for each , = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n 1);
such that
c(n 1, h, m, y, , ),
= (y
1
:
1
, . . . , y
m
:
m
, ), E(n, h).
If h ,= by 4.7 we can derive there exists a positive integer q p such
that q < n 1, /(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
96 M. Avon
Suppose i q, or in other words x
i
var(h).
We dene h
1
= h + (y
1
,
1
), and if m > 1 for each j = 1 . . . m 1
h
j+1
= h
j
+ (y
j+1
,
j+1
).
We have
1
E(n 1, h), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore
(
1
)
h
x
i
/t is dened;
for each j = 1 . . . m1 h
j
K(n 1) and by 4.7
/(n 1; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
for each = 1 . . . j y
V
b
() so y
/ V
b
(t),
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n 1, h
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
h
j
x
i
/t is dened;
h
m
K(n 1) and by 4.7
/(n 1; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
for each = 1 . . . m y
V
b
() so y
/ V
b
(t),
for each = 1 . . . m V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
E(n 1, h
m
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
h
m
x
i
/t is dened;
it results
k
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
Well prove the following. Given h K(n) such that k
i
_ h we know there
exist a positive integer q such that q < n, y
1
, . . . , y
q
1 such that y
,= y
for ,= ,
1
, . . . ,
q
E such that /(n; h; y
1
:
1
, . . . , y
q
:
q
).
By lemma 4.8 we know that i q and for each j = 1 . . . i y
j
= x
j
,
j
=
j
.
If i < q then assume for each j = i + 1 . . . q y
j
/ V
b
(t), V
b
(t) V
b
(
j
) = .
Also assume E(n, h).
Then
k
x
i
/t =
h
x
i
/t.
Well prove the following. If there exists h K(n) such that E(n, h),
x
i
/ var(h) then
k
x
i
/t = .
Our denition process uses induction on n 2, therefore in the initial step we have
n = 2. If k K(2) and k ,= then there exist x
1
1,
1
E such that /(2; k; x
1
:
1
).
This implies k = + (x
1
,
1
) and
1
E
s
(1, ).
Let t E() be such that #(t) #(
1
). Clearly for each ()
#(, t, ) = #(, t, ) = #(t) #(
1
) = #(,
1
, ) = #(,
1
, ) .
We dene kx
1
/t = . Clearly dom() = , K,
var() = = x
1
x
1
= var(k) x
1
.
It also results k
i1
= _ = kx
i
/t.
A dierent approach to logic 97
Suppose we dene
0
= and
1
=
0
+(x
1
, #(t)) = +(x
1
, #(t)). We have K(1),
1
E
s
(1, ), x
1
(1 var()), so k = + (x
1
,
1
) K(1)
+
. This implies
(k) = + (x
1
, s)[ s #(,
1
, ) .
Now since #(t) #(,
1
, ) we have
1
(k).
Let E(2, k) such that V
b
(t) V
b
() = . Of course
E(2, k) = E
(1, k) E
a
(2, k) E
b
(2, k) E
c
(2, k) E
d
(2, k) E
e
(2, k).
Suppose E
(1, k).
Now suppose E
a
(2, k). This means E
a
(2, k), k K(1)
+
. We have seen
that k = + (x
1
,
1
), where K(1),
1
E
s
(1, ), x
1
(1 var()). It follows that
E(1, ).
We dene
k
x
1
/t = E() = E(kx
1
/t).
Let (kx
1
/t) and dene
0
= and
1
=
0
+ (x
1
, #(t)) = + (x
1
, #(t)). We
have seen that
1
(k). Since E
a
(2, k) we have
#(k, ,
1
) = #(, , ) = #(kx
1
/t,
k
x
1
/t, ) .
Of course V
b
(
k
x
1
/t) = V
b
() V
b
() V
b
(t).
The condition (
k
x
1
/t = is clearly satisifed.
Suppose h K(2) such that k
1
_ h. We know there exist y
1
1,
1
E such that
/(2; h; y
1
:
1
). We know that y
1
= x
1
and
1
=
1
, therefore h = h
1
= k
1
= k, and
clearly
k
x
1
/t =
h
x
1
/t.
Finally suppose there exists h K(2) such that E(2, h), x
1
/ var(h). We have
k
x
1
/t = and this holds independently from the assumption, in fact this is the
denition of
k
x
1
/t.
Lets examine the case where E
b
(2, k). This means E
b
(2, k), k K(1)
+
. We
have seen that k = + (x
1
,
1
), where K(1),
1
E
s
(1, ), x
1
(1 var()). It
results E
b
(2, k) = x
1
, so = x
1
.
We dene
k
x
1
/t = t E() = E(kx
1
/t).
Let (kx
1
/t) and dene
0
= and
1
=
0
+ (x
1
, #(t)) = + (x
1
, #(t)). We
have seen that
1
(k). We have
#(k, ,
1
) = #(k, x
1
,
1
) = #(t) = #(, t, ) = #(kx
1
/t,
k
x
1
/t, ) .
Of course V
b
(
k
x
1
/t) = V
b
(t) V
b
() V
b
(t).
The following conditions hold: var(k), = x
1
,
k
x
1
/t = t. So the following
condition is satised:
var(k), = x
1
k
x
1
/t = t, ,= x
1
k
x
1
/t = .
98 M. Avon
Suppose h K(2) such that k
1
_ h. We know there exist y
1
1,
1
E such that
/(2; h; y
1
:
1
). We know that y
1
= x
1
and
1
=
1
, therefore h = h
1
= k
1
= k, and
clearly
k
x
1
/t =
h
x
1
/t.
Finally suppose there exists h K(2) such that E(2, h), x
1
/ var(h). In this
case, by lemma 2.8 V
f
() var(h). But since E
b
(2, k) we have also V
f
() = x
1
.
It comes out that x
1
var(h), against our assumption. So there doesnt exist h K(2)
such that E(2, h), x
1
/ var(h).
Now assume E
c
(2, k). This implies E
c
(2, k) ,= , so k K(1), k = . This is
against our assumption that k ,= , so we must exclude the case where E
c
(2, k). The
same way we have to exclude the cases where E
d
(2, k) and E
e
(2, k).
Weve seen the only two real cases are E
a
(2, k), E
b
(2, k), and the denition
of
k
x
1
/t depends on which case is veried. Clearly E
a
(2, k) and E
b
(2, k) are disjoint
sets, so the denition we have set out is correct.
This wraps up the initial step of our denition process. To deal with the inductive
step let n 2, suppose we have given our denitions and veried the results at step n,
and lets go on with step n + 1.
Let k K(n + 1) such that k ,= . Let p be a positive integer such that p < n + 1,
x
1
, . . . , x
p
1 such that x
i
,= x
j
for i ,= j,
1
, . . . ,
p
E such that
/(n + 1; k; x
1
:
1
, . . . , x
p
:
p
).
Let i = 1 . . . p, t E(k
i1
) such that
for each
i1
(k
i1
) #(k
i1
, t,
i1
) #(k
i1
,
i
,
i1
),
for each j = 1 . . . p: j ,= i x
j
/ V
b
(t),
for each j = i + 1 . . . p V
b
(t) V
b
(
j
) = .
Consider the case where i = p.
If k K(n) there exist a positive integer q such that q < n, y
1
, . . . , y
q
1 such
that y
,= y
for ,= ,
1
, . . . ,
q
E such that /(n; k; y
1
:
1
, . . . , y
q
:
q
). Clearly
/(n + 1; k; y
1
:
1
, . . . , y
q
:
q
) also holds, so by lemma 4.3 q = p, for each = 1 . . . p
y
= x
and
, /(n; k; x
1
:
1
, . . . , x
p
:
p
).
For this reason, by the inductive hypothesis, kx
i
/t is already dened and we have
kx
i
/t = k
p1
. We have also
dom(kx
i
/t) = 1, . . . , p 1;
kx
i
/t K;
var(kx
i
/t) = var(k) x
i
;
k
i1
_ kx
i
/t;
if we dene kx
i
/t = (u, ) then for each j = 1 . . . i 1 u
j
= x
j
, for each j =
i . . . p 1 u
j
= x
j+1
;
for each = (u, r) (kx
i
/t) we have that for each j = 1 . . . i 1 u
j
= x
j
, for
each j = i . . . p 1 u
j
= x
j+1
;
A dierent approach to logic 99
for each = (u, r) (kx
i
/t) if we dene
i1
=
/dom(k
i1
)
and dene
0
=
and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
);
then
p
(k).
If on the contrary k / K(n) then we dene kx
i
/t = k
p1
.
If p > 1 then we have k
p1
K(n) and k
p1
,= . Using lemma 4.2 we obtain
dom(k
p1
) = 1, . . . , p 1;
if we dene k
p1
= (u
p1
,
p1
) then for each j = 1 . . . p 1 (u
p1
)
j
= x
j
,
(
p1
)
j
=
j
.
Therefore the following hold
dom(kx
i
/t) = 1, . . . , p 1;
kx
i
/t K;
var(kx
i
/t) = var(k) x
i
;
k
i1
= k
p1
_ k
p1
= kx
i
/t;
if we dene kx
i
/t = (u, ) then for each j = 1 . . . i 1 u
j
= x
j
, for each
j = i . . . p 1 u
j
= x
j+1
;
for each = (u, r) (kx
i
/t) we have that for each j = 1 . . . i 1 u
j
= x
j
, for
each j = i . . . p 1 u
j
= x
j+1
.
Moreover let = (u, r) (kx
i
/t), we dene
0
= and for each j = 1 . . . p
if j < p then
j
=
j1
+ (u
j
, r
j
),
p
=
p1
+ (x
p
, #(k
p1
, t, )).
We need to show that
p
(k).
We have that dom() = 1, . . . , p 1 so by lemma 4.9
p1
= ,
p
= + (x
p
, #(k
p1
, t, )) .
It also results k = k
p1
+ (x
p
,
p
), where k
p1
K(n),
p
E
s
(n, k
p1
),
x
p
1var(k
p1
), (k
p1
), #(k
p1
, t, ) #(k
p1
,
p
, ). Therefore we can conrm
that
p
(k).
If p = 1 then kx
i
/t = k
p1
= .
We have
dom(kx
i
/t) = = 1, . . . , p 1;
kx
i
/t = K;
var(kx
i
/t) = = var(k) x
i
;
if we dene kx
i
/t = (u, ) then for each j = 1 . . . 0 u
j
= x
j
, for each j = 1 . . . 0
u
j
= x
j+1
;
100 M. Avon
for each = (u, r) (kx
i
/t) we have that = and for each j = 1 . . . 0 u
j
= x
j
,
for each j = 1 . . . 0 u
j
= x
j+1
;
Moreover let = (u, r) (kx
i
/t) (this implies = ) and dene
0
= and
1
=
0
+ (x
1
, #(, t, )) = + (x
1
, #(, t, )). We need to verify that
1
(k).
We have k = k
1
= + (x
1
,
1
), K(n),
1
E
s
(n, ), x
1
1 var(), (),
#(, t, ) #(,
1
, ). Therefore we can conrm that
1
(k).
We now turn to examine the case where i < p.
If k K(n) there exist a positive integer q such that q < n, y
1
, . . . , y
q
1 such
that y
,= y
for ,= ,
1
, . . . ,
q
E such that /(n; k; y
1
:
1
, . . . , y
q
:
q
). Clearly
/(n + 1; k; y
1
:
1
, . . . , y
q
:
q
) also holds, so by lemma 4.3 q = p, for each = 1 . . . p
y
= x
and
, /(n; k; x
1
:
1
, . . . , x
p
:
p
).
By the inductive hypothesis
k
p1
x
i
/t is dened and belongs to K;
x
p
1 var(k
p1
x
i
/t);
(
p
)
k
p1
x
i
/t is dened and belongs to E
s
(k
p1
x
i
/t).
Moreover, kx
i
/t is already dened and
kx
i
/t = k
p1
x
i
/t + (x
p
, (
p
)
k
p1
x
i
/t).
The inductive hypothesis also ensures that
dom(kx
i
/t) = 1, . . . , p 1;
kx
i
/t K;
var(kx
i
/t) = var(k) x
i
;
k
i1
_ kx
i
/t;
if we dene kx
i
/t = (u, ) then for each j = 1 . . . i 1 u
j
= x
j
, for each j =
i . . . p 1 u
j
= x
j+1
;
for each = (u, r) (kx
i
/t) we have that for each j = 1 . . . i 1 u
j
= x
j
, for
each j = i . . . p 1 u
j
= x
j+1
;
for each = (u, r) (kx
i
/t) if we dene
i1
=
/dom(k
i1
)
and dene
0
=
and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
);
then
p
(k).
If on the contrary k / K(n) then we consider that k
p1
K(n) and k
p1
,= .
Therefore there exist a positive integer q < n, y
1
, . . . , y
q
1 such that y
,= y
for
,= ,
1
, . . . ,
q
E such that /(n; k
p1
; y
1
:
1
, . . . , y
q
:
q
).
We recall that /(n + 1; k; x
1
:
1
, . . . , x
p
:
p
) also holds.
If we dene k
p1
= (u
p1
,
p1
) then lemma 4.2 tells us that
A dierent approach to logic 101
1, . . . , q = dom(k
p1
) = 1, . . . , p 1 and therefore q = p 1;
for each j = 1 . . . p 1 x
j
= (u
p1
)
j
= y
j
,
j
= (
p1
)
j
=
j
;
as a consequence of the former results, /(n; k
p1
; x
1
:
1
, . . . , x
p1
:
p1
).
By the inductive hypothesis k
p1
x
i
/t is dened, it belongs to K and
var(k
p1
x
i
/t) = var(k
p1
) x
i
. Therefore x
p
1 var(k
p1
x
i
/t).
We also consider that
p
E
s
(n, k
p1
) and V
b
(t) V
b
(
p
) = , so (
p
)
k
p1
x
i
/t is
also dened and belongs to E(k
p1
x
i
/t).
We want to show that (
p
)
k
p1
x
i
/t E
s
(k
p1
x
i
/t), so we still need to prove that
for each (k
p1
x
i
/t) #(k
p1
x
i
/t, (
p
)
k
p1
x
i
/t, ) is a set.
Let = (u, r) (k
p1
x
i
/t) and we dene
i1
=
/dom(k
i1
)
and
0
= and for
each j = 1 . . . p 1
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
).
Then #(k
p1
,
p
,
p1
) = #(k
p1
x
i
/t, (
p
)
k
p1
x
i
/t, ) .
Since
p
E
s
(n, k
p1
) we have that #(k
p1
,
p
,
p1
) is a set and
#(k
p1
x
i
/t, (
p
)
k
p1
x
i
/t, ) is a set too.
So we can dene kx
i
/t = k
p1
x
i
/t + (x
p
, (
p
)
k
p1
x
i
/t), and kx
i
/t K.
By the inductive hypothesis dom(k
p1
x
i
/t) = 1, . . . , p 2, so
dom(kx
i
/t) = 1, . . . , p 1. Moreover
var(kx
i
/t) = var(k
p1
x
i
/t) x
p
= (var(k
p1
) x
i
) x
p
= (var(k
p1
) x
p
) x
i
= var(k) x
i
.
Also, clearly, k
i1
_ k
p1
x
i
/t _ kx
i
/t.
We now dene kx
i
/t = (v, ), k
p1
x
i
/t = (u, ). By the inductive hypothesis we
have that for each j = 1 . . . i 1 u
j
= x
j
and for each j = i . . . p 2 u
j
= x
j+1
.
Furthermore for each j = 1 . . . p 2 v
j
= u
j
, v
p1
= x
p
.
So we derive that for each j = 1 . . . i 1 v
j
= u
j
= x
j
; for each j = i . . . p 2
v
j
= u
j
= x
j+1
, and it follows that for each j = i . . . p 1 v
j
= x
j+1
.
Let = (w, s) (kx
i
/t). We have w = v, so for each j = 1 . . . i 1 w
j
= v
j
= x
j
,
for each j = i . . . p 1 w
j
= v
j
= x
j+1
.
Let = (w, s) (kx
i
/t), we dene
i1
=
/dom(k
i1
)
and
0
= and for each
j = 1 . . . p
if j < i then
j
=
j1
+ (w
j
, s
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (w
j1
, s
j1
);
102 M. Avon
We want to show that
p
(k).
Clearly there exist
p1
(k
p1
x
i
/t), c #(k
p1
x
i
/t, (
p
)
k
p1
x
i
/t,
p1
)
such that =
p1
+ (x
p
, c).
We now dene
p1
= (w
, s
). It results
dom(
p1
) = dom(k
p1
x
i
/t) = 1, . . . p 2 .
For each j = 1 . . . p 2 we have w
j
= w
j
, s
j
= s
j
. Therefore for each j = 1 . . . p 1
if j < i then
j
=
j1
+ (w
j
, s
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (w
j1
, s
j1
);
Clearly
i1
=
/dom(k
i1
)
= (w
/dom(k
i1
)
, s
/dom(k
i1
)
) =
= ((w
/{1,...p2}
)
/dom(k
i1
)
, (s
/{1,...p2}
)
/dom(k
i1
)
) =
= ((w
)
/dom(k
i1
)
, (s
)
/dom(k
i1
)
) = (
p1
)
/dom(k
i1
)
.
We can apply the inductive hypothesis and obtain that
p1
(k
p1
), and
#(k
p1
,
p
,
p1
) = #(k
p1
x
i
/t, (
p
)
k
p1
x
i
/t,
p1
) .
To show that
p
(k) we consider that k = k
p
= k
p1
+ (x
p
,
p
), k
p1
K(n),
p
E
s
(n, k
p1
), x
p
1 var(k
p1
). Therefore k K(n)
+
.
Moreover
p
=
p1
+ (w
p1
, s
p1
) =
p1
+ (x
p
, c), and since
p1
(k
p1
),
c #(k
p1
,
p
,
p1
) we have that
p
(k).
In the next step of our denition, for each E(n+1, k) such that V
b
(t) V
b
() =
Well dene
k
x
i
/t.
Well show that
k
x
i
/t E(kx
i
/t).
Well prove that for each = (u, r) (kx
i
/t), if we dene
i1
=
/dom(k
i1
)
and dene
0
= and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
);
then #(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ) .
Well prove that V
b
(
k
x
i
/t) V
b
() V
b
(t).
Remember that
E(n+1, k) = E
(n, k)E
a
(n+1, k)E
b
(n+1, k)E
c
(n+1, k)E
d
(n+1, k)E
e
(n+1, k).
The denition of
k
x
i
/t depends on the set to which belongs to, actually may
belong to more than one of these sets, but this problem will be addressed later when well
show that the denitions match each other.
A dierent approach to logic 103
Suppose E
(n, k). This means E(n, k), k K(n). In this case, by the
inductive hypothesis,
k
x
i
/t has already been already dened at step n and has all the
properties we require at this stage of our denition.
Now suppose E
a
(n +1, k). This implies E
a
(n + 1, k), k K(n)
+
.
We have k = k
p
= k
p1
+(x
p
,
p
), k
p1
K(n),
p
E
s
(n, k
p1
), x
p
1var(k
p1
).
Therefore E(n, k
p1
), x
p
/ V
b
().
If i = p then we dene
k
x
i
/t = E(k
p1
) = E(kx
i
/t).
Let = (u, r) (kx
i
/t), we dene
0
= and for each j = 1 . . . p
if j < p then
j
=
j1
+ (u
j
, r
j
),
p
=
p1
+ (x
p
, #(k
p1
, t, )).
We have already seen that
p
(k).
We need to show that #(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ).
If p = 1 then () so = =
p1
.
If p > 1 we have that dom() = 1, . . . , p 1 so by lemma 4.9
p1
=
In both cases
p
= + (x
p
, #(k
p1
, t, )).
We have (k
p1
), #(k
p1
, t, ) #(k
p1
,
p
, ). Therefore
#(k, ,
p
) = #(k
p1
, , ) = #(kx
i
/t,
k
x
i
/t, ) .
Moreover V
b
(
k
x
i
/t) = V
b
() V
b
() V
b
(t).
If i < p we consider that k
p1
K(n) and k
p1
,= . Therefore there exist a positive
integer q < n, y
1
, . . . , y
q
1 such that y
,= y
for ,= ,
1
, . . . ,
q
E such that
/(n; k
p1
; y
1
:
1
, . . . , y
q
:
q
).
We recall that /(n + 1; k; x
1
:
1
, . . . , x
p
:
p
) also holds.
If we dene k
p1
= (u
p1
,
p1
) then lemma 4.2 tells us that
1, . . . , q = dom(k
p1
) = 1, . . . , p 1 and therefore q = p 1;
for each j = 1 . . . p 1 x
j
= (u
p1
)
j
= y
j
,
j
= (
p1
)
j
=
j
;
as a consequence of the former results, /(n; k
p1
; x
1
:
1
, . . . , x
p1
:
p1
).
Clearly k
p1
x
i
/t is dened; since E(n, k
p1
)
k
p1
x
i
/t is dened too, and it
belongs to E(k
p1
x
i
/t).
So we can dene
k
x
i
/t =
k
p1
x
i
/t E(k
p1
x
i
/t).
We need to show that
k
x
i
/t E(kx
i
/t). We consider that
k
p1
x
i
/t K;
x
p
1 var(k
p1
x
i
/t);
(
p
)
k
p1
x
i
/t E
s
(k
p1
x
i
/t).
kx
i
/t = k
p1
x
i
/t + (x
p
, (
p
)
k
p1
x
i
/t).
104 M. Avon
Moreover we can show that x
p
/ V
b
(
k
p1
x
i
/t). In fact, by the inductive hypothesis,
V
b
(
k
p1
x
i
/t) V
b
() V
b
(t). We know that V
b
() 1 var(k), so x
p
/ V
b
(). We
also know that x
p
/ V
b
(t), hence x
p
/ V
b
(
k
p1
x
i
/t).
Using lemma 4.12 we obtain that
k
x
i
/t =
k
p1
x
i
/t E(kx
i
/t).
Let = (w, s) (kx
i
/t), we dene
i1
=
/dom(k
i1
)
and
0
= and for each
j = 1 . . . p
if j < i then
j
=
j1
+ (w
j
, s
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (w
j1
, s
j1
);
We have proved that
p
(k) and we need to show that
#(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ) .
Clearly there exist
p1
(k
p1
x
i
/t), c #(k
p1
x
i
/t, (
p
)
k
p1
x
i
/t,
p1
)
such that =
p1
+ (x
p
, c).
We have #(kx
i
/t,
k
x
i
/t, ) = #(kx
i
/t,
k
p1
x
i
/t, ).
Since
k
p1
x
i
/t E(k
p1
x
i
/t) and x
p
/ V
b
(
k
p1
x
i
/t), by lemma 4.12, we
obtain #(kx
i
/t,
k
p1
x
i
/t, ) = #(k
p1
x
i
/t,
k
p1
x
i
/t,
p1
), and therefore
#(kx
i
/t,
k
x
i
/t, ) = #(k
p1
x
i
/t,
k
p1
x
i
/t,
p1
) .
We now dene
p1
= (w
, s
). It results
dom(
p1
) = dom(k
p1
x
i
/t) = 1, . . . p 2 .
For each j = 1 . . . p 2 we have w
j
= w
j
, s
j
= s
j
. Therefore for each j = 1 . . . p 1
if j < i then
j
=
j1
+ (w
j
, s
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (w
j1
, s
j1
);
Clearly
i1
=
/dom(k
i1
)
= (w
/dom(k
i1
)
, s
/dom(k
i1
)
) =
= ((w
/{1,...p2}
)
/dom(k
i1
)
, (s
/{1,...p2}
)
/dom(k
i1
)
) =
= ((w
)
/dom(k
i1
)
, (s
)
/dom(k
i1
)
) = (
p1
)
/dom(k
i1
)
.
We can apply the inductive hypothesis and obtain that
p1
(k
p1
), and
#(k
p1
, ,
p1
) = #(k
p1
x
i
/t,
k
p1
x
i
/t,
p1
) .
So far we have proved that
#(kx
i
/t,
k
x
i
/t, ) = #(k
p1
, ,
p1
) .
To complete our proof we need a further step, consisting in proving that
#(k, ,
p
) = #(k
p1
, ,
p1
) .
A dierent approach to logic 105
Here we consider that k
p1
K, x
p
1 var(k
p1
),
p
E
s
(k
p1
),
k = k
p1
+ (x
p
,
p
), E(k
p1
), x
p
/ V
b
(),
p
=
p1
+ (w
p1
, s
p1
) (k).
By lemma 4.11 there exist (k
p1
), d #(k
p1
,
p
, ) such that
p
= +(x
p
, d).
Clearly =
p1
, x
p
= w
p1
, d = s
p1
.
By lemma 4.12 we have #(k, ,
p
) = #(k
p1
, ,
p1
).
Finally it results V
b
(
k
x
i
/t) = V
b
(
k
p1
x
i
/t) V
b
() V
b
(t).
Now suppose E
b
(n +1, k). This implies E
b
(n + 1, k), k K(n)
+
.
We have k = k
p
= k
p1
+(x
p
,
p
), k
p1
K(n),
p
E
s
(n, k
p1
), x
p
1var(k
p1
).
Therefore = x
p
.
If i = p we dene
k
x
i
/t = t E(k
p1
) = E(kx
i
/t).
Let = (u, r) (kx
i
/t), we dene
0
= and for each j = 1 . . . p
if j < p then
j
=
j1
+ (u
j
, r
j
),
p
=
p1
+ (x
p
, #(k
p1
, t, )).
We have already seen that
p
(k).
We need to show that #(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ).
Clearly #(kx
i
/t,
k
x
i
/t, ) = #(k
p1
, t, ), so what we need to show is
#(k, x
p
,
p
) = #(k
p1
, t, ) .
There exist (k
p1
), s #(k
p1
,
p
, ) such that
p
= +(x
p
, s). By lemma 4.13
it results #(k, x
p
,
p
) = s.
Since
p
=
p1
+(x
p
, #(k
p1
, t, )) we have =
p1
and s = #(k
p1
, t, ). Therefore
#(k, x
p
,
p
) = #(k
p1
, t, ).
Moreover V
b
(
k
x
i
/t) = V
b
(t) V
b
() V
b
(t).
If i < p we dene
k
x
i
/t = = x
p
.
We need to show that
k
x
i
/t E(kx
i
/t). We consider that
k
p1
x
i
/t K;
x
p
1 var(k
p1
x
i
/t);
(
p
)
k
p1
x
i
/t E
s
(k
p1
x
i
/t).
kx
i
/t = k
p1
x
i
/t + (x
p
, (
p
)
k
p1
x
i
/t).
By lemma 4.13 we have
k
x
i
/t = x
p
E(kx
i
/t).
Let = (w, s) (kx
i
/t), we dene
i1
=
/dom(k
i1
)
and
0
= and for each
j = 1 . . . p
if j < i then
j
=
j1
+ (w
j
, s
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (w
j1
, s
j1
);
106 M. Avon
We have proved that
p
(k) and we need to show that
#(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ) .
By lemma 4.11 there exist (k
p1
), d #(k
p1
,
p
, ) such that
p
= +(x
p
, d).
Clearly =
p1
, x
p
= w
p1
, d = s
p1
.
Using lemma 4.13 we obtain
#(k, ,
p
) = #(k, x
p
,
p
) = s
p1
.
Clearly there exist
p1
(k
p1
x
i
/t), c #(k
p1
x
i
/t, (
p
)
k
p1
x
i
/t,
p1
)
such that =
p1
+ (x
p
, c). And clearly c = s
p1
.
Using lemma 4.13 we obtain
#(kx
i
/t,
k
x
i
/t, ) = #(kx
i
/t, x
p
, ) = s
p1
.
So we can derive #(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ).
To nish with the current case we see that V
b
(
k
x
i
/t) = V
b
() V
b
() V
b
(t).
We turn to the case where E
c
(n +1, k). This implies E
c
(n+1, k), k K(n).
There exist a positive integer m and ,
1
, . . . ,
m
E(n, k) such that
= ()(
1
, . . . ,
m
);
for each (k) #(k, , ) is a function with m arguments and
(#(k,
1
, ), . . . , #(k,
m
, )) is a member of its domain.
Since k K(n) there exist a positive integer q such that q < n, y
1
, . . . , y
q
1 such
that y
,= y
for ,= ,
1
, . . . ,
q
E such that /(n; k; y
1
:
1
, . . . , y
q
:
q
). Clearly
/(n + 1; k; y
1
:
1
, . . . , y
q
:
q
) also holds, so by lemma 4.3 q = p, for each = 1 . . . p
y
= x
and
, /(n; k; x
1
:
1
, . . . , x
p
:
p
).
We have V
b
() = V
b
() V
b
(
1
) V
b
(
m
) and since V
b
(t) V
b
() = we have
V
b
(t) V
b
() = ;
for each j = 1 . . . m V
b
(t) V
b
(
j
) = .
By the inductive hypothesis
k
x
i
/t is dened and belongs to E(kx
i
/t) and for
each j = 1 . . . m (
j
)
k
x
i
/t is dened and belongs to E(kx
i
/t). So we can dene
k
x
i
/t = (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) .
We need to show that
k
x
i
/t E(kx
i
/t). To show this we use lemma 3.10.
Let = (u, r) (kx
i
/t), we just need to show that #(kx
i
/t,
k
x
i
/t, ) is a
function with m arguments and
(#(kx
i
/t, (
1
)
k
x
i
/t, ), . . . , #(kx
i
/t, (
m
)
k
x
i
/t, )) is a member of its domain.
We dene
i1
=
/dom(k
i1
)
and dene
0
= , for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
A dierent approach to logic 107
if j > i then
j
=
j1
+ (u
j1
, r
j1
).
By the inductive hypothesis
p
(k), #(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ), for
each j = 1 . . . m #(k,
j
,
p
) = #(kx
i
/t, (
j
)
k
x
i
/t, ).
Therefore #(kx
i
/t,
k
x
i
/t, ) = #(k, ,
p
) is a function with m arguments and
(#(kx
i
/t, (
1
)
k
x
i
/t, ), . . . , #(kx
i
/t, (
m
)
k
x
i
/t, )) is equal to
(#(k,
1
,
p
), . . . , #(k,
m
,
p
)) and so is a member of the domain of
#(kx
i
/t,
k
x
i
/t, ).
We have proved that
k
x
i
/t E(kx
i
/t).
Moreover for each = (u, r) (kx
i
/t), if we dene
i1
=
/dom(k
i1
)
and dene
0
= and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
);
then
p
(k) and
#(k, ,
p
) = #(k, ()(
1
, . . . ,
m
),
p
) = #(k, ,
p
)(#(k,
1
,
p
), . . . , #(k,
m
,
p
)) =
= #(k{x
i
/t},
k
{x
i
/t}, )(#(k{x
i
/t}, (
1
)
k
{x
i
/t}, ), . . . , #(k{x
i
/t}, (
m
)
k
{x
i
/t}, )) =
= #(k{x
i
/t}, (
k
{x
i
/t})((
1
)
k
{x
i
/t}, . . . , (
m
)
k
{x
i
/t}), ) =
= #(k{x
i
/t},
k
{x
i
/t}, ) .
Finally
V
b
(
k
x
i
/t) = V
b
((
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t)) =
= V
b
((
k
x
i
/t)) V
b
((
1
)
k
x
i
/t) V
b
((
m
)
k
x
i
/t)
V
b
() V
b
(
1
) V
b
(
m
) V
b
(t) = V
b
() V
b
(t) .
We examine the case where E
d
(n +1, k). This implies E
d
(n+1, k), k K(n).
There exist f T, a positive integer m and
1
, . . . ,
m
E(n, k) such that
= (f)(
1
, . . . ,
m
);
for each (k) A
f
(#(k,
1
, ), . . . , #(k,
m
, )) holds true.
Since k K(n) we have /(n; k; x
1
:
1
, . . . , x
p
:
p
).
It results V
b
() = V
b
(
1
) V
b
(
m
) and since V
b
(t) V
b
() = we have
for each j = 1 . . . m V
b
(t) V
b
(
j
) = .
By the inductive hypothesis for each j = 1 . . . m (
j
)
k
x
i
/t is dened and belongs
to E(kx
i
/t). So we can dene
k
x
i
/t = (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) .
We need to show that
k
x
i
/t E(kx
i
/t). To show this we use lemma 3.11.
We have kx
i
/t K, f T, for each j = 1 . . . m (
j
)
k
x
i
/t E(kx
i
/t).
108 M. Avon
Given = (u, r) (kx
i
/t) we need to show that
A
f
(#(kx
i
/t, (
1
)
k
x
i
/t, ), . . . , #(kx
i
/t, (
1
)
k
x
i
/t, )) holds true.
We dene
i1
=
/dom(k
i1
)
and dene
0
= , for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
).
By the inductive hypothesis
p
(k) and for each j = 1 . . . m
#(k,
j
,
p
) = #(kx
i
/t, (
j
)
k
x
i
/t, ).
We have seen A
f
(#(k,
1
,
p
), . . . , #(k,
j
,
p
)) holds true, so
A
f
(#(kx
i
/t, (
1
)
k
x
i
/t, ), . . . , #(kx
i
/t, (
1
)
k
x
i
/t, )) also holds.
Consequently, we have proved that
k
x
i
/t E(kx
i
/t).
Moreover for each = (u, r) (kx
i
/t), if we dene
i1
=
/dom(k
i1
)
and dene
0
= and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
);
then
p
(k) and
#(k, ,
p
) = #(k, (f)(
1
, . . . ,
m
),
p
) = P
f
(#(k,
1
,
p
), . . . , #(k,
m
,
p
)) =
= P
f
(#(kx
i
/t, (
1
)
k
x
i
/t, ), . . . , #(kx
i
/t, (
m
)
k
x
i
/t, )) =
= #(kx
i
/t, (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t), ) =
= #(kx
i
/t,
k
x
i
/t, ) .
Furthermore
V
b
(
k
x
i
/t) = V
b
((f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t)) =
= V
b
((
1
)
k
x
i
/t) V
b
((
m
)
k
x
i
/t)
V
b
(
1
) V
b
(
m
) V
b
(t) = V
b
() V
b
(t) .
Finally lets consider the case where E
e
(n +1, k). This implies E
e
(n +1, k),
k K(n). There exist
a positive integer m,
a function y whose domain is 1, . . . , m such that for each j = 1 . . . m
y
j
1 var(k), and for each , = 1 . . . m ,= y
,= y
,
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n),
E(n)
such that = (y
1
:
1
, . . . , y
m
:
m
, ) and c(n, k, m, y, , ).
Let k
1
= k + (y
1
,
1
), and if m > 1 for each j = 1 . . . m1 k
j+1
= k
j
+ (y
j+1
,
j+1
).
We intend to dene
k
x
i
/t as follows.
A dierent approach to logic 109
If m = 1
k
x
i
/t = (y
1
: (
1
)
k
x
i
/t,
k
1
x
i
/t);
if m > 1
k
x
i
/t = (y
1
: (
1
)
k
x
i
/t, y
2
: (
2
)
k
1
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) .
To accept this denition we need to verify it relies on well dened concepts. In other
words we have to verify that (
1
)
k
x
i
/t is dened, if m > 1 then for each j = 2 . . . m
(
j
)
k
j1
x
i
/t is dened, and nally that
k
m
x
i
/t is dened.
Since k K(n) we have /(n; k; x
1
:
1
, . . . , x
p
:
p
).
It results
1
E(n, k) and since V
b
(
1
) V
b
() we have V
b
(
1
) V
b
(t) = . This
ensures (
1
)
k
x
i
/t is dened and belongs to E(kx
i
/t).
Suppose m > 1 and let j = 2 . . . m, we want to verify that (
j
)
k
j1
x
i
/t is dened.
We have k
j1
K(n) and /(n; k
j1
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
, . . . , y
j1
:
j1
) follows
by lemma 4.7. For each = 1 . . . j 1 y
V
b
() so y
/ V
b
(t) and V
b
(
) V
b
() so
V
b
(
)V
b
(t) = . We have
j
E(n, k
j1
) and also V
b
(
j
) V
b
() so V
b
(
j
)V
b
(t) = .
Therefore (
j
)
k
j1
x
i
/t is dened and belongs to E(k
j1
x
i
/t).
To verify that
k
m
x
i
/t is dened we consider that k
m
K(n) and by lemma 4.7
/(n; k
m
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
, . . . , y
m
:
m
). For each = 1 . . . m y
V
b
()
so y
/ V
b
(t) and V
b
(
) V
b
() so V
b
(
) V
b
(t) = . We have E(n, k
m
) and
also V
b
() V
b
() so V
b
() V
b
(t) = . Therefore
k
m
x
i
/t is dened and belongs to
E(k
m
x
i
/t).
At this point we accept the proposed denition of
k
x
i
/t, but we also need to prove
that
k
x
i
/t E(kx
i
/t).
We dene h = kx
i
/t; a function whose domain is 1, . . . , m such that
1
= (
1
)
k
x
i
/t, if m > 1 for each j = 2 . . . m
j
= (
j
)
k
j1
x
i
/t; =
k
m
x
i
/t.
With these denitions clearly
k
x
i
/t = (y
1
:
1
, . . . , y
m
:
m
, ) .
We should be able to apply lemma 4.15. We have var(h) = var(k) x
i
var(k) and
so 1 var(k) 1 var(h). Moreover
h K;
m is a positive integer;
y is a function whose domain is 1, . . . , m such that for each j = 1 . . . m
y
j
1 var(h), and for each , = 1 . . . m ,= y
,= y
;
is a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E;
E.
We then dene h
1
= h + (y
1
,
1
), and if m > 1 for each j = 1 . . . m 1
h
j+1
= h
j
+ (y
j+1
,
j+1
). Clearly we need to prove c(h, m, y, , ).
We rst verify that for each j = 1 . . . m h
j
= k
j
x
i
/t.
110 M. Avon
We have /(n; k
1
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
) so k
1
= (k
1
)
p
+ (y
1
,
1
),
but k
1
= k + (y
1
,
1
) also holds so k = (k
1
)
p
. Consequently
k
1
x
i
/t = (k
1
)
p
x
i
/t + (y
1
, (
1
)
(k
1
)
p
x
i
/t) = kx
i
/t + (y
1
, (
1
)
k
x
i
/t) = h
1
.
Now suppose m > 1 and let j = 1 . . . m1.
We have /(n; k
j+1
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
, . . . , y
j+1
:
j+1
),
so k
j+1
= (k
j+1
)
p+j
+ (y
j+1
,
j+1
), but k
j+1
= k
j
+ (y
j+1
,
j+1
) also holds so
(k
j+1
)
p+j
= k
j
. It follows
k
j+1
x
i
/t = (k
j+1
)
p+j
x
i
/t + (y
j+1
, (
j+1
)
(k
j+1
)
p+j
x
i
/t)
= k
j
x
i
/t + (y
j+1
, (
j+1
)
k
j
x
i
/t) = h
j
+ (y
j+1
,
j+1
) = h
j+1
.
Proving c(h, m, y, , ) means showing that
1
E
s
(h) ;
if m > 1 then for each j = 1 . . . m1 h
j
K
j+1
E
s
(h
j
);
h
m
K E(h
m
).
We begin by proving that
1
E
s
(h), in other words (
1
)
k
x
i
/t E
s
(kx
i
/t). Let
= (u, r) (kx
i
/t), we dene
i1
=
/dom(k
i1
)
,
0
= and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
).
Then
p
(k) and #(k,
1
,
p
) = #(kx
i
/t, (
1
)
k
x
i
/t, ). Since c(n, k, m, y, , )
it results
1
E
s
(n, k) so #(k,
1
,
p
) is a set and so is #(kx
i
/t, (
1
)
k
x
i
/t, ).
Suppose m > 1 and j = 1 . . . m1, we need to verify that h
j
K and
j+1
E
s
(h
j
).
In other words we need to verify k
j
x
i
/t K and (
j+1
)
k
j
x
i
/t E
s
(k
j
x
i
/t).
Clearly k
j
x
i
/t belongs to K, and we have veried that (
j+1
)
k
j
x
i
/t E(k
j
x
i
/t).
Let = (u, r) (k
j
x
i
/t), wed like to verify that #(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t, )
is a set. We dene
i1
=
/dom(k
i1
)
,
0
= and for each = 1 . . . p +j
if < i then
=
1
+ (u
, r
),
if = i then
=
1
+ (x
i
, #(k
i1
, t,
i1
)),
if > i then
=
1
+ (u
1
, r
1
).
Then
p+j
(k
j
) and #(k
j
,
j+1
,
p+j
) = #(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t, ).
Since c(n, k, m, y, , ) it results
j+1
E
s
(n, k
j
) so #(k
j
,
j+1
,
p+j
) is a set and so is
#(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t, ).
Finally we need to verify that h
m
K E(h
m
). In other words we need to verify
k
m
x
i
/t K and
k
m
x
i
/t E(k
m
x
i
/t). This has been proved above.
At this point c(h, m, y, , ) is proved so by lemma 4.15 we obtain
k
x
i
/t = (y
1
:
1
, . . . , y
m
:
m
, ) E(h) = E(kx
i
/t) .
By lemma 4.15 we also obtain
V
b
(
k
x
i
/t) = y
1
, . . . , y
m
V
b
(
1
) V
b
(
m
) V
b
() . (4.0.1)
A dierent approach to logic 111
This will be used later.
Another point we have to verify is the following. Let = (u, r) (kx
i
/t), we
dene
i1
=
/dom(k
i1
)
,
0
= and for each j = 1 . . . p
if j < i then
j
=
j1
+ (u
j
, r
j
),
if j = i then
j
=
j1
+ (x
i
, #(k
i1
, t,
i1
)),
if j > i then
j
=
j1
+ (u
j1
, r
j1
).
It has been shown that
p
(k), we need to prove #(k, ,
p
) = #(kx
i
/t,
k
x
i
/t, ).
Of course we have #(k, ,
p
) = #(k
m
, ,
m
)[
m
(k
m
),
p
_
m
, and using
lemma 4.15 we derive
#(kx
i
/t,
k
x
i
/t, ) = #(h,
k
x
i
/t, ) = #(h
m
, ,
m
)[
m
(h
m
), _
m
=
= #(k
m
x
i
/t,
k
m
x
i
/t,
m
)[
m
(k
m
x
i
/t), _
m
.
So what we need to show is
{#(k
m
, ,
m
)|
m
(k
m
),
p
m
} = {#(k
m
{x
i
/t},
k
m
{x
i
/t},
m
)|
m
(k
m
{x
i
/t}),
m
} .
Suppose w #(k
m
, ,
m
)[
m
(k
m
),
p
_
m
.
There exists
m
= (v, c) (k
m
) such that
p
_
m
and w = #(k
m
, ,
m
).
Clearly dom(
m
) = dom(k
m
) = 1, . . . , p +m.
We dene
1
= + (v
p+1
, c
p+1
) and if m > 1 for each j = 1 . . . m1
j+1
=
j
+ (v
p+j+1
, c
p+j+1
). Our goal is to show that
m
(k
m
x
i
/t) and #(k
m
x
i
/t,
k
m
x
i
/t,
m
) = #(k
m
, ,
m
) .
First of all we dene
0
= and for each j = 1 . . . p + m
j
=
j1
+ (v
j
, c
j
). By
lemma 4.9 we derive that
p+m
=
m
. Therefore
p
_
p+m
.
There exists C T such that C 1, . . . p +m,
p
= (
p+m
)
/C
.
We have C = dom(
p
) = 1, . . . , p, and therefore
p
= (
p+m
)
/{1,...,p}
= (v
/{1,...,p}
, c
/{1,...,p}
) =
p
.
We dene k
0
= k. We use backward induction on j to show that for each j = m. . . 1
p+j
(k
j
) and
p+j1
(k
j1
), v
p+j
= y
j
, c
p+j
#(k
j1
,
j
,
p+j1
) .
Clearly
p+m
=
m
(k
m
).
We have k
m
= k
m1
+(y
m
,
m
), k
m1
K(n), y
m
1 var(k
m1
),
m
E
s
(n, k
m1
).
By lemma 4.11 this implies there exist (k
m1
), s #(k
m1
,
m
, ) such that
p+m1
+ (v
p+m
, c
p+m
) =
p+m
= + (y
m
, s) .
By lemma 2.1 we obtain
p+m1
= (k
m1
), v
p+m
= y
m
, c
p+m
= s #(k
m1
,
m
,
p+m1
) .
112 M. Avon
If m > 1 we need an inductive step. Let j = m. . . 2 and assume
p+j
(k
j
) and
p+j1
(k
j1
), v
p+j
= y
j
, c
p+j
#(k
j1
,
j
,
p+j1
) .
We need to prove
p+j2
(k
j2
), v
p+j1
= y
j1
, c
p+j1
#(k
j2
,
j1
,
p+j2
) .
We have k
j1
= k
j2
+ (y
j1
,
j1
), k
j2
K(n), y
j1
1 var(k
j2
),
j1
E
s
(n, k
j2
).
By lemma 4.11 this implies there exist (k
j2
), s #(k
j2
,
j1
, ) such that
p+j2
+ (v
p+j1
, c
p+j1
) =
p+j1
= + (y
j1
, s) .
By lemma 2.1 we obtain
p+j2
= (k
j2
), v
p+j1
= y
j1
, c
p+j1
= s #(k
j2
,
j1
,
p+j2
) .
To show that
m
(k
m
x
i
/t) we show by induction on j that for each j = 1 . . . m
j
(h
j
).
We begin by showing that
1
(h
1
) using lemma 4.11. We have h
1
= h + (y
1
,
1
)
and h K, y
1
1 var(h),
1
E
s
(h). We have (h) and
1
= + (y
1
, c
p+1
).
To show
1
(h
1
) we just need to show that c
p+1
#(h,
1
, ).
In other words we have to prove c
p+1
#(kx
i
/t, (
1
)
k
x
i
/t, ).
But we have proved that c
p+1
#(k,
1
,
p
) = #(k,
1
,
p
), and since we have been
able to dene (
1
)
k
x
i
/t we can assume #(k,
1
,
p
) = #(kx
i
/t, (
1
)
k
x
i
/t, ).
So c
p+1
#(h,
1
, ) and
1
(h
1
) are proved.
Suppose m > 1 and let j = 1 . . . m 1. We assume
j
(h
j
) and try to show
j+1
(h
j+1
), using lemma 4.11. We have h
j+1
= h
j
+ (y
j+1
,
j+1
) and
h
j
K, y
j+1
1 var(h
j
),
j+1
E
s
(h
j
).
We have also
j+1
=
j
+ (v
p+j+1
, c
p+j+1
) =
j
+ (y
j+1
, c
p+j+1
).
To show
j+1
(h
j+1
) we just need to show that c
p+j+1
#(h
j
,
j+1
,
j
).
In other words we have to prove c
p+j+1
#(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t,
j
).
We have proved that c
p+j+1
#(k
j
,
j+1
,
p+j
).
We dene two functions z, q over 1, . . . , p as follows: for each = 1 . . . p
if < i then z
= u
, q
= r
;
if = i then z
= x
i
, q
= #(k
i1
, t,
i1
);
if > i then z
= u
1
, q
= r
1
.
Clearly we have
0
= and for each = 1 . . . p
=
1
+ (z
, q
). Therefore by
lemma 4.9 we have
p
= (z, q).
Since (v
/{1,...,p}
, c
/{1,...,p}
) =
p
=
p
= (z, q) we have that for each = 1 . . . p v
= z
and c
= q
.
A dierent approach to logic 113
Moreover we consider that = (u, r) is a state-like pair whose domain is 1, . . . , p1
and
m
= (v, c) is a state-like pair whose domain is 1, . . . , p +m.
We have dened
1
= + (v
p+1
, c
p+1
) and if m > 1 for each = 1 . . . m1
+1
=
+ (v
p++1
, c
p++1
).
We dene
j
= (u
, r
j
) = 1, . . . , p1+j
and for each = 1 . . . p 1 +j
if p 1 then u
= u
, r
= r
;
if > p 1 then u
= v
+1
, r
= c
+1
.
For each = 1 . . . p +j
if < i then
=
1
+ (v
, c
) =
1
+ (u
, r
) =
1
+ (u
, r
);
if = i then
=
1
+ (v
, c
) =
1
+ (x
i
, #(k
i1
, t,
i1
));
if i < p then
=
1
+(v
, c
) =
1
+(u
1
, r
1
) =
1
+(u
1
, r
1
);
if > p then
=
1
+ (v
, c
) =
1
+ (u
1
, r
1
).
Also consider that
i1
=
/dom(k
i1
)
= ((
j
)
/dom()
)
/dom(k
i1
)
= (
j
)
/dom(k
i1
)
.
Since we have been able to dene k
j
x
i
/t and (
j+1
)
k
j
x
i
/t, we must have
#(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t,
j
) = #(k
j
,
j+1
,
p+j
) .
At this point we have proved c
p+j+1
#(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t,
j
), and so also
j+1
(h
j+1
) is proved. This also completes the proof of
m
(k
m
x
i
/t).
We still need to prove #(k
m
x
i
/t,
k
m
x
i
/t,
m
) = #(k
m
, ,
m
).
We consider that = (u, r) is a state-like pair whose domain is 1, . . . , p 1 and
m
= (v, c) is a state-like pair whose domain is 1, . . . , p +m.
We have dened
1
= + (v
p+1
, c
p+1
) and if m > 1 for each = 1 . . . m1
+1
=
+ (v
p++1
, c
p++1
).
We dene
m
= (u
, r
m
) = 1, . . . , p 1 +m and for each = 1 . . . p 1 +m
if p 1 then u
= u
, r
= r
;
if > p 1 then u
= v
+1
, r
= c
+1
.
For each = 1 . . . p +m
if < i then
=
1
+ (v
, c
) =
1
+ (u
, r
) =
1
+ (u
, r
);
if = i then
=
1
+ (v
, c
) =
1
+ (x
i
, #(k
i1
, t,
i1
));
if i < p then
=
1
+(v
, c
) =
1
+(u
1
, r
1
) =
1
+(u
1
, r
1
);
if > p then
=
1
+ (v
, c
) =
1
+ (u
1
, r
1
).
Also consider that
i1
=
/dom(k
i1
)
= ((
m
)
/dom()
)
/dom(k
i1
)
= (
m
)
/dom(k
i1
)
.
Since we have been able to dene k
m
x
i
/t and
k
m
x
i
/t, we must have
#(k
m
x
i
/t,
k
m
x
i
/t,
m
) = #(k
m
, ,
p+m
) = w .
So we conclude that w #(k
m
x
i
/t,
k
m
x
i
/t,
m
)[
m
(k
m
x
i
/t), _
m
.
114 M. Avon
For the converse implication we assume
w #(k
m
x
i
/t,
k
m
x
i
/t,
m
)[
m
(k
m
x
i
/t), _
m
,
and try to show that w #(k
m
, ,
m
)[
m
(k
m
),
p
_
m
.
There exists
m
= (u
, r
) (k
m
x
i
/t) such that _
m
and
w = #(k
m
x
i
/t,
k
m
x
i
/t,
m
) .
Clearly dom(
m
) = dom(k
m
x
i
/t) = 1, . . . , p +m1.
We dene
1
=
p
+ (u
p
, r
p
) and if m > 1 for each j = 1 . . . m1
j+1
=
j
+ (u
p+j
, r
p+j
). Our goal is to show that
m
(k
m
) and #(k
m
, ,
m
) = #(k
m
x
i
/t,
k
m
x
i
/t,
m
) .
First of all we dene
0
= and for each j = 1 . . . p +m1
j
=
j1
+ (u
j
, r
j
). By
lemma 4.9 we derive that
p+m1
=
m
. Therefore _
p+m1
.
There exists C T such that C 1, . . . p +m1, = (
p+m1
)
/C
.
We have C = dom() = 1, . . . , p 1, and therefore
= (
p+m1
)
/{1,...,p1}
= (u
/{1,...,p1}
, r
/{1,...,p1}
) =
p1
.
We dene h
0
= h. We use backward induction on j to show that for each j = m. . . 1
p+j1
(h
j
),
p+j2
(h
j1
), u
p+j1
= y
j
, r
p+j1
#(h
j1
,
j
,
p+j2
) .
Clearly
p+m1
=
m
(h
m
).
We have h
m
= h
m1
+ (y
m
,
m
), h
m1
K, y
m
1 var(h
m1
),
m
E
s
(h
m1
).
By lemma 4.11 this implies there exist (h
m1
), s #(h
m1
,
m
, ) such that
p+m2
+ (u
p+m1
, r
p+m1
) =
p+m1
= + (y
m
, s) .
By lemma 2.1 we obtain
p+m2
= (h
m1
), u
p+m1
= y
m
, r
p+m1
#(h
m1
,
m
,
p+m2
) .
If m > 1 we need an inductive step. Let j = m. . . 2 and assume
p+j1
(h
j
),
p+j2
(h
j1
), u
p+j1
= y
j
, r
p+j1
#(h
j1
,
j
,
p+j2
) .
We need to prove
p+j3
(h
j2
), u
p+j2
= y
j1
, r
p+j2
#(h
j2
,
j1
,
p+j3
) .
We have h
j1
= h
j2
+ (y
j1
,
j1
), h
j2
K, y
j1
1 var(h
j2
),
j1
E
s
(h
j2
).
By lemma 4.11 this implies there exist (h
j2
), s #(h
j2
,
j1
, ) such that
p+j3
+ (u
p+j2
, r
p+j2
) =
p+j2
= + (y
j1
, s) .
By lemma 2.1 we obtain
p+j3
= (h
j2
), u
p+j2
= y
j1
, r
p+j2
#(h
j2
,
j1
,
p+j3
) .
A dierent approach to logic 115
To show that
m
(k
m
) we show by induction on j that for each j = 1 . . . m
j
(k
j
).
We begin by showing that
1
(k
1
) using lemma 4.11. We have k
1
= k + (y
1
,
1
)
and k K, y
1
1 var(k),
1
E
s
(k).
Moreover
p
(k) and
1
=
p
+ (u
p
, r
p
) =
p
+ (y
1
, r
p
), so to prove
1
(k
1
) we
just need to show that r
p
#(k,
1
,
p
).
But we have proved r
p
#(h,
1
,
p1
) = #(kx
i
/t, (
1
)
k
x
i
/t, ). Since we have
been able to dene (
1
)
k
x
i
/t we can assume #(k,
1
,
p
) = #(kx
i
/t, (
1
)
k
x
i
/t, ).
Therefore r
p
#(k,
1
,
p
) and
1
(k
1
) are proved.
Suppose m > 1 and let j = 1 . . . m 1. We assume
j
(k
j
) and try to show
j+1
(k
j+1
), using lemma 4.11. We have k
j+1
= k
j
+ (y
j+1
,
j+1
), where k
j
K,
y
j+1
1 var(k
j
),
j+1
E
s
(k
j
).
Moreover
j+1
=
j
+ (u
p+j
, r
p+j
) =
j
+ (y
j+1
, r
p+j
) and
j
(k
j
), so to show that
j+1
(k
j+1
) we just need to prove r
p+j
#(k
j
,
j+1
,
j
).
We have proved r
p+j
#(h
j
,
j+1
,
p+j1
), that we can rewrite
r
p+j
#(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t,
p+j1
) .
Well try to exploit this. For each = 0 . . . p we dene
p
.
Let u
, r
= u
and r
= r
.
We have
0
= and for each = 1 . . . p +j 1
=
1
+ (u
, r
) =
1
+ (u
, r
).
Therefore, by lemma 4.9,
p+j1
= (u
, r
).
For each = 1 . . . p 1 we have u
= u
= u
, r
= r
= r
.
We have
0
= and for each = 1 . . . p +j
if < i then
=
1
+ (u
, r
) =
1
+ (u
, r
);
if = i then
1
+ (x
i
, #(k
i1
, t,
i1
));
if i < p then
=
1
+ (u
1
, r
1
) =
1
+ (u
1
, r
1
);
if = p + 1 then
1
=
p
+ (u
p
, r
p
) =
p
+ (u
p
, r
p
) =
1
+ (u
1
, r
1
);
if > p + 1 then
p
=
p1
+ (u
1
, r
1
) =
1
+ (u
1
, r
1
).
Also consider that =
p1
_
p+j1
, so = (
p+j1
)
/dom()
and
i1
=
/dom(k
i1
)
= ((
p+j1
)
/dom()
)
/dom(k
i1
)
= (
p+j1
)
/dom(k
i1
)
.
Since we have been able to dene k
j
x
i
/t and (
j+1
)
k
j
x
i
/t, we must have
#(k
j
x
i
/t, (
j+1
)
k
j
x
i
/t,
p+j1
) = #(k
j
,
j+1
,
p+j
) .
It follows that r
p+j
#(k
j
,
j+1
,
p+j
) = #(k
j
,
j+1
,
j
). So we have proved
j+1
(k
j+1
) and the proof of
m
(k
m
) is nished.
We still need to show that #(k
m
, ,
m
) = #(k
m
x
i
/t,
k
m
x
i
/t,
m
).
116 M. Avon
For each = 1 . . . p 1 we have u
= u
, r
= r
.
We have
0
= and for each = 1 . . . p +m
if < i then
=
1
+ (u
, r
) =
1
+ (u
, r
);
if = i then
1
+ (x
i
, #(k
i1
, t,
i1
));
if i < p then
=
1
+ (u
1
, r
1
) =
1
+ (u
1
, r
1
);
if = p + 1 then
1
=
p
+ (u
p
, r
p
) =
p
+ (u
p
, r
p
) =
1
+ (u
1
, r
1
);
if > p + 1 then
p
=
p1
+ (u
1
, r
1
) =
1
+ (u
1
, r
1
).
Also consider that _
m
, so = (
m
)
/dom()
and
i1
=
/dom(k
i1
)
= ((
m
)
/dom()
)
/dom(k
i1
)
= (
m
)
/dom(k
i1
)
.
Since we have been able to dene k
m
x
i
/t and
k
m
x
i
/t, we must have
#(k
m
, ,
m
) = #(k
m
, ,
p+m
) = #(k
m
x
i
/t,
k
m
x
i
/t,
m
) = w .
Therefore we conclude that w #(k
m
, ,
m
)[
m
(k
m
),
p
_
m
.
To nish with the case E
e
(n+1, k) we need to show V
b
(
k
x
i
/t) V
b
()V
b
(t).
Using (4.0.1) on page 110 and the inductive hypothesis we obtain
V
b
(
k
x
i
/t) = y
1
, . . . , y
m
V
b
(
1
) V
b
(
m
) V
b
() =
= y
1
, . . . , y
m
V
b
((
1
)
k
x
i
/t) V
b
((
m
)
k
m1
x
i
/t) V
b
(
k
m
x
i
/t)
y
1
, . . . , y
m
(V
b
(
1
) V
b
(t)) (V
b
(
m
) V
b
(t)) (V
b
() V
b
(t)) =
= y
1
, . . . , y
m
V
b
(
1
) V
b
(
m
) V
b
() V
b
(t) =
= V
b
() V
b
(t) .
We have dened
k
x
i
/t for each E(n+1, k) such that V
b
(t) V
b
() = . Recall
that
E(n+1, k) = E
(n, k)E
a
(n+1, k)E
b
(n+1, k)E
c
(n+1, k)E
d
(n+1, k)E
e
(n+1, k).
and recall that the denition of
k
x
i
/t depends on the set to which belongs to.
Actually may belong to more than one of these sets. We need to check that, in every
case in which belongs to two of the six sets, the two denitions of
k
x
i
/t match each
other.
We split the task in two steps. The rst step requires to verify that
for each w a, b, c, d, e if E
(n, k) E
w
(n + 1, k) then
(
k
x
i
/t)
w
=
k
x
i
/t.
The second step requires to verify that
for each w
1
, w
2
a, b, c, d, e if w
1
,= w
2
, E
w
1
(n + 1, k) E
w
2
(n + 1, k) then
(
k
x
i
/t)
w
1
= (
k
x
i
/t)
w
2
.
A dierent approach to logic 117
We begin with the rst step and examine the case where E
(n, k) E
a
(n +1, k).
Of course E(n, k) E
a
(n + 1, k) and k K(n)
+
.
We have k = k
p
= k
p1
+(x
p
,
p
), k
p1
K(n),
p
E
s
(n, k
p1
), x
p
1var(k
p1
).
Therefore E(n, k
p1
), x
p
/ V
b
().
Consider the case where i = p. Here we have (
k
x
i
/t)
a
= .
We also see that k
p1
K(n), E(n, k
p1
), x
i
/ var(k
p1
). At the beginning of
our denition we declared the intention to show the truth of some properties. Clearly we
will show these properties are true at step n + 1, and we can assume their truth at step
n. One of those properties tells us that in this case
k
x
i
/t = . So
(
k
x
i
/t)
a
= =
k
x
i
/t .
We now examine the case where i < p. Here we dened (
k
x
i
/t)
a
=
k
p1
x
i
/t.
It also holds true that k
p1
K(n), k
i
_ k
p1
, /(n; k
p1
; x
1
:
1
, . . . , x
p1
:
p1
),
E(n, k
p1
). Another declared property tells us that
k
x
i
/t =
k
p1
x
i
/t. So
(
k
x
i
/t)
a
=
k
p1
x
i
/t =
k
x
i
/t .
Lets turn to examine the case where E
(n, k) E
b
(n +1, k).
Of course E(n, k) E
b
(n + 1, k) and k K(n)
+
.
We have k = k
p
= k
p1
+(x
p
,
p
), k
p1
K(n),
p
E
s
(n, k
p1
), x
p
1var(k
p1
).
Therefore = x
p
.
Since E(n, k) the following condition holds:
var(k), = x
i
k
x
i
/t = t, ,= x
i
k
x
i
/t = .
Consider the case where i = p. Here we dened (
k
x
i
/t)
b
= t and since = x
p
= x
i
we have
k
x
i
/t = t = (
k
x
i
/t)
b
.
Turn to the case where i < p. Here we dened (
k
x
i
/t)
b
= and since = x
p
,= x
i
we have
k
x
i
/t = = (
k
x
i
/t)
b
.
Lets examine the case where E
(n, k) E
c
(n +1, k).
Of course E(n, k) E
c
(n + 1, k).
Since E(n, k) the following condition holds:
n > 1, there exist h K(n1): h _ k, a positive integer m, ,
1
, . . . ,
m
E(n1, h)
such that = ()(
1
, . . . ,
m
), E(n, h), for each (h) #(h, , ) is a function
with m arguments, (#(h,
1
, ), . . . , #(h,
m
, )) is a member of the domain of #(h, , ).
If h ,= there exists a positive integer q p such that q < n 1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
118 M. Avon
If i q, or in other words x
i
var(h), we can dene
h
x
i
/t, and similarly we can
dene (
j
)
h
x
i
/t, and it results
k
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
Since E
c
(n + 1, k) the following condition holds:
there exist a positive integer r and ,
1
, . . . ,
r
E(n, k) such that
= ()(
1
, . . . ,
r
);
for each (k) #(k, , ) is a function with r arguments and
(#(k,
1
, ), . . . , #(k,
r
, )) is a member of its domain;
(
k
x
i
/t)
c
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
r
)
k
x
i
/t).
This implies that r = m, = and for each j = 1 . . . m
j
=
j
. Therefore
(
k
x
i
/t)
c
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) .
Suppose h ,= and i q, in this case we have
k
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
We have k
i
_ h, /(n; h; x
1
:
1
, . . . , x
q
:
q
), ,
1
, . . . ,
m
E(n, h). We can apply
one of our declared properties and obtain that
k
x
i
/t =
h
x
i
/t, for each j = 1 . . . m
(
j
)
k
x
i
/t = (
j
)
h
x
i
/t. Therefore
(
k
x
i
/t)
c
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) =
k
x
i
/t .
Consider instead the case where h = or h ,= i > q. In this case x
i
/ var(h), and
,
1
, . . . ,
m
E(n, h), so by one of our declared properties
(
k
x
i
/t)
c
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= ()(
1
, . . . ,
m
) = =
k
x
i
/t .
Lets examine the case where E
(n, k) E
d
(n +1, k).
Of course E(n, k) E
d
(n + 1, k).
Since E(n, k) the following condition holds:
n > 1, there exist h K(n1): h _ k, a positive integer m,
1
, . . . ,
m
E(n1, h),
f T, such that = (f)(
1
, . . . ,
m
), E(n, h), for each (h)
A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
If h ,= there exists a positive integer q p such that q < n 1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), we can dene (
j
)
h
x
i
/t, and it results
k
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
A dierent approach to logic 119
Since E
d
(n+1, k) there exist f T, a positive integer r and
1
, . . . ,
r
E(n, k)
such that
= (f)(
1
, . . . ,
r
);
for each (k) A
f
(#(k,
1
, ), . . . , #(k,
r
, )) holds true;
(
k
x
i
/t)
d
= (f)((
1
)
k
x
i
/t, . . . , (
r
)
k
x
i
/t).
This implies that r = m and for each j = 1 . . . m
j
=
j
. Therefore
(
k
x
i
/t)
d
= (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) .
Suppose h ,= and i q, in this case we have
k
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
We have k
i
_ h, /(n; h; x
1
:
1
, . . . , x
q
:
q
),
1
, . . . ,
m
E(n, h). We can apply one
of our declared properties and obtain that for each j = 1 . . . m (
j
)
k
x
i
/t = (
j
)
h
x
i
/t.
Therefore
(
k
x
i
/t)
d
= (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) =
k
x
i
/t .
Consider instead the case where h = or h ,= i > q. In this case x
i
/ var(h), and
1
, . . . ,
m
E(n, h), so by one of our declared properties
(
k
x
i
/t)
d
= (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= (f)(
1
, . . . ,
m
) = =
k
x
i
/t .
Lets examine the case where E
(n, k) E
e
(n +1, k).
Of course E(n, k) E
e
(n + 1, k).
Since E(n, k) the following condition holds:
n > 1, there exist h K(n 1): h _ k, a positive integer m, E(n 1),
a function y whose domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var(h)
and for each , = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n 1);
such that
c(n 1, h, m, y, , ),
= (y
1
:
1
, . . . , y
m
:
m
, ), E(n, h).
If h ,= there exists a positive integer q p such that q < n 1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
Suppose i q, or in other words x
i
var(h).
We dene h
1
= h+(y
1
,
1
), and if m > 1 for each j = 1 . . . m1 h
j+1
= h
j
+(y
j+1
,
j+1
).
We have
1
E(n 1, h), V
b
(t) V
b
(
1
) = , therefore (
1
)
h
x
i
/t is dened;
for each j = 1 . . . m1 h
j
K(n 1) and
/(n 1; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
j+1
E(n 1, h
j
), V
b
(t) V
b
(
j+1
) = , (
j+1
)
h
j
x
i
/t is dened;
120 M. Avon
h
m
K(n 1) and
/(n 1; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
E(n 1, h
m
), V
b
(t) V
b
() = ,
h
m
x
i
/t is dened;
it results
k
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
Since E
e
(n + 1, k) there exist
a positive integer r,
a function z whose domain is 1, . . . , r such that for each j = 1 . . . r
z
j
1 var(k), and for each , = 1 . . . r ,= z
,= z
,
a function whose domain is 1, . . . , r such that for each j = 1 . . . r
j
E(n),
E(n)
such that = (z
1
:
1
, . . . , z
r
:
r
, ), c(n, k, r, z, , ) and
(
k
x
i
/t)
e
= (z
1
: (
1
)
k
x
i
/t, . . . , z
r
: (
r
)
k
r1
x
i
/t,
k
r
x
i
/t) ,
where k
1
= k+(y
1
,
1
), and if m > 1 for each j = 1 . . . m1 k
j+1
= k
j
+(y
j+1
,
j+1
).
Clearly r = m, z = y, = , = , therefore
(
k
x
i
/t)
e
= (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) .
Suppose h ,= and i q, we have
k
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) .
We recall that k K(n), /(n; k; x
1
:
1
, . . . , x
p
:
p
),
1
E(n, k) and
V
b
(
1
) V
b
(t) = . This ensures (
1
)
k
x
i
/t is dened, and we have h K(n), k
i
_ h,
/(n; h; x
1
:
1
, . . . , x
q
:
q
),
1
E(n, h). By one of our declared properties we obtain
that (
1
)
k
x
i
/t = (
1
)
h
x
i
/t.
If m > 1 let j = 1 . . . m1, we want to show that (
j+1
)
k
j
x
i
/t = (
j+1
)
h
j
x
i
/t.
We recall that k
j
K(n) and /(n; k
j
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
, . . . , y
j
:
j
), for each
= i + 1 . . . p x
/ V
b
(t) and V
b
(
) V
b
(t) = , for each = 1 . . . j y
/ V
b
(t) and
V
b
(
) V
b
(t) = ,
j+1
E(n, k
j
), V
b
(
j+1
) V
b
(t) = . As a result of these conditions
we were able to dene (
j+1
)
k
j
x
i
/t.
We have also h
j
K(n), /(n; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
k
i
_ h _ h
j
,
j+1
E(n, h
j
). By one of our declared properties we obtain that
(
j+1
)
k
j
x
i
/t = (
j+1
)
h
j
x
i
/t .
We also want to show that
k
m
x
i
/t) =
h
m
x
i
/t).
We recall that k
m
K(n) and /(n; k
m
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
, . . . , y
m
:
m
), for
each = i + 1 . . . p x
/ V
b
(t) and V
b
(
) V
b
(t) = , for each = 1 . . . m y
/ V
b
(t)
and V
b
(
) V
b
(t) = . We have E(n, k
m
) and also V
b
() V
b
(t) = . As a result of
these conditions we were able to dene
k
m
x
i
/t.
A dierent approach to logic 121
We have also h
m
K(n), /(n; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
k
i
_ h _ h
m
, E(n, h
m
). By one of our declared properties we obtain that
m
x
i
/t =
h
m
x
i
/t .
Hence
k
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) =
= (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) = (
k
x
i
/t)
e
.
We now consider the alternative case h = or h ,= i > q. In this case
k
x
i
/t = = (y
1
:
1
, . . . , y
m
:
m
, ) .
We could dene (
1
)
k
x
i
/t, and we have h K(n),
1
E(n, h) (follows by
c(n 1, h, m, y, , )), x
i
/ var(h). By one declared property (
1
)
k
x
i
/t =
1
.
If m > 1 suppose j = 1 . . . m 1. We want to show that (
j+1
)
k
j
x
i
/t =
j+1
.
Recall we were able to dene (
j+1
)
k
j
x
i
/t. Recall that c(n 1, h, m, y, , ) holds, so
h
j
K(n),
j+1
E(n, h
j
). Moreover, for each = 1 . . . m since y
1 var(k) it also
results y
,= x
i
, so x
i
/ var(h
j
). By one declared property (
j+1
)
k
j
x
i
/t =
j+1
.
We also need to show
k
m
x
i
/t = . Recall we were able to dene ()
k
m
x
i
/t.
Recall that c(n1, h, m, y, , ) holds, so h
m
K(n), E(n, h
m
). Moreover, for each
= 1 . . . m since y
,= x
i
, so x
i
/ var(h
m
). By one
declared property
k
m
x
i
/t = .
Therefore
k
x
i
/t = = (y
1
:
1
, . . . , y
m
:
m
, ) =
= (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) = (
k
x
i
/t)
e
.
We now turn to the second step of our task. This requires to verify that
for each w
1
, w
2
a, b, c, d, e if w
1
,= w
2
, E
w
1
(n + 1, k) E
w
2
(n + 1, k) then
(
k
x
i
/t)
w
1
= (
k
x
i
/t)
w
2
.
Within denition 2.7 we have seen that for many values of w
1
, w
2
it results
E
w
1
(n + 1, k) E
w
2
(n + 1, k) = .
In fact, we have seen that in all the cases in which w
1
, w
2
b, c, d, e and w
1
,= w
2
E
w
1
(n + 1, k) E
w
2
(n + 1, k) = .
Moreover, we have proved that E
a
(n + 1, k) E
b
(n + 1, k) = .
Therefore we just need to examine three cases: E
a
(n + 1, k) E
c
(n + 1, k),
E
a
(n + 1, k) E
d
(n + 1, k), E
a
(n + 1, k) E
e
(n + 1, k).
We start with the case where E
a
(n +1, k) E
c
(n +1, k). Clearly belongs to
E
a
(n + 1, k) E
c
(n + 1, k).
122 M. Avon
Since E
c
(n + 1, k) the following condition holds: there exist a positive integer m
and ,
1
, . . . ,
m
E(n, k) such that
= ()(
1
, . . . ,
m
);
(
k
x
i
/t)
c
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t).
Since E
a
(n + 1, k) we have E(n, k
p1
), x
p
/ V
b
(). We distinguish the case
where i = p from the case where i < p.
If i = p then (
k
x
i
/t)
a
= . Given that E(n, k
p1
) we can use assump-
tion 2.1.10 to obtain that n > 1 and there exist h K(n 1) such that h _ k
p1
and a
positive integer r, ,
1
, . . .
r
E(n 1, h) such that = ()(
1
, . . . ,
r
).
Clearly m = r, = E(n 1, h) and for each j = 1 . . . m
j
=
j
E(n 1, h).
Therefore ,
1
, . . . ,
m
E(n 1, h).
We can apply one of our declared properties. In fact h K(n) and x
i
/ var(h), so
k
x
i
/t = and for each j = 1 . . . m (
j
)
k
x
i
/t =
j
. Therefore
(
k
x
i
/t)
c
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= ()(
1
, . . . ,
m
) = = (
k
x
i
/t)
a
.
Now suppose i < p. Here (
k
x
i
/t)
a
=
k
p1
x
i
/t. Since E(n, k
p1
) we can
apply one of our inductive assumptions and obtain the following:
n > 1, there exist h K(n 1): h _ k
p1
, a positive integer r,
,
1
, . . . ,
r
E(n 1, h) such that = ()(
1
, . . . ,
r
).
If h ,= there exists a positive integer q p 1 such that q < n 1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), we can dene
h
x
i
/t, and similarly we can
dene (
j
)
h
x
i
/t, and it results
k
p1
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
r
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
p1
x
i
/t = .
Clearly m = r, = E(n 1, h) and for each j = 1 . . . m
j
=
j
E(n 1, h).
Therefore ,
1
, . . . ,
m
E(n 1, h).
Moreover, if x
i
var(h) it results
(
k
x
i
/t)
a
=
k
p1
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) ;
otherwise (
k
x
i
/t)
a
=
k
p1
x
i
/t = .
Suppose x
i
var(h). It follows that k
i
_ h. Since ,
1
, . . . ,
m
E(n1, h) we can
apply one of our declared inductive assumptions and get
k
x
i
/t =
h
x
i
/t and for
each j = 1 . . . m (
j
)
k
x
i
/t = (
j
)
h
x
i
/t. Therefore
(
k
x
i
/t)
a
=
k
p1
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) =
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) = (
k
x
i
/t)
c
.
A dierent approach to logic 123
Suppose instead x
i
/ var(h). Since ,
1
, . . . ,
m
E(n1, h) by one of our declared
inductive assumptions we obtain
k
x
i
/t = and for each j = 1 . . . m (
j
)
k
x
i
/t =
j
.
Therefore
(
k
x
i
/t)
c
= (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= ()(
1
, . . . ,
m
) = = (
k
x
i
/t)
a
.
We now examine the case where E
a
(n +1, k) E
d
(n +1, k). Clearly belongs
to E
a
(n + 1, k) E
d
(n + 1, k).
Since E
d
(n + 1, k) the following condition holds: there exist f T, a positive
integer m and
1
, . . . ,
m
E(n, k) such that
= (f)(
1
, . . . ,
m
);
(
k
x
i
/t)
d
= (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t).
Since E
a
(n + 1, k) we have E(n, k
p1
), x
p
/ V
b
(). We distinguish the case
where i = p from the case where i < p.
If i = p then (
k
x
i
/t)
a
= . Given that E(n, k
p1
) we can use assump-
tion 2.1.10 to obtain that n > 1 and there exist h K(n1): h _ k
p1
, g T, a positive
integer r,
1
, . . .
r
E(n 1, h): = (g)(
1
, . . . ,
r
).
Clearly f = g, m = r and for each j = 1 . . . m
j
=
j
E(n 1, h). Therefore
1
, . . . ,
m
E(n 1, h).
We can apply one of our declared properties. In fact h K(n) and x
i
/ var(h), so
for each j = 1 . . . m (
j
)
k
x
i
/t =
j
. Therefore
(
k
x
i
/t)
d
= (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= (f)(
1
, . . . ,
m
) = = (
k
x
i
/t)
a
.
Now suppose i < p. Here (
k
x
i
/t)
a
=
k
p1
x
i
/t. Since E(n, k
p1
) we can
apply one of our inductive assumptions and obtain the following:
n > 1, there exist h K(n 1): h _ k
p1
, g T, a positive integer r,
1
, . . . ,
r
E(n 1, h) such that = (g)(
1
, . . . ,
r
).
If h ,= there exists a positive integer q p 1 such that q < n 1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), we can dene (
j
)
h
x
i
/t, and it results
k
p1
x
i
/t = (g)((
1
)
h
x
i
/t, . . . , (
r
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
p1
x
i
/t = .
Clearly f = g, m = r and for each j = 1 . . . m
j
=
j
E(n 1, h). Therefore
1
, . . . ,
m
E(n 1, h).
Moreover, if x
i
var(h) it results
(
k
x
i
/t)
a
=
k
p1
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) ;
124 M. Avon
otherwise (
k
x
i
/t)
a
=
k
p1
x
i
/t = .
Suppose x
i
var(h). It follows that k
i
_ h. Since
1
, . . . ,
m
E(n 1, h) we
can apply one of our declared inductive assumptions and get, for each j = 1 . . . m,
(
j
)
k
x
i
/t = (
j
)
h
x
i
/t. Therefore
(
k
x
i
/t)
a
=
k
p1
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) =
= (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) = (
k
x
i
/t)
d
.
Suppose instead x
i
/ var(h). Since
1
, . . . ,
m
E(n 1, h) by one of our declared
inductive assumptions we obtain, for each j = 1 . . . m, (
j
)
k
x
i
/t =
j
. Therefore
(
k
x
i
/t)
d
= (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) =
= (f)(
1
, . . . ,
m
) = = (
k
x
i
/t)
a
.
Finally we turn to the case where E
a
(n +1, k) E
e
(n +1, k). Clearly belongs
to E
a
(n + 1, k) E
e
(n + 1, k).
Since E
e
(n + 1, k) the following condition holds: there exist
a positive integer m,
a function y whose domain is 1, . . . , m such that for each j = 1 . . . m
y
j
1 var(k), and for each , = 1 . . . m ,= y
,= y
,
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n),
E(n)
such that = (y
1
:
1
, . . . , y
m
:
m
, ), c(n, k, m, y, , ) and
(
k
x
i
/t)
e
= (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) .
where k
1
= k +(y
1
,
1
) and if m > 1 for each j = 1 . . . m1 k
j+1
= k
j
+(y
j+1
,
j+1
).
Since E
a
(n + 1, k) we have E(n, k
p1
), x
p
/ V
b
(). We distinguish the case
where i = p from the case where i < p.
If i = p then (
k
x
i
/t)
a
= . Given that E(n, k
p1
) we can use assump-
tion 2.1.10 to obtain that
n > 1 and there exist
h K(n 1): h _ k
p1
,
a positive integer r,
a function z whose domain is 1, . . . , r such that for each j = 1 . . . r
z
j
1 var(h), and for each , = 1 . . . r ,= z
,= z
,
a function whose domain is 1, . . . , r such that for each j = 1 . . . r
j
E(n1),
E(n 1)
such that = (z
1
:
1
, . . . , z
r
:
r
, ) and c(n 1, h, r, z, , ).
Clearly m = r, y = z, = , = .
We dene h
1
= h + (y
1
,
1
), and if m > 1 for each j = 1 . . . m 1
h
j+1
= h
j
+ (y
j+1
,
j+1
).
A dierent approach to logic 125
We could dene (
1
)
k
x
i
/t, and we have h K(n),
1
E(n, h) (follows by
c(n 1, h, m, y, , )), x
i
/ var(h). By one declared property (
1
)
k
x
i
/t =
1
.
If m > 1 suppose j = 1 . . . m 1. We want to show that (
j+1
)
k
j
x
i
/t =
j+1
.
Recall we were able to dene (
j+1
)
k
j
x
i
/t. Recall that c(n 1, h, m, y, , ) holds, so
h
j
K(n),
j+1
E(n, h
j
). Moreover, for each = 1 . . . m since y
1 var(k) it also
results y
,= x
i
, so x
i
/ var(h
j
). By one declared property (
j+1
)
k
j
x
i
/t =
j+1
.
We also need to show
k
m
x
i
/t = . Recall we were able to dene ()
k
m
x
i
/t.
Recall that c(n1, h, m, y, , ) holds, so h
m
K(n), E(n, h
m
). Moreover, for each
= 1 . . . m since y
,= x
i
, so x
i
/ var(h
m
). By one
declared property
k
m
x
i
/t = .
Therefore
(
k
x
i
/t)
e
= (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) =
= (y
1
:
1
, . . . , y
m
:
m
, ) = = (
k
x
i
/t)
a
.
Now suppose i < p. Here (
k
x
i
/t)
a
=
k
p1
x
i
/t. Since E(n, k
p1
) we can
apply one of our inductive assumptions and obtain the following:
n > 1, there exist h K(n 1): h _ k
p1
, a positive integer r, E(n 1),
a function z whose domain is 1, . . . , r such that for each j = 1 . . . r z
j
1 var(h)
and for each , = 1 . . . r ,= z
,= z
;
a function whose domain is 1, . . . , r such that for each j = 1 . . . r
j
E(n 1);
such that
c(n 1, h, r, z, , ),
= (z
1
:
1
, . . . , z
r
:
r
, ), E(n, h).
If h ,= there exists a positive integer q p 1 such that q < n 1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
Suppose i q, or in other words x
i
var(h).
We dene h
1
= h+(z
1
,
1
), and if r > 1 for each j = 1 . . . r 1 h
j+1
= h
j
+(z
j+1
,
j+1
).
We have
1
E(n 1, h), V
b
(t) V
b
(
1
) = , therefore (
1
)
h
x
i
/t is dened;
for each j = 1 . . . r 1 h
j
K(n 1) and
/(n 1; h
j
; x
1
:
1
, . . . , x
q
:
q
, z
1
:
1
, . . . , z
j
:
j
),
for each = 1 . . . j z
/ V
b
(t) and V
b
(t) V
b
(
) = ,
j+1
E(n 1, h
j
),
V
b
(t) V
b
(
j+1
) = , therefore (
j+1
)
h
j
x
i
/t is dened;
h
r
K(n 1) and /(n 1; h
r
; x
1
:
1
, . . . , x
q
:
q
, z
1
:
1
, . . . , z
r
:
r
),
for each = 1 . . . r z
/ V
b
(t) and V
b
(t) V
b
(
) = ,
E(n 1, h
r
), V
b
(t) V
b
() = , therefore
h
r
x
i
/t is dened;
it results
k
p1
x
i
/t = (z
1
: (
1
)
h
x
i
/t, . . . , z
r
: (
r
)
h
r1
x
i
/t,
h
r
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
p1
x
i
/t = .
126 M. Avon
Clearly m = r, y = z, = and = .
Therefore, if x
i
var(h) it results
(
k
x
i
/t)
a
=
k
p1
x
i
/t =
= (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) ;
otherwise (
k
x
i
/t)
a
=
k
p1
x
i
/t = .
Suppose x
i
var(h).
We recall that k K(n), /(n; k; x
1
:
1
, . . . , x
p
:
p
),
1
E(n, k) and
V
b
(
1
) V
b
(t) = . This ensures (
1
)
k
x
i
/t is dened, and we have h K(n), k
i
_ h,
/(n; h; x
1
:
1
, . . . , x
q
:
q
),
1
E(n, h). By one of our declared properties we obtain
that (
1
)
k
x
i
/t = (
1
)
h
x
i
/t.
If m > 1 let j = 1 . . . m1, we want to show that (
j+1
)
k
j
x
i
/t = (
j+1
)
h
j
x
i
/t.
We recall that k
j
K(n) and /(n; k
j
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
, . . . , y
j
:
j
), for each
= i + 1 . . . p x
/ V
b
(t) and V
b
(
) V
b
(t) = , for each = 1 . . . j y
/ V
b
(t) and
V
b
(
) V
b
(t) = ,
j+1
E(n, k
j
), V
b
(
j+1
) V
b
(t) = . As a result of these conditions
we were able to dene (
j+1
)
k
j
x
i
/t.
We have also h
j
K(n), /(n; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
k
i
_ h _ h
j
,
j+1
E(n, h
j
). By one of our declared properties we obtain that
(
j+1
)
k
j
x
i
/t = (
j+1
)
h
j
x
i
/t .
We also want to show that
k
m
x
i
/t) =
h
m
x
i
/t).
We recall that k
m
K(n) and /(n; k
m
; x
1
:
1
, . . . , x
p
:
p
, y
1
:
1
, . . . , y
m
:
m
), for
each = i + 1 . . . p x
/ V
b
(t) and V
b
(
) V
b
(t) = , for each = 1 . . . m y
/ V
b
(t)
and V
b
(
) V
b
(t) = . We have E(n, k
m
) and also V
b
() V
b
(t) = . As a result of
these conditions we were able to dene
k
m
x
i
/t.
We have also h
m
K(n), /(n; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
k
i
_ h _ h
m
, E(n, h
m
). By one of our declared properties we obtain that
m
x
i
/t =
h
m
x
i
/t .
Hence
(
k
x
i
/t)
a
=
k
p1
x
i
/t =
= (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) =
= (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) =
= (
k
x
i
/t)
e
.
Now let x
i
/ var(h). In this case (
k
x
i
/t)
a
= = (y
1
:
1
, . . . , y
m
:
m
, ).
As seen above, we could dene (
1
)
k
x
i
/t, and we have h K(n),
1
E(n, h),
x
i
/ var(h). By one declared property (
1
)
k
x
i
/t =
1
.
If m > 1 suppose j = 1 . . . m1. We want to show that (
j+1
)
k
j
x
i
/t =
j+1
. As seen
above, we were able to dene (
j+1
)
k
j
x
i
/t. We have also h
j
K(n),
A dierent approach to logic 127
j+1
E(n, h
j
). Moreover, for each = 1 . . . m since y
,= x
i
, so x
i
/ var(h
j
). By one declared property (
j+1
)
k
j
x
i
/t =
j+1
.
We also need to show
k
m
x
i
/t = . As seen above, we were able to dene
()
k
m
x
i
/t. We have also h
m
K(n), E(n, h
m
). Moreover, for each = 1 . . . m
since y
,= x
i
, so x
i
/ var(h
m
). By one declared property
m
x
i
/t = .
Therefore
(
k
x
i
/t)
a
= (y
1
:
1
, . . . , y
m
:
m
, ) =
= (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) =
= (
k
x
i
/t)
e
.
k
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
a4. there exist h K(n): h _ k, f T, a positive integer m,
1
, . . . ,
m
E(n, h)
such that = (f)(
1
, . . . ,
m
), E(n + 1, h),
for each (h) A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
If h ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have
V
b
(t) V
b
(
j
) V
b
(t) V
b
() = ,
128 M. Avon
we can dene (
j
)
h
x
i
/t, and it results
k
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
a5. there exist h K(n): h _ k, a positive integer m, E(n), a function y whose
domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var(h) and for each
, = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n);
such that
c(n, h, m, y, , ), = (y
1
:
1
, . . . , y
m
:
m
, ), E(n + 1, h).
If h ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; h; x
1
:
1
, . . . , x
q
:
q
).
Suppose i q, or in other words x
i
var(h).
We dene h
1
= h + (y
1
,
1
), and if m > 1 for each j = 1 . . . m 1
h
j+1
= h
j
+ (y
j+1
,
j+1
).
We have
1
E(n, h), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore (
1
)
h
x
i
/t is
dened;
for each j = 1 . . . m1 h
j
K(n) and by 4.7
/(n; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
for each = 1 . . . j y
V
b
() so y
/ V
b
(t),
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n, h
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
h
j
x
i
/t is dened;
h
m
K(n) and by 4.7
/(n; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
for each = 1 . . . m y
V
b
() so y
/ V
b
(t),
for each = 1 . . . m V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
E(n, h
m
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
h
m
x
i
/t is dened;
it results
k
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
In this case too we need to remember that
E(n+1, k) = E
(n, k)E
a
(n+1, k)E
b
(n+1, k)E
c
(n+1, k)E
d
(n+1, k)E
e
(n+1, k).
Suppose E
k
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
b4. n > 1, there exist h K(n 1): h _ k, f T, a positive integer m,
1
, . . . ,
m
E(n 1, h) such that = (f)(
1
, . . . ,
m
), E(n, h), for each
(h) A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
If h ,= by 4.7 we can derive there exists a positive integer q p such that q < n1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have
V
b
(t) V
b
(
j
) V
b
(t) V
b
() = ,
we can dene (
j
)
h
x
i
/t, and it results
k
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
b5. n > 1, there exist h K(n 1): h _ k, a positive integer m, E(n 1),
a function y whose domain is 1, . . . , m such that
for each j = 1 . . . m y
j
1 var(h) and for each , = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that
for each j = 1 . . . m
j
E(n 1);
such that
c(n 1, h, m, y, , ),
= (y
1
:
1
, . . . , y
m
:
m
, ), E(n, h).
If h ,= by 4.7 we can derive there exists a positive integer q p such that q < n1,
/(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
Suppose i q, or in other words x
i
var(h).
We dene h
1
= h + (y
1
,
1
), and if m > 1 for each j = 1 . . . m 1
h
j+1
= h
j
+ (y
j+1
,
j+1
).
130 M. Avon
We have
1
E(n1, h), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore (
1
)
h
x
i
/t
is dened;
for each j = 1 . . . m1 h
j
K(n 1) and by 4.7
/(n 1; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
for each = 1 . . . j y
V
b
() so y
/ V
b
(t),
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n 1, h
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
h
j
x
i
/t is dened;
h
m
K(n 1) and by 4.7
/(n 1; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
for each = 1 . . . m y
V
b
() so y
/ V
b
(t),
for each = 1 . . . m V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
E(n 1, h
m
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
h
m
x
i
/t is dened;
it results
k
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
Clearly if b1. holds then a1. holds too, if b2. holds then a2. holds too. If b3. holds
then a3. holds too, if b4. holds then a4. holds too. Finally if b5. holds then a5. holds too.
We turn to the case where E
a
(n +1, k). This implies E
a
(n+1, k), k K(n)
+
.
We have k = k
p
= k
p1
+(x
p
,
p
), k
p1
K(n),
p
E
s
(n, k
p1
), x
p
1var(k
p1
).
Therefore E(n, k
p1
), x
p
/ V
b
(). We have to distinguish the case where i < p from
the one where i = p.
First we suppose i < p. In this case 1 i p 1, so k
p1
,= . As we have seen above
/(n; k
p1
; x
1
:
1
, . . . , x
p1
:
p1
) holds. So we can apply the inductive hypothesis to
and obtain that one of the following ve conditions holds:
( and
k
p1
x
i
/t = .
var(k
p1
), = x
i
k
p1
x
i
/t = t, ,= x
i
k
p1
x
i
/t = .
n > 1, there exist h K(n 1): h _ k
p1
, a positive integer m,
,
1
, . . . ,
m
E(n 1, h) such that = ()(
1
, . . . ,
m
), E(n, h), for each
(h) #(h, , ) is a function with m arguments, (#(h,
1
, ), . . . , #(h,
m
, ))
is a member of the domain of #(h, , ).
If h ,= by 4.7 we can derive there exists a positive integer q p 1 such that
q < n 1, /(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have
V
b
(t) V
b
() V
b
(t) V
b
() = ,
A dierent approach to logic 131
we can dene
h
x
i
/t, and similarly we can dene (
j
)
h
x
i
/t, and it results
k
p1
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
p1
x
i
/t = .
n > 1, there exist h K(n 1): h _ k
p1
, f T, a positive integer m,
1
, . . . ,
m
E(n 1, h) such that = (f)(
1
, . . . ,
m
), E(n, h), for each
(h) A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
If h ,= by 4.7 we can derive there exists a positive integer q p 1 such that
q < n 1, /(n 1; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have
V
b
(t) V
b
(
j
) V
b
(t) V
b
() = ,
we can dene (
j
)
h
x
i
/t, and it results
k
p1
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
p1
x
i
/t = .
n > 1, there exist h K(n 1): h _ k
p1
, a positive integer m, E(n 1), a
function y whose domain is 1, . . . , m such that for each j = 1 . . . m y
j
1var(h)
and for each , = 1 . . . m ,= y
,= y
1
= h + (y
1
,
1
), and if m > 1 for each j = 1 . . . m1
h
j+1
= h
j
+ (y
j+1
,
j+1
).
We have
1
E(n1, h), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore (
1
)
h
x
i
/t
is dened;
for each j = 1 . . . m1 h
j
K(n 1) and by 4.7
/(n 1; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
for each = 1 . . . j y
V
b
() so y
/ V
b
(t),
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n 1, h
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
h
j
x
i
/t is dened;
h
m
K(n 1) and by 4.7
/(n 1; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
for each = 1 . . . m y
V
b
() so y
/ V
b
(t),
for each = 1 . . . m V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
132 M. Avon
E(n 1, h
m
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
h
m
x
i
/t is dened;
it results
k
p1
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
p1
x
i
/t = .
In this case i < p we dened
k
x
i
/t =
k
p1
x
i
/t, therefore one of the following
ve conditions holds:
( and
k
x
i
/t = .
var(k), = x
i
k
x
i
/t = t, ,= x
i
k
x
i
/t = .
there exist h K(n): h _ k, a positive integer m, ,
1
, . . . ,
m
E(n, h) such that
= ()(
1
, . . . ,
m
), E(n + 1, h), for each (h) #(h, , ) is a function
with m arguments, (#(h,
1
, ), . . . , #(h,
m
, )) is a member of the domain of
#(h, , ).
If h ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have
V
b
(t) V
b
() V
b
(t) V
b
() = ,
we can dene
h
x
i
/t, and similarly we can dene (
j
)
h
x
i
/t, and it results
k
x
i
/t = (
h
x
i
/t)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
there exist h K(n): h _ k, f T, a positive integer m,
1
, . . . ,
m
E(n, h)
such that = (f)(
1
, . . . ,
m
), E(n + 1, h),
for each (h) A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
If h ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; h; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(h), since we have
V
b
(t) V
b
(
j
) V
b
(t) V
b
() = ,
we can dene (
j
)
h
x
i
/t, and it results
k
x
i
/t = (f)((
1
)
h
x
i
/t, . . . , (
m
)
h
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
there exist h K(n): h _ k, a positive integer m, E(n), a function y whose
domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var(h) and for each
, = 1 . . . m ,= y
,= y
1
= h + (y
1
,
1
), and if m > 1 for each j = 1 . . . m1
h
j+1
= h
j
+ (y
j+1
,
j+1
).
We have
1
E(n, h), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore (
1
)
h
x
i
/t is
dened;
for each j = 1 . . . m1 h
j
K(n) and by 4.7
/(n; h
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
for each = 1 . . . j y
V
b
() so y
/ V
b
(t),
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n, h
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
h
j
x
i
/t is dened;
h
m
K(n) and by 4.7
/(n; h
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
for each = 1 . . . m y
V
b
() so y
/ V
b
(t),
for each = 1 . . . m V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
E(n, h
m
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
h
m
x
i
/t is dened;
it results
k
x
i
/t = (y
1
: (
1
)
h
x
i
/t, . . . , y
m
: (
m
)
h
m1
x
i
/t,
h
m
x
i
/t) .
Otherwise (when h = or h ,= i > q)
k
x
i
/t = .
We now consider the case where i = p, in which we dened
k
x
i
/t = .
Since E(n, k
p1
) we can apply assumption 2.1.10 to establish that one of the following
ve conditions holds:
c1. (.
c2. there exists j = 1 . . . p 1 such that = x
j
.
c3.
h K(n 1) : h _ k
p1
, m positive integer , ,
1
, . . .
m
E(n 1, h) :
= ()(
1
, . . . ,
m
), E(n, h),
(h) ( #(h, , ) is a function with m arguments,
(#(h,
1
, ), . . . , #(h,
m
, )) is a member of the domain of #(h, , ).
134 M. Avon
c4.
h K(n 1) : h _ k
p1
, f T, m positive integer ,
1
, . . .
m
E(n 1, h) :
= (f)(
1
, . . . ,
m
), E(n, h),
(h) ( A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
c5.
there exist
h K(n 1) : h _ k
p1
,
a positive integer m,
a function y whose domain is 1, . . . , m such that for each j = 1 . . . m
y
j
1 var(h), and for each , = 1 . . . m ,= y
,= y
,
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n 1),
E(n 1)
such that
c(n 1, h, m, y, , ),
= (y
1
:
1
, . . . , y
m
:
m
, ), E(n, h).
If c1. holds then ( and
k
x
i
/t = so a1. holds.
If c2. holds then var(k), ,= x
i
,
k
x
i
/t = so a2. holds.
If c3. holds then there exist h K(n): h _ k
p1
_ k, a positive integer m,
,
1
, . . . ,
m
E(n, h) such that = ()(
1
, . . . ,
m
), E(n+1, h), for each (h)
#(h, , ) is a function with m arguments, (#(h,
1
, ), . . . , #(h,
m
, )) is a member of
the domain of #(h, , ).
Moreover if h ,= then by 4.7 we can derive there exists a positive integer q p
such that q < n, /(n; h; x
1
:
1
, . . . , x
q
:
q
). Suppose p = i q, this would imply
that q = p, so h = k
p
= k. But h _ k
p1
also holds. So dom(h) = 1, . . . , p and
dom(h) dom(k
p1
) 1, . . . , p 1. This is a contradiction, so we must have i > q.
Therefore we have h = (h ,= i > q) and
k
x
i
/t = . This implies that a3. is
satised.
If c4. holds then there exist h K(n): h _ k
p1
_ k, f T, a positive integer m,
1
, . . . ,
m
E(n, h) such that = (f)(
1
, . . . ,
m
), E(n + 1, h),
for each (h) A
f
(#(h,
1
, ), . . . , #(h,
m
, )).
Moreover if h ,= then by 4.7 we can derive there exists a positive integer q p
such that q < n, /(n; h; x
1
:
1
, . . . , x
q
:
q
). Suppose p = i q, this would imply
that q = p, so h = k
p
= k. But h _ k
p1
also holds. So dom(h) = 1, . . . , p and
dom(h) dom(k
p1
) 1, . . . , p 1. This is a contradiction, so we must have i > q.
Therefore we have h = (h ,= i > q) and
k
x
i
/t = . This implies that a4. is
satised.
A dierent approach to logic 135
If c5. holds then there exist h K(n): h _ k
p1
_ k, a positive integer m, E(n),
a function y whose domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var(h)
and for each , = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n);
such that
c(n, h, m, y, , ), = (y
1
:
1
, . . . , y
m
:
m
, ), E(n + 1, h).
Moreover if h ,= then by 4.7 we can derive there exists a positive integer q p
such that q < n, /(n; h; x
1
:
1
, . . . , x
q
:
q
). Suppose p = i q, this would imply
that q = p, so h = k
p
= k. But h _ k
p1
also holds. So dom(h) = 1, . . . , p and
dom(h) dom(k
p1
) 1, . . . , p 1. This is a contradiction, so we must have i > q.
Therefore we have h = (h ,= i > q) and
k
x
i
/t = . This implies that a5. is
satised.
Lets examine the case where E
b
(n +1, k). This implies E
b
(n + 1, k),
k K(n)
+
.
We have k = k
p
= k
p1
+(x
p
,
p
), k
p1
K(n),
p
E
s
(n, k
p1
), x
p
1var(k
p1
).
Therefore = x
p
var(k).
If i = p we have = x
i
and
k
x
i
/t = t.
If i < p we have ,= x
i
and
k
x
i
/t = .
This implies that a2. is satised.
We now consider the case where E
c
(n +1, k). This implies E
c
(n + 1, k),
k K(n).
There exist a positive integer m and ,
1
, . . . ,
m
E(n, k) such that
= ()(
1
, . . . ,
m
);
for each (k) #(k, , ) is a function with m arguments and
(#(k,
1
, ), . . . , #(k,
m
, )) is a member of its domain.
We have also /(n; k; x
1
:
1
, . . . , x
p
:
p
). We can dene
k
x
i
/t, (
j
)
k
x
i
/t and
it results
k
x
i
/t = (
k
x
i
/t)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) .
This implies that a3. is satised.
The case where E
d
(n +1, k) is similar. In fact this implies E
d
(n + 1, k),
k K(n).
There exist f T, a positive integer m and
1
, . . . ,
m
E(n, k) such that
= (f)(
1
, . . . ,
m
);
for each (k) A
f
(#(k,
1
, ), . . . , #(k,
m
, )) holds true.
136 M. Avon
We have also /(n; k; x
1
:
1
, . . . , x
p
:
p
). We can dene (
j
)
k
x
i
/t and it results
k
x
i
/t = (f)((
1
)
k
x
i
/t, . . . , (
m
)
k
x
i
/t) .
This implies that a4. is satised.
Finally we examine the case where E
e
(n +1, k). This implies E
e
(n + 1, k),
k K(n). There exist
a positive integer m,
a function y whose domain is 1, . . . , m such that for each j = 1 . . . m
y
j
1 var(k), and for each , = 1 . . . m ,= y
,= y
,
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n),
E(n)
such that = (y
1
:
1
, . . . , y
m
:
m
, ) and c(n, k, m, y, , ).
We have /(n; k; x
1
:
1
, . . . , x
p
:
p
).
Let k
1
= k + (y
1
,
1
), and if m > 1 for each j = 1 . . . m1 k
j+1
= k
j
+ (y
j+1
,
j+1
).
In our assumptions (
1
)
k
x
i
/t is dened, if m > 1 then for each j = 2 . . . m
(
j
)
k
j1
x
i
/t is dened, and nally that
k
m
x
i
/t is dened. It results
k
x
i
/t = (y
1
: (
1
)
k
x
i
/t, . . . , y
m
: (
m
)
k
m1
x
i
/t,
k
m
x
i
/t) .
This implies that a5. is satised.
Another step has been completed. We maintain the assumption that E(n + 1, k)
is such that V
b
(t) V
b
() = . To go on with the next step we assume h K(n + 1) is
such that k
i
_ h.
We know there exist a positive integer u such that u < n + 1, w
1
, . . . , w
u
1 such
that w
,= w
for ,= ,
1
, . . . ,
u
E such that /(n + 1; h; w
1
:
1
, . . . , w
u
:
u
).
By lemma 4.8 we know that i u and for each j = 1 . . . i w
j
= x
j
,
j
=
j
.
If i < u then we assume for each j = i + 1 . . . u w
j
/ V
b
(t), V
b
(t) V
b
(
j
) = . We also
assume E(n + 1, h).
We need to show that
k
x
i
/t =
h
x
i
/t.
We have just seen that one of the following ve conditions holds:
a1. ( and
k
x
i
/t = .
a2. var(k), = x
i
k
x
i
/t = t, ,= x
i
k
x
i
/t = .
A dierent approach to logic 137
a3. there exist K(n): _ k, a positive integer m,
,
1
, . . . ,
m
E(n, ) such that = ()(
1
, . . . ,
m
), E(n + 1, ), for each
() #(, , ) is a function with m arguments, (#(,
1
, ), . . . , #(,
m
, ))
is a member of the domain of #(, , ).
If ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(), since we have
V
b
(t) V
b
() V
b
(t) V
b
() = ,
we can dene
x
i
/t, and similarly we can dene (
j
)
x
i
/t, and it results
k
x
i
/t = (
x
i
/t)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) .
Otherwise (when = or ,= i > q)
k
x
i
/t = .
a4. there exist K(n): _ k, f T, a positive integer m,
1
, . . . ,
m
E(n, )
such that = (f)(
1
, . . . ,
m
), E(n + 1, ),
for each () A
f
(#(,
1
, ), . . . , #(,
m
, )).
If ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(), since we have
V
b
(t) V
b
(
j
) V
b
(t) V
b
() = ,
we can dene (
j
)
x
i
/t, and it results
k
x
i
/t = (f)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) .
Otherwise (when = or ,= i > q)
k
x
i
/t = .
a5. there exist K(n): _ k, a positive integer m, E(n), a function y whose
domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var() and for each
, = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n);
such that
c(n, , m, y, , ), = (y
1
:
1
, . . . , y
m
:
m
, ), E(n + 1, ).
If ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
).
Suppose i q, or in other words x
i
var().
We dene
1
= + (y
1
,
1
), and if m > 1 for each j = 1 . . . m 1
j+1
=
j
+ (y
j+1
,
j+1
).
We have
1
E(n, ), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore (
1
)
x
i
/t is
dened;
for each j = 1 . . . m1
j
K(n) and by 4.7
/(n;
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
for each = 1 . . . j y
V
b
() so y
/ V
b
(t),
138 M. Avon
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n,
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
j
x
i
/t is dened;
m
K(n) and by 4.7
/(n;
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
for each = 1 . . . m y
V
b
() so y
/ V
b
(t),
for each = 1 . . . m V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
E(n,
m
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
m
x
i
/t is dened;
it results
k
x
i
/t = (y
1
: (
1
)
x
i
/t, . . . , y
m
: (
m
)
m1
x
i
/t,
m
x
i
/t) .
Otherwise (when = or ,= i > q)
k
x
i
/t = .
Given that E(n, h + 1) we have also to accept that one of the following ve
conditions holds:
d1. ( and
h
x
i
/t = .
d2. var(h), = x
i
h
x
i
/t = t, ,= x
i
h
x
i
/t = .
d3. there exist K(n): _ h, a positive integer r,
,
1
, . . . ,
r
E(n, ) such that = ()(
1
, . . . ,
r
), E(n + 1, ), for each
() #(, , ) is a function with r arguments, (#(,
1
, ), . . . , #(,
r
, )) is
a member of the domain of #(, , ).
If ,= by 4.7 we can derive there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
).
If i q
, or in other words x
i
var(), since we have
V
b
(t) V
b
() V
b
(t) V
b
() = ,
we can dene
x
i
/t, and similarly we can dene (
j
)
x
i
/t, and it results
h
x
i
/t = (
x
i
/t)((
1
)
x
i
/t, . . . , (
r
)
x
i
/t) .
Otherwise (when = or ,= i > q
)
h
x
i
/t = .
d4. there exist K(n): _ h, g T, a positive integer r,
1
, . . . ,
r
E(n, ) such
that = (g)(
1
, . . . ,
r
), E(n + 1, ),
for each () A
g
(#(,
1
, ), . . . , #(,
r
, )).
If ,= by 4.7 we can derive there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
).
If i q
, or in other words x
i
var(), since we have
V
b
(t) V
b
(
j
) V
b
(t) V
b
() = ,
A dierent approach to logic 139
we can dene (
j
)
x
i
/t, and it results
h
x
i
/t = (g)((
1
)
x
i
/t, . . . , (
r
)
x
i
/t) .
Otherwise (when = or ,= i > q
)
h
x
i
/t = .
d5. there exist K(n): _ h, a positive integer r, E(n), a function z whose
domain is 1, . . . , r such that for each j = 1 . . . r z
j
1 var() and for each
, = 1 . . . r ,= z
,= z
;
a function whose domain is 1, . . . , r such that for each j = 1 . . . r
j
E(n);
such that
c(n, , r, z, , ), = (z
1
:
1
, . . . , z
r
:
r
, ), E(n + 1, ).
If ,= by 4.7 we can derive there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
).
Suppose i q
, or in other words x
i
var().
We dene
1
= + (z
1
,
1
), and if r > 1 for each j = 1 . . . r 1
j+1
=
j
+ (z
j+1
,
j+1
).
We have
1
E(n, ), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore (
1
)
x
i
/t is
dened;
for each j = 1 . . . r 1
j
K(n) and by 4.7
/(n;
j
; w
1
:
1
, . . . , w
q
:
q
, z
1
:
1
, . . . , z
j
:
j
),
for each = 1 . . . j z
V
b
() so z
/ V
b
(t),
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n,
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
j
x
i
/t is dened;
r
K(n) and by 4.7
/(n;
r
; w
1
:
1
, . . . , w
q
:
q
, z
1
:
1
, . . . , z
r
:
r
),
for each = 1 . . . r z
V
b
() so z
/ V
b
(t),
for each = 1 . . . r V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
E(n,
r
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
r
x
i
/t is dened;
it results
h
x
i
/t = (z
1
: (
1
)
x
i
/t, . . . , z
r
: (
r
)
r1
x
i
/t,
r
x
i
/t) .
Otherwise (when = or ,= i > q
)
h
x
i
/t = .
If a1. occurs then d1. also occurs and
k
x
i
/t = =
h
x
i
/t.
If a2. occurs then d2. also occurs and = x
i
h
x
i
/t = t =
k
x
i
/t,
,= x
i
h
x
i
/t = =
k
x
i
/t.
We now consider the case where a3. occurs. As a consequence d3. occurs too.
There exist K(n): _ k, a positive integer m,
,
1
, . . . ,
m
E(n, ) such that = ()(
1
, . . . ,
m
), E(n+1, ), for each ()
140 M. Avon
#(, , ) is a function with m arguments, (#(,
1
, ), . . . , #(,
m
, )) is a member of
the domain of #(, , ).
There exist K(n): _ h, a positive integer r,
,
1
, . . . ,
r
E(n, ) such that = ()(
1
, . . . ,
r
), E(n +1, ), for each ()
#(, , ) is a function with r arguments, (#(,
1
, ), . . . , #(,
r
, )) is a member of
the domain of #(, , ).
We have
()(
1
, . . . ,
m
) = = ()(
1
, . . . ,
r
) ,
and therefore r = m, = and for each j = 1 . . . m
j
=
j
.
It follows that K(n), _ h, ,
1
, . . . ,
m
E(n, ), E(n + 1, ), for each
() #(, , ) is a function with m arguments, (#(,
1
, ), . . . , #(,
m
, )) is a
member of the domain of #(, , ).
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose ,= and there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene
x
i
/t, (
j
)
x
i
/t. Moreover K(n),
i
= h
i
=
i
_ .
We know that /(n; ; w
1
:
1
, . . . , w
q
:
q
). Lemma 4.8 conrms that i q
and for
each j = 1 . . . i w
j
= x
j
,
j
=
j
.
Our assumptions also ensure that if i < q
w
j
/ V
b
(t),
V
b
(t) V
b
(
j
) = . We have also seen that ,
1
, . . . ,
m
E(n, ).
So by the inductive hypothesis we obtain
x
i
/t =
x
i
/t, (
j
)
x
i
/t = (
j
)
x
i
/t ,
and then
k
x
i
/t = (
x
i
/t)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
= (
x
i
/t)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
h
x
i
/t .
We now need to consider another subcase of our a3. and d3. case.
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose = or ,= and there doesnt exist a positive integer q
u such
that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene
x
i
/t, (
j
)
x
i
/t. Moreover K(n),
,
1
, . . . ,
m
E(n, ).
A dierent approach to logic 141
If = then clearly x
i
/ var(). Otherwise we know there exists a positive integer
q
u such that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
). Clearly i > q
, so for each
j = 1 . . . q
w
j
= x
j
and x
i
/ var() holds in this case too.
Therefore
k
x
i
/t = (
x
i
/t)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
= ()(
1
, . . . ,
m
) = =
h
x
i
/t .
We turn to consider a third subcase of our a3. and d3. case.
Suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose ,= and there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene
x
i
/t, (
j
)
x
i
/t. Moreover K(n),
,
1
, . . . ,
m
E(n, ).
If = then clearly x
i
/ var(). Otherwise we know there exists a positive integer
q p such that q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
). Clearly i > q so x
i
/ var() still
holds.
Thus we get
k
x
i
/t = = ()(
1
, . . . ,
m
) =
= (
x
i
/t)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
h
x
i
/t .
There is still another subcase to consider.
Suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose = or ,= and there doesnt exist a positive integer q
u such
that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Here clearly
k
x
i
/t = =
h
x
i
/t.
Lets turn to the case where a4. holds, and accordingly d4. holds too.
There exist K(n): _ k, f T, a positive integer m,
1
, . . . ,
m
E(n, ) such
that = (f)(
1
, . . . ,
m
), E(n + 1, ), for each ()
A
f
(#(,
1
, ), . . . , #(,
m
, )).
There exist K(n): _ h, g T, a positive integer r,
1
, . . . ,
r
E(n, ) such that
= (g)(
1
, . . . ,
r
), E(n + 1, ), for each () A
g
(#(,
1
, ), . . . , #(,
r
, )).
We have
(f)(
1
, . . . ,
m
) = = (g)(
1
, . . . ,
r
) ,
and therefore g = f, r = m, for each j = 1 . . . m
j
=
j
.
142 M. Avon
It follows that K(n), _ h,
1
, . . . ,
m
E(n, ), E(n + 1, ), for each
() A
f
(#(,
1
, ), . . . , #(,
m
, )).
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose ,= and there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene (
j
)
x
i
/t. Moreover K(n),
i
= h
i
=
i
_ .
We know that /(n; ; w
1
:
1
, . . . , w
q
:
q
). Lemma 4.8 conrms that i q
and for
each j = 1 . . . i w
j
= x
j
,
j
=
j
.
Our assumptions also ensure that if i < q
w
j
/ V
b
(t),
V
b
(t) V
b
(
j
) = . We have also seen that
1
, . . . ,
m
E(n, ).
So by the inductive hypothesis we obtain (
j
)
x
i
/t = (
j
)
x
i
/t and then
k
x
i
/t = (f)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
= (f)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
h
x
i
/t .
We now need to consider another subcase of our a4. and d4. case.
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose = or ,= and there doesnt exist a positive integer q
u such
that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene (
j
)
x
i
/t. Moreover K(n),
1
, . . . ,
m
E(n, ).
If = then clearly x
i
/ var(). Otherwise we know there exists a positive integer
q
u such that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
). Clearly i > q
, so for each
j = 1 . . . q
w
j
= x
j
and x
i
/ var() holds in this case too.
Therefore
k
x
i
/t = (f)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
= (f)(
1
, . . . ,
m
) = =
h
x
i
/t .
We turn to consider a third subcase of our a4. and d4. case.
Suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose ,= and there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene (
j
)
x
i
/t. Moreover K(n),
1
, . . . ,
m
E(n, ).
A dierent approach to logic 143
If = then clearly x
i
/ var(). Otherwise we know there exists a positive integer
q p such that q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
). Clearly i > q so x
i
/ var() still
holds.
Thus we get
k
x
i
/t = = (f)(
1
, . . . ,
m
) =
= (f)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) =
h
x
i
/t .
There is still another subcase to consider.
Suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose = or ,= and there doesnt exist a positive integer q
u such
that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Here clearly
k
x
i
/t = =
h
x
i
/t.
Finally we examine the case where a5. holds, and accordingly d5. also occurs.
There exist K(n): _ k, a positive integer m, E(n), a function y whose
domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var() and for each
, = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n);
such that
c(n, , m, y, , ), = (y
1
:
1
, . . . , y
m
:
m
, ), E(n + 1, ).
There exist K(n): _ h, a positive integer r, E(n), a function z whose domain
is 1, . . . , r such that for each j = 1 . . . r z
j
1 var() and for each , = 1 . . . r
,= z
,= z
;
a function whose domain is 1, . . . , r such that for each j = 1 . . . r
j
E(n);
such that
c(n, , r, z, , ), = (z
1
:
1
, . . . , z
r
:
r
, ), E(n + 1, ).
We have E
e
(n + 1, ) and
(y
1
:
1
, . . . , y
m
:
m
, ) = = (z
1
:
1
, . . . , z
r
:
r
, ) .
Therefore r = m, z = y, = and = .
It follows that K(n), _ h, for each j = 1 . . . m y
j
1var(); c(n, , m, y, , ),
E(n + 1, ).
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose ,= and there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
144 M. Avon
Consider that K(n) and we can dene (
1
)
x
i
/t. Moreover K(n),
i
= h
i
=
i
_ .
We know that /(n; ; w
1
:
1
, . . . , w
q
:
q
). Lemma 4.8 conrms that i q
and for
each j = 1 . . . i w
j
= x
j
,
j
=
j
.
Our assumptions also ensure that if i < q
w
j
/ V
b
(t),
V
b
(t) V
b
(
j
) = . We have also seen that
1
E(n, ).
So by the inductive hypothesis we obtain
(
1
)
x
i
/t = (
1
)
x
i
/t .
Now suppose m > 1 and let j = 1 . . . m 1. It results
j
K(n) and we can dene
(
j+1
)
j
x
i
/t. Moreover
j
K(n), (
j
)
i
=
i
= h
i
=
i
_
j
.
We know that /(n;
j
; w
1
:
1
, . . . , w
q
:
q
, y
1
:
1
, . . . , y
j
:
j
).
For each = 1 . . . j y
/ V
b
(t), V
b
(t) V
b
(
) = . Moreover
j+1
E(n,
j
). Therefore
(
j+1
)
j
x
i
/t = (
j+1
)
j
x
i
/t .
We still need to show that
m
x
i
/t =
m
x
i
/t.
To this end we see that
m
K(n) and we can dene
m
x
i
/t. Moreover
m
K(n), (
m
)
i
=
i
= h
i
=
i
_
m
.
We know that /(n;
m
; w
1
:
1
, . . . , w
q
:
q
, y
1
:
1
, . . . , y
m
:
m
).
For each = 1 . . . m y
/ V
b
(t), V
b
(t) V
b
(
) = . Moreover E(n,
m
). Therefore
m
x
i
/t =
m
x
i
/t .
Finally we can establish
k
x
i
/t = (y
1
: (
1
)
x
i
/t, . . . , y
m
: (
m
)
m1
x
i
/t,
m
x
i
/t) =
= (y
1
: (
1
)
x
i
/t, . . . , y
m
: (
m
)
m1
x
i
/t,
m
x
i
/t) =
h
x
i
/t .
We now need to consider another subcase of our a5. and d5. case.
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose = or ,= and there doesnt exist a positive integer q
u such
that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene (
1
)
x
i
/t. Moreover K(n),
1
E(n, ).
If = then clearly x
i
/ var(). Otherwise we know there exists a positive integer
q
u such that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
). Clearly i > q
, so for each
j = 1 . . . q
w
j
= x
j
and x
i
/ var() holds in this case too.
We obtain that (
1
)
x
i
/t =
1
.
A dierent approach to logic 145
Now suppose m > 1 and let j = 1 . . . m 1. It results
j
K(n) and we can dene
(
j+1
)
j
x
i
/t. Recall that c(n, , m, y, , ) holds, so
j
K(n),
j+1
E(n,
j
). For
each = 1 . . . j y
,= x
i
. This implies
that x
i
/ var(
j
).
Therefore (
j+1
)
j
x
i
/t =
j+1
.
We still need to show that
m
x
i
/t = .
It results
m
K(n) and we can dene
m
x
i
/t. Recall that c(n, , m, y, , )
holds, so
m
K(n), E(n,
m
). For each = 1 . . . m y
,= x
i
. This implies that x
i
/ var(
m
).
Therefore
m
x
i
/t = .
Finally we establish
k
x
i
/t = (y
1
: (
1
)
x
i
/t, . . . , y
m
: (
m
)
m1
x
i
/t,
m
x
i
/t) =
= (y
1
:
1
, . . . , y
m
:
m
, ) = =
h
x
i
/t .
We turn to consider a third subcase of our a5. and d5. case.
Suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose ,= and there exists a positive integer q
u such that q
< n,
/(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Consider that K(n) and we can dene (
1
)
x
i
/t. Moreover K(n), and
because of c(n, , m, y, , ) we have
1
E(n, ).
If = then clearly x
i
/ var(). Otherwise we know there exists a positive integer
q p such that q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
). Clearly i > q so x
i
/ var() still
holds.
We obtain that (
1
)
x
i
/t =
1
.
Now suppose m > 1 and let j = 1 . . . m 1. It results
j
K(n) and we can dene
(
j+1
)
j
x
i
/t. Recall that c(n, , m, y, , ) holds, so
j
K(n),
j+1
E(n,
j
). For
each = 1 . . . j y
,= x
i
. This
implies that x
i
/ var(
j
).
Therefore (
j+1
)
j
x
i
/t =
j+1
.
We still need to show that
m
x
i
/t = .
It results
m
K(n) and we can dene
m
x
i
/t. Recall that c(n, , m, y, , ) holds,
so
m
K(n), E(n,
m
). For each = 1 . . . my
,= x
i
. This implies that x
i
/ var(
m
).
Therefore
m
x
i
/t = .
146 M. Avon
Finally we establish
h
x
i
/t = (y
1
: (
1
)
x
i
/t, . . . , y
m
: (
m
)
m1
x
i
/t,
m
x
i
/t) =
= (y
1
:
1
, . . . , y
m
:
m
, ) = =
k
x
i
/t .
There is still another subcase to consider.
Suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Also suppose = or ,= and there doesnt exist a positive integer q
u such
that q
< n, /(n; ; w
1
:
1
, . . . , w
q
:
q
), i q
.
Here clearly
k
x
i
/t = =
h
x
i
/t.
Our denition process requires just a nal step. As in the former step, we maintain
the assumption that E(n+1, k) is such that V
b
(t)V
b
() = . In addition we assume
that h K(n + 1) is such that E(n + 1, h), x
i
/ var(h). We want to prove that
k
x
i
/t = .
Because of E(n + 1, k) and V
b
(t) V
b
() = one of the following ve conditions
holds:
a1. ( and
k
x
i
/t = .
a2. var(k), = x
i
k
x
i
/t = t, ,= x
i
k
x
i
/t = .
a3. there exist K(n): _ k, a positive integer m,
,
1
, . . . ,
m
E(n, ) such that = ()(
1
, . . . ,
m
), E(n + 1, ), for each
() #(, , ) is a function with m arguments, (#(,
1
, ), . . . , #(,
m
, ))
is a member of the domain of #(, , ).
If ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(), since we have
V
b
(t) V
b
() V
b
(t) V
b
() = ,
we can dene
x
i
/t, and similarly we can dene (
j
)
x
i
/t, and it results
k
x
i
/t = (
x
i
/t)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) .
Otherwise (when = or ,= i > q)
k
x
i
/t = .
a4. there exist K(n): _ k, f T, a positive integer m,
1
, . . . ,
m
E(n, )
such that = (f)(
1
, . . . ,
m
), E(n + 1, ),
for each () A
f
(#(,
1
, ), . . . , #(,
m
, )).
A dierent approach to logic 147
If ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
).
If i q, or in other words x
i
var(), since we have
V
b
(t) V
b
(
j
) V
b
(t) V
b
() = ,
we can dene (
j
)
x
i
/t, and it results
k
x
i
/t = (f)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) .
Otherwise (when = or ,= i > q)
k
x
i
/t = .
a5. there exist K(n): _ k, a positive integer m, E(n), a function y whose
domain is 1, . . . , m such that for each j = 1 . . . m y
j
1 var() and for each
, = 1 . . . m ,= y
,= y
;
a function whose domain is 1, . . . , m such that for each j = 1 . . . m
j
E(n);
such that
c(n, , m, y, , ), = (y
1
:
1
, . . . , y
m
:
m
, ), E(n + 1, ).
If ,= by 4.7 we can derive there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
).
Suppose i q, or in other words x
i
var().
We dene
1
= + (y
1
,
1
), and if m > 1 for each j = 1 . . . m 1
j+1
=
j
+ (y
j+1
,
j+1
).
We have
1
E(n, ), V
b
(t) V
b
(
1
) V
b
(t) V
b
() = , therefore (
1
)
x
i
/t is
dened;
for each j = 1 . . . m1
j
K(n) and by 4.7
/(n;
j
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
j
:
j
),
for each = 1 . . . j y
V
b
() so y
/ V
b
(t),
for each = 1 . . . j V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
j+1
E(n,
j
), V
b
(t) V
b
(
j+1
) V
b
(t) V
b
() = ,
therefore (
j+1
)
j
x
i
/t is dened;
m
K(n) and by 4.7
/(n;
m
; x
1
:
1
, . . . , x
q
:
q
, y
1
:
1
, . . . , y
m
:
m
),
for each = 1 . . . m y
V
b
() so y
/ V
b
(t),
for each = 1 . . . m V
b
(t) V
b
(
) V
b
(t) V
b
() = ,
E(n,
m
), V
b
(t) V
b
() V
b
(t) V
b
() = ,
therefore
m
x
i
/t is dened;
it results
k
x
i
/t = (y
1
: (
1
)
x
i
/t, . . . , y
m
: (
m
)
m1
x
i
/t,
m
x
i
/t) .
Otherwise (when = or ,= i > q)
k
x
i
/t = .
Since E(n + 1, h) we can apply assumption 2.1.10 to establish that one of the
following ve conditions holds:
148 M. Avon
e1. (.
e2. var(h).
e3.
K(n) : _ h, r positive integer , ,
1
, . . .
r
E(n, ) :
= ()(
1
, . . . ,
r
), E(n + 1, ),
() ( #(, , ) is a function with r arguments,
(#(,
1
, ), . . . , #(,
r
, )) is a member of the domain of #(, , ).
e4.
K(n) : _ h, g T, r positive integer ,
1
, . . .
r
E(n, ) :
= (g)(
1
, . . . ,
r
), E(n + 1, ),
() ( A
g
(#(,
1
, ), . . . , #(,
r
, )).
e5.
there exist
K(n) : _ h,
a positive integer r,
a function z whose domain is 1, . . . , r such that for each j = 1 . . . r
z
j
1 var(), and for each , = 1 . . . r ,= z
,= z
,
a function whose domain is 1, . . . , r such that for each j = 1 . . . r
j
E(n),
E(n)
such that
c(n, , r, z, , ),
= (z
1
:
1
, . . . , z
r
:
r
, ), E(n + 1, ).
If a1. occurs then clearly
k
x
i
/t = .
If a2. occurs then e2. also holds. Since var(h) and x
i
/ var(h) we have ,= x
i
,
so
k
x
i
/t = .
If a3. occurs then e3. also holds. We have
()(
1
, . . . ,
m
) = = ()(
1
, . . . ,
r
) ,
and therefore r = m, = and for each j = 1 . . . m
j
=
j
.
It follows that K(n), _ h, ,
1
, . . . ,
m
E(n, ), E(n + 1, ), for each
() #(, , ) is a function with m arguments, (#(,
1
, ), . . . , #(,
m
, )) is a
member of the domain of #(, , ).
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
A dierent approach to logic 149
Consider that K(n) and we can dene
x
i
/t, (
j
)
x
i
/t. Also consider that
K(n), ,
1
, . . . ,
m
E(n, ), and since _ h, x
i
/ var(). Therefore
x
i
/t = , (
j
)
x
i
/t =
j
.
It follows that
k
x
i
/t = (
x
i
/t)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) = ()(
1
, . . . ,
m
) = .
Now suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Here its easier, as we immediately get
k
x
i
/t = .
If a4. occurs then e4. also holds. We have
(f)(
1
, . . . ,
m
) = = (g)(
1
, . . . ,
r
) ,
and therefore g = f, r = m, for each j = 1 . . . m
j
=
j
.
It follows that K(n), _ h,
1
, . . . ,
m
E(n, ), E(n + 1, ), for each
() A
f
(#(,
1
, ), . . . , #(,
m
, )).
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Consider that K(n) and we can dene (
j
)
x
i
/t. Also consider that K(n),
1
, . . . ,
m
E(n, ), and since _ h, x
i
/ var(). Therefore (
j
)
x
i
/t =
j
.
It follows that
k
x
i
/t = (f)((
1
)
x
i
/t, . . . , (
m
)
x
i
/t) = (f)(
1
, . . . ,
m
) = .
Now suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Here its easier, as we immediately get
k
x
i
/t = .
If a5. occurs then e5. also holds. We have E
e
(n + 1, ) and
(y
1
:
1
, . . . , y
m
:
m
, ) = = (z
1
:
1
, . . . , z
r
:
r
, ) .
Therefore r = m, z = y, = and = .
It follows that K(n), _ h, for each j = 1 . . . m y
j
1var(); c(n, , m, y, , ),
E(n + 1, ).
Suppose ,= and there exists a positive integer q p such that q < n,
/(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Consider that K(n) and we can dene (
1
)
x
i
/t. Also consider that K(n),
1
E(n, ), and since _ h, x
i
/ var(). Therefore (
1
)
x
i
/t =
1
.
Now suppose m > 1 and let j = 1 . . . m 1. It results
j
K(n) and we can dene
(
j+1
)
j
x
i
/t. Moreover
j
K(n),
j+1
E(n,
j
). For each = 1 . . . j y
/ var()
and since x
i
var() we have that y
,= x
i
. This implies that x
i
/ var(
j
).
150 M. Avon
Therefore (
j+1
)
j
x
i
/t =
j+1
.
We still need to show that
m
x
i
/t = .
It results
m
K(n) and we can dene
m
x
i
/t. Recall that c(n, , m, y, , )
holds, so
m
K(n), E(n,
m
). For each = 1 . . . m y
,= x
i
. This implies that x
i
/ var(
m
).
Therefore
m
x
i
/t = .
We conclude
k
x
i
/t = (y
1
: (
1
)
x
i
/t, . . . , y
m
: (
m
)
m1
x
i
/t,
m
x
i
/t) =
= (y
1
:
1
, . . . , y
m
:
m
, ) = .
Now suppose = or ,= and there doesnt exist a positive integer q p such that
q < n, /(n; ; x
1
:
1
, . . . , x
q
:
q
), i q.
Here its easier, as we immediately get
k
x
i
/t = .
The nal step of our denition process has been completed.
A dierent approach to logic 151
5. Proofs and deductive methodology
In chapter 2 we have seen that our language is identied by a 4-tuple (1, T, (, #). In
chapter 3 we have given some denitions which are important with respect to the deduc-
tive methodology. For instance we have dened the set S(k) of sentences with respect to
a context k. A sentence with respect to will simply be called a sentence.
At this point we need to dene what is a proof in our language. To dene this we
need to dene the notions of axiom and rule.
An axiom is a set A such that
A S()
for each A #() holds.
The property for each A #() holds states that axiom A is sound.
Given a positive integer n we indicate with S()
n
the set of all n-tuples (
1
, . . . ,
n
)
for
1
, . . . ,
n
S(). An n-ary rule is a set R S()
n+1
such that
for each (
1
, . . . ,
n
, ) R if #(
1
), . . . , #(
n
) hold then #() holds.
The property for each (
1
, . . . ,
n
, ) R if #(
1
), . . . , #(
n
) hold then #() holds
states that rule R is sound.
Both in the denition of axiom and rule we have included a requirement of soundness.
A deductive system is built on top of a language / = (1, T, (, #), and is identied
by a pair (/, !) where / is a set of axioms in / and ! is a set of rules in /.
Given a language /, T = (/, !) deductive system in /, ,
1
, . . . ,
m
sentences in
/, we say that (
1
, . . . ,
m
) is a proof of in T if and only if
there exists A / such that
1
A;
if m > 1 then for each j = 2 . . . m one of the following holds
there exists A / such that
j
A,
there exist an n-ary rule R ! and i
1
, . . . , i
n
< j such that
(
i
1
, . . . ,
i
n
,
j
) R;
m
= .
Given T = (/, !) deductive system in / and sentence in / we say that is deriv-
able in T and write
D
if and only if there exist
1
, . . . ,
m
sentences in / such that
(
1
, . . . ,
m
) is a proof of in T.
A deductive system T = (/, !) is said to be sound if and only if for each sentence
in / if
D
then #() holds. In the next lemma we easily prove that each of our systems
152 M. Avon
is sound.
Lemma 5.1. Let T = (/, !) be a deductive system in /. Then T is sound.
Proof.
Let be a sentence in /. Suppose
D
. There exist
1
, . . . ,
m
sentences in / such
that (
1
, . . . ,
m
) is a proof of in T. We can show that for each j = 1 . . . m #(
j
)
holds.
There exists A / such that
1
A, so #(
1
) holds.
If m > 1 suppose j = 2 . . . m.
If there exists A / such that
j
A then #(
j
) holds.
Otherwise there exist an n-ary rule R ! and i
1
, . . . , i
n
< j such that
(
i
1
, . . . ,
i
n
,
j
) R .
Since #(
i
1
), . . . , #(
i
n
) all hold then #(
j
) also holds.
We assume that all of these symbols: , , , , , , are in our set T (this is the
same assumption we made in chapter 3). We also add to T the membership predicate
and the equality predicate = (they have both been explained at the beginning of chapter
2).
We now need to list a set of axioms and rules that can be used in every language with
the aforementioned symbols within the set T. For every axiom/rule we rst prove a result
which ensures the soundness of the axiom/rule and then dene properly the axiom/rule
itself.
In our proofs well frequently use the following simple result.
Lemma 5.2. Let S be a set and q, r be functions over S such that for each S q()
and r() are true (in these assumptions q, r can be called predicates over S). Then
P
(x
1
) and try to prove for each S q().
Let S, clearly q() x
1
, so q() is true.
A dierent approach to logic 153
Conversely we suppose for each S q() and try to prove P
(x
1
).
Let x x
1
, there exists S such that x = q() is true.
We suppose P
(x
1
) and try to prove there exists S q().
There exists x in x
1
such that (x is true). There exists S such that x = q(), therefore
q() is true.
Conversely we suppose there exists S q() and try to prove P
(x
1
).
Clearly q() x
1
and q() is true, so P
(x
1
) is proved.
Now, to prove the other result, let x
1
= q()[ S, r().
We suppose P
(x
1
) and try to prove for each S if r() then q().
Let S and assume r(), clearly q() x
1
, so q() is true.
Conversely we suppose for each S if r() then q() and try to prove P
(x
1
).
Let x x
1
, there exists S such that r() and x = q() is true.
We suppose P
(x
1
) and try to prove there exists S : r() and q().
There exists x in x
1
such that x is true. So there exists S such that r() and x = q(),
therefore q() is true.
Conversely we suppose there exists S : r() and q() and try to prove P
(x
1
).
Clearly q() x
1
and q() is true, so P
(x
1
) is proved.
The rst rule we introduce is based on lemma 3.8. In fact that lemma allows us to
create a rule R
3.8
which is the set of all 3-tuples
[x
1
:
1
, . . . , x
m
:
m
, ()(,
1
)],
[x
1
:
1
, . . . , x
m
:
m
, ()(,
2
)],
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(
1
,
2
))]
such that
m is a positive integer, x
1
, . . . , x
m
1, x
i
,= x
j
for i ,= j,
1
, . . . ,
m
E,
H[x
1
:
1
, . . . , x
m
:
m
],
,
1
,
2
S(k[x
1
:
1
, . . . , x
m
:
m
]).
We continue the list of our rules with another simple one.
Lemma 5.3. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , S(k).
Under these assumptions we have
()(, ), ()(, ), ()(, ) S(k),
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S(),
154 M. Avon
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, ()(, )]) then
#([x
1
:
1
, . . . , x
m
:
m
, ()(, )]) and #([x
1
:
1
, . . . , x
m
:
m
, ()(, )]).
Proof.
We have
P
(P
(P
(P
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , S(k).
Under these assumptions we have
A dierent approach to logic 155
()(, ), () (()(, ), ) , () (()(, ), ) S(k),
[x
1
:
1
, . . . , x
m
:
m
, () (()(, ), )] S(),
[x
1
:
1
, . . . , x
m
:
m
, () (()(, ), )] S().
Moreover #([x
1
:
1
, . . . , x
m
:
m
, () (()(, ), )]) and
#([x
1
:
1
, . . . , x
m
:
m
, () (()(, ), )]) are both true.
Proof.
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, () (()(, ), )]) as follows:
P
(P
(P
(P
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , , S(k).
Under these assumptions we have
()(, ), ()(, ), ()(, ) S(k),
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S(),
156 M. Avon
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S().
Moreover if
#([x
1
:
1
, . . . , x
m
:
m
, ()(, )]),
#([x
1
:
1
, . . . , x
m
:
m
, ()(, )])
then #([x
1
:
1
, . . . , x
m
:
m
, ()(, )]).
Proof.
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, ()(, )]) as follows:
P
(P
(P
(P
[x
1
:
1
, . . . , x
m
:
m
, ()(, )],
[x
1
:
1
, . . . , x
m
:
m
, ()(, )],
[x
1
:
1
, . . . , x
m
:
m
, ()(, )]
such that
m is a positive integer, x
1
, . . . , x
m
1, x
i
,= x
j
for i ,= j,
1
, . . . ,
m
E,
H[x
1
:
1
, . . . , x
m
:
m
],
, , S(k[x
1
:
1
, . . . , x
m
:
m
]).
Lemma 5.6. Let m be a positive integer. Let x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j.
Let
1
, . . . ,
m+1
E and assume H[x
1
:
1
, . . . , x
m+1
:
m+1
].
Dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
]. Of course H[x
1
:
1
, . . . , x
m
:
m
] also holds
and we dene h = k[x
1
:
1
, . . . , x
m
:
m
]. Let S(h).
A dierent approach to logic 157
Let t E(h) such that (h) #(h, t, ) #(h,
m+1
, ).
Let t
, ) #(h,
m+1
, ).
Let S(k) such that V
b
(t) V
b
() = , V
b
(t
) V
b
() = .
Then we can dene
k
x
m+1
/t,
k
x
m+1
/t
))] S()
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t
)] S().
Moreover if
#([x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)])
#([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(t, t
))])
then #([x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t
)]).
Proof.
We dene k
0
= and for each i = 1 . . . m+ 1 k
i
= k[x
1
:
1
, . . . , x
i
:
i
].
We saw in remark 3.3 that for each i = 1 . . . m+ 1
i
E
s
(k
i1
), k
i
= k
i1
+ (x
i
,
i
), dom(k
i
) = 1, . . . , i .
There exists a positive integer n such that k K(n), E(n, k) and for each
i = 1 . . . m+1
i
E
s
(n1, k
i1
). Clearly k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] = k
m+1
also
holds, so we have /(n; k; x
1
:
1
, . . . , x
m+1
:
m+1
).
Moreover h = k
m
so t E(k
m
) is such that (k
m
) #(k
m
, t, ) #(k
m
,
m+1
, ).
We have V
b
(t) 1 var(k
m
) = 1 x
1
, . . . , x
m
, so for each j = 1 . . . m x
j
/ V
b
(t).
Therefore we can dene
k
x
m+1
/t E(kx
m+1
/t) = E(h), and clearly the same
holds for t
, so we can dene
k
x
m+1
/t
E(kx
m+1
/t
) = E(h).
By denition 4.16 we know that for each (h) there exists (k) such that
#(h,
k
x
m+1
/t, ) = #(k, , ). Since #(k, , ) is true or false then so is
#(h,
k
x
m+1
/t, ). Therefore
k
x
m+1
/t S(h).
Clearly the same holds for t
, so
k
x
m+1
/t
S(h).
We can derive that ()(,
k
x
m+1
/t), ()(,
k
x
m+1
/t
) S(h).
Furtherly, (=)(t, t
)) S(h). Therefore
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)] S()
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(t, t
))] S()
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t
)] S().
Suppose the following both hold
a. #([x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)])
b. #([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(t, t
))])
158 M. Avon
We can rewrite a. like this:
P
(#(h, ()(,
k
x
m+1
/t), )[ (h))
P
(P
(#(h, , ), #(h,
k
x
m+1
/t, )) [ (h)) .
And we can rewrite b. like this:
P
)), )[ (h))
P
(P
), )) [ (h))
P
(P
(#(h, , ), P
=
(#(h, t, ), #(h, t
, ))) [ (h)) .
We have to show #([x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t
(P
(#(h, , ), #(h,
k
x
m+1
/t
, )) [ (h)) .
In other words we need to show that for each (h)
if #(h, , ) then #(h,
k
x
m+1
/t
, ).
Let = (u, r) (h) and assume #(h, , ).
We have dom() = dom(h) = dom(k
m
), so
/dom(k
m
)
= . Lets dene
0
= ,
for each j = 1 . . . m
j
=
j1
+ (u
j
, r
j
),
m+1
=
m
+ (x
m+1
, #(k
m
, t, )) .
Because of a. #(h,
k
x
m+1
/t, ) holds, so #(k, ,
m+1
) holds too.
Because of b. #(k
m
, t
, ) = #(h, t
, ) = #(h, t, ) = #(k
m
, t, ).
Therefore #(h,
k
x
m+1
/t
, ) holds too.
Lemma 5.6 allows us to create a rule R
5.6
which is the set of all 3-tuples
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)],
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(t, t
))],
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t
)]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(h),
t E(h), (h) #(h, t, ) #(h,
m+1
, ),
t
, ) #(h,
m+1
, ),
S(k), V
b
(t) V
b
() = , V
b
(t
) V
b
() = .
A dierent approach to logic 159
Lemma 5.7. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , S(k).
Under these assumptions we have
()(, ) S(k),
[x
1
:
1
, . . . , x
m
:
m
, ] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, ]) then #([x
1
:
1
, . . . , x
m
:
m
, ()(, )])
also holds.
Proof.
Suppose #([x
1
:
1
, . . . , x
m
:
m
, ]) holds. It can be rewritten as
P
(#(k, , )[ (k)) .
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, ()(, )]) as
P
(P
(P
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , , E(k), S(k).
Under these assumptions we have
()(, (=)(, )), ()(, (=)(, )), ()(, (=)(, )) S(k);
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))] S();
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))] S();
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))] S().
Moreover if
#([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]),
160 M. Avon
#([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))])
then #([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]).
Proof.
We rewrite #([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]) as
P
(P
(P
(#(k, , ), P
=
(#(k, , ), #(k, , ))) [ (k)) .
Similarly we rewrite #([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]) as
P
(P
(#(k, , ), P
=
(#(k, , ), #(k, , ))) [ (k)) ,
and we rewrite #([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]) as
P
(P
(#(k, , ), P
=
(#(k, , ), #(k, , ))) [ (k)) .
If
#([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]),
#([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))])
both hold, then for each (k) if #(k, , ) then #(k, , ) is equal to #(k, , ),
which is equal to #(k, , ).
This implies that #([x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]) holds.
Lemma 5.8 allows us to create a rule R
5.8
which is the set of all 3-tuples
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))],
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))],
[x
1
:
1
, . . . , x
m
:
m
, ()(, (=)(, ))]
such that
m is a positive integer, x
1
, . . . , x
m
1, x
i
,= x
j
for i ,= j,
1
, . . . ,
m
E,
H[x
1
:
1
, . . . , x
m
:
m
],
, , E(k[x
1
:
1
, . . . , x
m
:
m
]), S(k[x
1
:
1
, . . . , x
m
:
m
]).
Lemma 5.9. Let m be a positive integer. Let x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j.
Let
1
, . . . ,
m+1
E and assume H[x
1
:
1
, . . . , x
m+1
:
m+1
].
Dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
]. Of course H[x
1
:
1
, . . . , x
m
:
m
] also holds
and we dene h = k[x
1
:
1
, . . . , x
m
:
m
]. Let S(h).
Let t E(h) such that (h) #(h, t, ) #(h,
m+1
, ).
Let S(k) such that V
b
(t) V
b
() = .
Then we can dene
k
x
m+1
/t S(h) and furthermore
()( (x
m+1
:
m+1
, )) S(h),
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)] S(),
A dierent approach to logic 161
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)]) then
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))]) .
Proof.
As we have seen in lemma 5.6, in these assumptions we can prove
k
x
m+1
/t S(h).
Since
m+1
E
s
(h), x
m+1
1 var(h), k = h + (x
m+1
,
m+1
) we can apply
lemma 3.1 and obtain that ()( (x
m+1
:
m+1
, )) S(h).
Therefore
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))] S().
Suppose #([x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)]) holds, it can be rewritten
as
P
(#(h, ()(,
k
x
m+1
/t), )[ (h)) ,
P
(P
(#(h, , ), #(h,
k
x
m+1
/t, ))[ (h)) .
We need to prove #([x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))]), and
this can be rewritten
P
(P
(P
(#(h, , ), P
m+1
=
m
+ (x
m+1
, #(h, t, )).
By denition 4.16 it results
m+1
(k) and #(h,
k
x
m+1
/t, ) = #(k, ,
m+1
),
so #(k, ,
m+1
) holds true. Clearly =
m
_
m+1
, so our proof is nished.
Lemma 5.9 allows us to create a rule R
5.9
which is the set of all pairs
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)],
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(h),
t E(h), (h) #(h, t, ) #(h,
m+1
, ),
S(k), V
b
(t) V
b
() = .
162 M. Avon
Lemma 5.10. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , , S(k).
Under these assumptions we have
()(()(, ), ), ()(, ()(, )) S(k),
[x
1
:
1
, . . . , x
m
:
m
, ()(()(, ), )] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(, ))] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, ()(()(, ), )]) then
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(, ))]).
Proof.
We assume #([x
1
:
1
, . . . , x
m
:
m
, ()(()(, ), )]) which can be rewritten
P
(P
(P
(P
(P
(P
(#(k, , ), P
(#(h, [x
m+1
:
m+1
, ()(, )], )[ (h))
P
(#(h, () ( (x
m+1
:
m+1
, ()(, ))) , )[ (h))
P
(P
(P
(P
(#(h, ()(, [x
m+1
:
m+1
, ]), )[ (h))
P
(P
(#(h, , ), #(h, [x
m+1
:
m+1
, ], )) [ (h))
P
(P
(P
(#(h, , ), P
[x
1
:
1
, . . . , x
m
:
m
, [x
m+1
:
m+1
, ()(, )]],
[x
1
:
1
, . . . , x
m
:
m
, ()(, [x
m+1
:
m+1
, ])]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(h) S(k) and S(k).
Lemma 5.12. Let m be a positive integer. Let x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j.
Let
1
, . . . ,
m+1
E and assume H[x
1
:
1
, . . . , x
m+1
:
m+1
].
Dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
]. Of course H[x
1
:
1
, . . . , x
m
:
m
] also holds,
and we dene h = k[x
1
:
1
, . . . , x
m
:
m
]. Let S(k) and S(h) S(k).
Under these assumptions we have
()(, ) S(k),
[x
m+1
:
m+1
, ()(, )] S(h),
() ((x
m+1
:
m+1
, )) S(h),
() (() ((x
m+1
:
m+1
, )) , ) S(h),
[x
1
:
1
, . . . , x
m
:
m
, [x
m+1
:
m+1
, ()(, )]] S(),
[x
1
:
1
, . . . , x
m
:
m
, () (() ((x
m+1
:
m+1
, )) , )] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, [x
m+1
:
m+1
, ()(, )]]) then
#([x
1
:
1
, . . . , x
m
:
m
, () (() ((x
m+1
:
m+1
, )) , )]) .
Proof.
Using remark 3.3 (and the notation in it) we can easily determine that h = k
m
K,
k = k
m+1
= k
m
+(x
m+1
,
m+1
) = h+(x
m+1
,
m+1
),
m+1
E
s
(h), x
m+1
1 var(h).
Clearly ()(, ) S(k) also holds so
[x
m+1
:
m+1
, ()(, )] S(h),
() ((x
m+1
:
m+1
, )) S(h),
() (() ((x
m+1
:
m+1
, )) , ) S(h)
and this implies
[x
1
:
1
, . . . , x
m
:
m
, [x
m+1
:
m+1
, ()(, )]] S(),
[x
1
:
1
, . . . , x
m
:
m
, () (() ((x
m+1
:
m+1
, )) , )] S().
A dierent approach to logic 165
Suppose #([x
1
:
1
, . . . , x
m
:
m
, [x
m+1
:
m+1
, ()(, )]]). It can be rewritten
as
P
(#(h, [x
m+1
:
m+1
, ()(, )], )[ (h))
P
(#(h, () ( (x
m+1
:
m+1
, ()(, ))) , )[ (h))
P
(P
(P
(P
(P
(#(h, () ((x
m+1
:
m+1
, )) , ), #(h, , )) [ (h))
P
(P
(P
[x
1
:
1
, . . . , x
m
:
m
, [x
m+1
:
m+1
, ()(, )]],
[x
1
:
1
, . . . , x
m
:
m
, () (() ((x
m+1
:
m+1
, )) , )]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(k) and S(h) S(k).
The next rule is a variation of the former one.
166 M. Avon
Lemma 5.13. Let m be a positive integer. Let x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j.
Let
1
, . . . ,
m+1
E and assume H[x
1
:
1
, . . . , x
m+1
:
m+1
].
Dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
]. Of course H[x
1
:
1
, . . . , x
m
:
m
] also holds,
and we dene h = k[x
1
:
1
, . . . , x
m
:
m
]. Let S(h), S(k) and S(h) S(k).
Under these assumptions we have
()(, ) S(k),
[x
m+1
:
m+1
, ()(, )] S(h),
() (, [x
m+1
:
m+1
, ()(, )]) S(h),
() ((x
m+1
:
m+1
, )) S(h),
() (() ((x
m+1
:
m+1
, )) , ) S(h),
() (, () (() ((x
m+1
:
m+1
, )) , )) S(h),
[x
1
:
1
, . . . , x
m
:
m
, () (, [x
m+1
:
m+1
, ()(, )])] S(),
[x
1
:
1
, . . . , x
m
:
m
, () (, () (() ((x
m+1
:
m+1
, )) , ))] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, () (, [x
m+1
:
m+1
, ()(, )])]) then
#([x
1
:
1
, . . . , x
m
:
m
, () (, () (() ((x
m+1
:
m+1
, )) , ))]) .
Proof.
Just as in the proof of 5.12 we can derive
()(, ) S(k)
[x
m+1
:
m+1
, ()(, )] S(h),
() ((x
m+1
:
m+1
, )) S(h),
() (() ((x
m+1
:
m+1
, )) , ) S(h).
It clearly follows that
() (, [x
m+1
:
m+1
, ()(, )]) S(h),
() (, () (() ((x
m+1
:
m+1
, )) , )) S(h),
[x
1
:
1
, . . . , x
m
:
m
, () (, [x
m+1
:
m+1
, ()(, )])] S(),
[x
1
:
1
, . . . , x
m
:
m
, () (, () (() ((x
m+1
:
m+1
, )) , ))] S().
Suppose #([x
1
:
1
, . . . , x
m
:
m
, () (, [x
m+1
:
m+1
, ()(, )])]). It can be
rewritten as
P
(#(h, () (, [x
m+1
:
m+1
, ()(, )]) , )[ (h))
P
(P
(#(h, , ), #(h, [x
m+1
:
m+1
, ()(, )], )) [ (h))
P
(P
(#(h, , ), P
(P
(#(h, , ), P
(P
(P
(P
(#(h, , ), P
(#(h, () ((x
m+1
:
m+1
, )) , ), #(h, , ))) [ (h))
P
(P
(#(h, , ), P
(P
[x
1
:
1
, . . . , x
m
:
m
, () (, [x
m+1
:
m+1
, ()(, )])],
[x
1
:
1
, . . . , x
m
:
m
, () (, () (() ((x
m+1
:
m+1
, )) , ))]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(h), S(k) and S(h) S(k).
The next rule recalls the rule of standard logic which is called modus ponens and
can be itself called modus ponens .
Lemma 5.14. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , , S(k).
Under these assumptions we have
()(, ), ()(, ()(, )), ()(, ) S(k),
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(, ))] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, )] S().
Moreover if
#([x
1
:
1
, . . . , x
m
:
m
, ()(, )]),
168 M. Avon
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(, ))])
then #([x
1
:
1
, . . . , x
m
:
m
, ()(, )]).
Proof.
We assume that
#([x
1
:
1
, . . . , x
m
:
m
, ()(, )]),
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(, ))])
both hold.
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, ()(, )]) as
P
(P
(P
(P
(#(k, , ), P
(P
[x
1
:
1
, . . . , x
m
:
m
, ()(, )],
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(, ))],
[x
1
:
1
, . . . , x
m
:
m
, ()(, )]
such that
m is a positive integer, x
1
, . . . , x
m
1, x
i
,= x
j
for i ,= j,
1
, . . . ,
m
E,
H[x
1
:
1
, . . . , x
m
:
m
],
, , S(k[x
1
:
1
, . . . , x
m
:
m
]).
Lemma 5.15. Let m be a positive integer. Let x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j.
Let
1
, . . . ,
m+1
E and assume H[x
1
:
1
, . . . , x
m+1
:
m+1
].
Dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
]. Of course H[x
1
:
1
, . . . , x
m
:
m
] also holds
and we dene h = k[x
1
:
1
, . . . , x
m
:
m
]. Let S(h).
Let t E(h) such that (h) #(h, t, ) #(h,
m+1
, ).
Let S(k) such that V
b
(t) V
b
() = .
A dierent approach to logic 169
Then we can dene
k
x
m+1
/t S(h) and furthermore
()( (x
m+1
:
m+1
, )) S(h),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))]) then
#([x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)]) .
Proof.
As we have seen in lemma 5.6, in these assumptions we can prove
k
x
m+1
/t S(h).
Since
m+1
E
s
(h), x
m+1
1 var(h), k = h + (x
m+1
,
m+1
) we can apply
lemma 3.1 and obtain that ()( (x
m+1
:
m+1
, )) S(h).
Therefore
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)] S().
Suppose #([x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))]) holds, it can be
rewritten
P
(P
(P
(#(h, , ), P
(#(h, ()(,
k
x
m+1
/t), )[ (h)) ,
P
(P
(#(h, , ), #(h,
k
x
m+1
/t, ))[ (h)) .
Let (h) and suppose #(h, , ). We need to show #(h,
k
x
m+1
/t, ).
We have (h) = (kx
m+1
/t), dom() = 1, . . . , m. Let = (u, r) and dene
0
= , for each j = 1 . . . m
j
=
j1
+ (u
j
, r
j
),
m+1
=
m
+ (x
m+1
, #(h, t, )).
By denition 4.16 it results
m+1
(k) and #(h,
k
x
m+1
/t, ) = #(k, ,
m+1
).
Clearly =
m
_
m+1
, so #(k, ,
m+1
) holds and #(h,
k
x
m+1
/t, ) holds too.
Lemma 5.15 allows us to create a rule R
5.15
which is the set of all pairs
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, )))],
[x
1
:
1
, . . . , x
m
:
m
, ()(,
k
x
m+1
/t)]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(h),
170 M. Avon
t E(h), (h) #(h, t, ) #(h,
m+1
, ),
S(k), V
b
(t) V
b
() = .
Lemma 5.16. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
].
Let i = 1 . . . m such that for each j = i . . . m x
j
/ V
b
(
i
). Then
()(x
i
,
i
) S(k),
[x
1
:
1
, . . . , x
m
:
m
, ()(x
i
,
i
)] S(),
#([x
1
:
1
, . . . , x
m
:
m
, ()(x
i
,
i
)]).
Proof.
We have k
i1
K, x
i
1 var(k
i1
),
i
E
s
(k
i1
), k
i
= k
i1
+ (x
i
,
i
).
By lemma 4.13 x
i
E(k
i
) and for each
i
=
i1
+ (x
i
, s) (k
i
)
#(k
i
, x
i
,
i
) = s #(k
i1
,
i
,
i1
).
If i = m then we have proved x
i
E(k).
If i < m, since for each j = i + 1 . . . m x
j
/ V
b
(x
i
), we can apply lemma 3.14 and
derive that x
i
E(k) and for each (k) there exists
i
(k
i
) such that
i
_ and
#(k
i
, x
i
,
i
) = #(k, x
i
, ).
It also results
i
E
s
(k
i1
) and for each j = i . . . m x
j
/ V
b
(
i
). Therefore, by
lemma 3.14,
i
E(k) and for each (k) there exists
i1
(k
i1
) such that
i1
_ and #(k,
i
, ) = #(k
i1
,
i
,
i1
) is a set.
By lemma 3.13 we derive that ()(x
i
,
i
) S(k), and consequently
[x
1
:
1
, . . . , x
m
:
m
, ()(x
i
,
i
)] S() .
Moreover we can rewrite #([x
1
:
1
, . . . , x
m
:
m
, ()(x
i
,
i
)]) as follows
P
(#(k, ()(x
i
,
i
), )[ (k)) ,
P
(P
(#(k, x
i
, ), #(k,
i
, ))[ (k)) .
To show this we have to prove that for each (k) #(k, x
i
, ) belongs to #(k,
i
, ).
We know there exists
i
(k
i
) such that
i
_ and #(k
i
, x
i
,
i
) = #(k, x
i
, ).
We also know there exist
i1
(k
i1
), s #(k
i1
,
i
,
i1
) such that
i
=
i1
+ (x
i
, s) and #(k
i
, x
i
,
i
) = s #(k
i1
,
i
,
i1
).
Furthermore there exists
i1
(k
i1
) such that
i1
_ and
#(k,
i
, ) = #(k
i1
,
i
,
i1
).
We have
i1
=
/dom(k
i1
)
=
i1
, so
#(k, x
i
, ) = #(k
i
, x
i
,
i
) #(k
i1
,
i
,
i1
) = #(k,
i
, ) .
A dierent approach to logic 171
Lemma 5.16 permits us to create an axiom A
5.16
which is the set of all sentences
[x
1
:
1
, . . . , x
m
:
m
, ()(x
i
,
i
)] such that
m is a positive integer, x
1
, . . . , x
m
1, x
,= x
for ,= ,
1
, . . . ,
m
E,
H[x
1
:
1
, . . . , x
m
:
m
],
i = 1 . . . m,
for each j = i . . . m x
j
/ V
b
(
i
).
Lemma 5.17. Let m be a positive integer. Let x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j.
Let
1
, . . . ,
m+1
E and assume H[x
1
:
1
, . . . , x
m+1
:
m+1
].
Dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
]. Of course H[x
1
:
1
, . . . , x
m
:
m
] also holds,
we dene h = k[x
1
:
1
, . . . , x
m
:
m
]. Let S(h), t E(h).
Let E
s
(h) and x
m+1
/ V
b
().
Under these assumptions
()(x
m+1
, ) S(k),
()( (x
m+1
:
m+1
, ()(x
m+1
, ))) S(h),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, ()(x
m+1
, ))))] S(),
()(t,
m+1
) S(h),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t,
m+1
))] S(),
()(t, ) S(h),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t, ))] S().
Moreover if
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, ()(x
m+1
, ))))]) and
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t,
m+1
))])
then #([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t, ))]).
Proof.
We have k = h + (x
m+1
,
m+1
), where h K, x
m+1
1 var(h),
m+1
E
s
(h).
Using lemma 4.13 we can derive that x
m+1
E(k).
Since E
s
(h) and x
m+1
/ V
b
() we can apply lemma 4.12 and obtain that E(k)
and for each = + (x
m+1
, s) (k) #(k, , ) = #(h, , ) is a set.
Therefore, by lemma 3.13 we get ()(x
m+1
, ) S(k).
By lemma 3.1 we obtain ()( (x
m+1
:
m+1
, ()(x
m+1
, ))) S(h).
Clearly this implies that
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, ()(x
m+1
, ))))] S().
Furthermore we have t E(h),
m+1
E
s
(h), so ()(t,
m+1
) S(h). It clearly
follows that [x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t,
m+1
))] S().
172 M. Avon
We have also E
s
(h), so ()(t, ) S(h). It follows that
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t, ))] S().
We now assume
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, ()(x
m+1
, ))))]) and
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t,
m+1
))])
both hold and we try to prove #([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t, ))]).
We can rewrite
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, ()(x
m+1
, ))))])
as
P
(#(h, () (, () ( (x
m+1
:
m+1
, ()(x
m+1
, )))) , ) [ (h)) ,
P
(P
(#(h, , ) , #(h, () ( (x
m+1
:
m+1
, ()(x
m+1
, ))) , )) [ (h)) ,
P
(P
(#(h, , ) , P
(#(k, ()(x
m+1
, ), ) [ (k), _ )) [ (h)) ,
P
({P
(#(h, , ) , P
({P
(#(k, x
m+1
, ), #(k, , )) | (k), })) | (h)}) .
We can rewrite
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t,
m+1
))])
as
P
(P
(P
(#(h, , ), P
(#(h, t, ), #(h,
m+1
, )))[ (h)) .
We can rewrite
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t, ))])
as
P
(P
(P
(#(h, , ), P
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()( (x
m+1
:
m+1
, ()(x
m+1
, ))))],
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t,
m+1
))],
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()(t, ))]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(h),
t E(h),
E
s
(h), x
m+1
/ V
b
().
Lemma 5.18. Let m be a positive integer. Let x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , S(k).
Under these assumptions we have
() (, () (, ()())) , ()() S(k),
[x
1
:
1
, . . . , x
m
:
m
, () (, () (, ()()))] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()()] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, () (, () (, ()()))]) then
#([x
1
:
1
, . . . , x
m
:
m
, ()()]).
Proof.
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, () (, () (, ()()))]) as
P
(P
(P
(#(k, , ), P
(P
(#(k, , ), P
(#(k, , ), P
(P
1
, . . . ,
m
E and assume H[x
1
:
1
, . . . , x
m
:
m
]. Dene k = k[x
1
:
1
, . . . , x
m
:
m
]
and let , S(k).
Under these assumptions we have
() (()(, )) , ()(, ()()) S(k),
[x
1
:
1
, . . . , x
m
:
m
, () (()(, ))] S(),
[x
1
:
1
, . . . , x
m
:
m
, ()(, ()())] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, () (()(, ))]) then
#([x
1
:
1
, . . . , x
m
:
m
, ()(, ()())]).
Proof.
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, () (()(, ))]) as
P
(P
(P
(P
(P
(P
(#(k, , ), P
(#(k, , ), P
(#(k, , ))).
The last condition clearly implies #([x
1
:
1
, . . . , x
m
:
m
, ()(, ()())]).
Lemma 5.19 allows us to create a rule R
5.19
which is the set of all pairs
([x
1
:
1
, . . . , x
m
:
m
, () (()(, ))], [x
1
:
1
, . . . , x
m
:
m
, ()(, ()())])
such that
A dierent approach to logic 175
m is a positive integer, x
1
, . . . , x
m
1, x
i
,= x
j
for i ,= j,
1
, . . . ,
m
E,
H[x
1
:
1
, . . . , x
m
:
m
],
, S(k[x
1
:
1
, . . . , x
m
:
m
]).
Lemma 5.20. Let m be a positive integer. Let x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j.
Let
1
, . . . ,
m+1
E and assume H[x
1
:
1
, . . . , x
m+1
:
m+1
].
Dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
]. Of course H[x
1
:
1
, . . . , x
m
:
m
] also holds,
we dene h = k[x
1
:
1
, . . . , x
m
:
m
]. Let S(h), S(k).
Under these assumptions we have
()((x
m+1
:
m+1
, )) S(h),
() (()((x
m+1
:
m+1
, ))) S(h),
() (, () (()((x
m+1
:
m+1
, )))) S(h),
[x
1
:
1
, . . . , x
m
:
m
, () (, () (()((x
m+1
:
m+1
, ))))] S(),
()() S(k),
()((x
m+1
:
m+1
, ()())) S(h),
() (, ()((x
m+1
:
m+1
, ()()))) S(h),
[x
1
:
1
, . . . , x
m
:
m
, () (, ()((x
m+1
:
m+1
, ()())))] S().
Moreover if #([x
1
:
1
, . . . , x
m
:
m
, () (, () (()((x
m+1
:
m+1
, ))))]) then
#([x
1
:
1
, . . . , x
m
:
m
, () (, ()((x
m+1
:
m+1
, ()())))]) .
Proof.
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, () (, () (()((x
m+1
:
m+1
, ))))]) as
P
(#(h, () (, () (()((x
m+1
:
m+1
, )))) , )[ (h)) ,
P
(P
(P
(#(h, , ), P
(#(h, () ( (x
m+1
:
m+1
, )) , ))) [ (h)) ,
P
(P
(#(h, , ), P
(P
(#(k, , )[ (k), _ ),
for each (h) if #(h, , ) then it is false that (for each (k) such that _
#(k, , ) holds),
for each (h) if #(h, , ) then (there exists (k) such that _ and
#(k, , ) is false).
We can rewrite #([x
1
:
1
, . . . , x
m
:
m
, () (, ()((x
m+1
:
m+1
, ()())))]) as
P
(#(h, () (, ()((x
m+1
:
m+1
, ()()))) , )[ (h)) ,
P
(P
(P
(#(h, , ), P
(P
(#(h, , ), P
(P
(P
[x
1
:
1
, . . . , x
m
:
m
, () (, () (()((x
m+1
:
m+1
, ))))],
[x
1
:
1
, . . . , x
m
:
m
, () (, ()((x
m+1
:
m+1
, ()())))]
such that
m is a positive integer, x
1
, . . . , x
m+1
1, with x
i
,= x
j
for i ,= j,
1
, . . . ,
m+1
E,
H[x
1
:
1
, . . . , x
m+1
:
m+1
];
if we dene k = k[x
1
:
1
, . . . , x
m+1
:
m+1
] and h = k[x
1
:
1
, . . . , x
m
:
m
] then
S(h), S(k).
The next lemma is just a degenerate case of rule 5.12, we probably could modify that
lemma to enclose also this case, but we choose to treat it separately.
Lemma 5.21. Let x
1
1,
1
E and assume H[x
1
:
1
]. Dene k = k[x
1
:
1
]. Let
S(k) and S(k) S(). Under these assumptions we have
()(, ) S(k),
[x
1
:
1
, ()(, )] S(),
() ((x
1
:
1
, )) S(),
() (() ((x
1
:
1
, )) , ) S().
Moreover if #([x
1
:
1
, ()(, )]) then #(() (() ((x
1
:
1
, )) , )).
Proof.
Suppose #([x
1
:
1
, ()(, )]). By denition we have
#(()((x
1
:
1
, ()(, )))) ,
and then
P
(P
(#(() ((x
1
:
1
, ))), #()) ,
P
(P
[x
1
:
1
, ()(, )],
() (() ((x
1
:
1
, )) , )
such that x
1
1,
1
E, H[x
1
:
1
], S(k[x
1
:
1
]) and S(k[x
1
:
1
]) S().
Lemma 5.22. Let , , S(). We have
()(, ()(, )) S(),
()(()(, ), ) S().
Moreover if #(()(, ()(, ))) then #(()(()(, ), )).
Proof.
Suppose #(()(, ()(, ))) holds. It can be rewritten
P
(#(), P
(#(), #())) .
In turn, #(()(()(, ), )) can be rewritten
P
(P
(#(), P
(#(), #())) .
Lemma 5.22 allows us to create a rule R
5.22
which is the set of all pairs
i+1
=
i
+ (x
i+1
,
i+1
) then
m
(k) and for each i = 1 . . . m #(k, x
i
,
m
) =
i
.
Proof.
We rst show that H[x
1
: N, . . . , x
m
: N] holds. Let k
0
= .
First consider that N ( E() and #(, N, ) = #(N) = N is a set. Therefore
N E
s
(), so H[x
1
: N] holds. Let k
1
= k[x
1
: N], clearly var(k
1
) = x
1
.
Suppose m > 1 and let i = 1 . . . m 1. Assume H[x
1
: N, . . . , x
i
: N] and
k
i
= k[x
1
: N, . . . , x
i
: N], var(k
i
) = x
1
, . . . , x
i
. To prove H[x
1
: N, . . . , x
i+1
: N]
we just need to prove N E
s
(k
i
).
There exists a positive integer n such that k
i
K(n), so by lemma 3.9 N E(n, k
i
)
and for each (k
i
) #(k
i
, N, ) = #(N) is a set. So we have proved N E
s
(k
i
), and
A dierent approach to logic 179
it results H[x
1
: N, . . . , x
i+1
: N]. We also dene k
i+1
= k[x
1
: N, . . . , x
i+1
: N], and we
have var(k
i+1
) = x
1
, . . . , x
i+1
.
We have proved that H[x
1
: N, . . . , x
m
: N] holds.
Let i = 1 . . . m. Clearly k
i
= k
i1
+ (x
i
, N). There exists a positive integer n such
that k
i1
K(n), and N E
s
(k
i1
), so N E
s
(n, k
i1
). Moreover x
i
/ var(k
i1
), so
k
i
K(n)
+
.
It follows that x
i
E(n + 1, k
i
) E(k
i
), for each = + (x
i
, s) (k
i
)
#(k
i
, x
i
, ) = s #(k
i1
, N, ) = #(N) = N; V
b
(x
i
) = .
If i < m then for each j = i . . . m1 we can assume x
i
E(k
j
) and for each (k
j
)
#(k
j
, x
i
, ) N.
Clearly k
j+1
= k
j
+(x
j+1
, N). There exists a positive integer n such that x
i
E(n, k
j
),
so k
j
K(n). We have also N E
s
(k
j
), so N E
s
(n, k
j
). Moreover x
j+1
/ var(k
j
), so
k
j+1
K(n)
+
.
Since x
j+1
/ V
b
(x
i
) we have x
i
E(n + 1, k
j+1
) E(k
j+1
). In addition, for each
= + (x
j+1
, s) (k
j+1
) #(k
j+1
, x
i
, ) = #(k
j
, x
i
, ) N.
We have proved that x
i
E(k) and for each (k) #(k, x
i
, ) N.
Let
1
, . . . ,
m
N,
0
= and for each i = 0 . . . m1
i+1
=
i
+ (x
i+1
,
i+1
).
We have
0
= () = (k
0
).
Given i = 0 . . . m 1 we assume
i
(k
i
). We have k
i+1
= k
i
+ (x
i+1
, N) and
there exists a positive integer n such that k
i
K(n) and N E
s
(n, k
i
). Moreover
x
i+1
/ var(k
i
), so k
i+1
K(n)
+
. To prove that
i+1
(k
i+1
) we just need to prove
that
i+1
#(k
i
, N,
i
) = #(N) = N. This is true, of course.
So we have proved that
m
(k).
Now let i = 1 . . . m, we want to show that #(k, x
i
,
m
) =
i
.
We begin by showing that #(k
i
, x
i
,
i
) =
i
. We have k
i
= k
i1
+ (x
i
, N) and
there exists a positive integer n such that k
i1
K(n) and N E
s
(n, k
i1
). Moreover
x
i
/ var(k
i1
), so k
i
K(n)
+
, x
i
E(n + 1, k
i
). We have
i
=
i1
+ (x
i
,
i
) (k
i
)
and #(k
i
, x
i
,
i
) =
i
.
If i < m then let j = i . . . m 1, we assume #(k
j
, x
i
,
j
) =
i
and try to show
#(k
j+1
, x
i
,
j+1
) =
i
. Clearly k
j+1
= k
j
+ (x
j+1
, N). There exists a positive integer n
such that x
i
E(n, k
j
), so k
j
K(n). We have also N E
s
(k
j
), so
N E
s
(n, k
j
). Moreover x
j+1
/ var(k
j
), so k
j+1
K(n)
+
. Since x
j+1
/ V
b
(x
i
) we
have x
i
E(n + 1, k
j+1
).
It results
j+1
=
j
+ (x
j+1
,
j+1
) (k
j+1
) and
#(k
j+1
, x
i
,
j+1
) = #(k
j
, x
i
,
j
) =
i
.
180 M. Avon
To show that expression Th
1
belongs to S() we dene k = k[x : N, y : N, z : N].
By 6.1 we obtain that x, y, z E(k) and for each (k) #(k, x, ), #(k, y, ), #(k, z, )
are all members of N.
Moreover, [ E(k), for each (k) #(k, [, ) = #([) is a function with two
arguments and (#(k, x, ), #(k, y, )), (#(k, y, ), #(k, z, )), (#(k, x, ), #(k, z, )) are
members of its domain.
So, by lemma 3.10, ([)(x, y), ([)(x, z), ([)(y, z) all belong to E(k).
Moreover, for each (k)
#(k, ([)(x, y), ) = #([)(#(k, x, ), #(k, y, )) = N : (#(k, y, ) = #(k, x, ) ) ;
so #(k, ([)(x, y), ) is true or false. Therefore ([)(x, y) S(k). In the same way we
can show that ([)(y, z), ([)(x, z) S(k).
By lemma 3.7 we have
()(([)(x, y), ([)(y, z)) S(k), ()(()(([)(x, y), ([)(y, z)), ([)(x, z)) S(k).
By denition 3.4
[x : N, y : N, z : N, ()(()(([)(x, y), ([)(y, z)), ([)(x, z))] S() .
We have proved Th
1
is a sentence and well now show it has the correct meaning.
By theorem 3.6 #([x : N, y : N, z : N, ()(()(([)(x, y), ([)(y, z)), ([)(x, z))]) is equiv-
alent to
P
(P
(#(k, ()(([)(x, y), ([)(y, z)), ), #(k, ([)(x, z)), ))[ (k)) ,
P
(P
(P
(#(k, ([)(x, y), ), #(k, ([)(y, z), )), #(k, ([)(x, z)), ))[ (k)) ,
P
(P
(P
(P
(P
#([)(
1
,
2
),#([)(
2
,
3
)
, #([)(
1
,
3
)).
Finally this can be rewritten
for each
1
,
2
,
3
N if #([)(
1
,
2
) and #([)(
2
,
3
) then #([)(
1
,
3
).
A dierent approach to logic 181
This is the meaning of our sentence Th
1
and that meaning is exactly as expected.
Our proof of statement Th
1
will begin by trying to exploit the denition of symbol [.
To this end we need to add another constant symbol in our language. This is the sym-
bol that stands for the product (or multiplication) operation in the domain N of natural
numbers. Therefore #() is a function dened on NN and for each , N #()(, )
is the product of and , in other words #()(, ) = . Given two expressions , in
our language if ()(, ) is also an expression in our language then it can be abbreviated
as () (as used in mathematics).
Lemma 6.2. Let m be a positive integer, x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j.
We have H[x
1
: N, . . . , x
m
: N] and we dene k = k[x
1
: N, . . . , x
m
: N].
Suppose i, j = 1 . . . m, i ,= j, suppose c 1 var(k). Then
[x
1
: N, . . . , x
m
: N, () (([)(x
i
, x
j
), () ( (c : N, (=)(x
j
, (x
i
c)))))] S();
#( [x
1
: N, . . . , x
m
: N, () (([)(x
i
, x
j
), () ( (c : N, (=)(x
j
, (x
i
c)))))]) is true.
Proof.
We have also H[x
1
: N, . . . , x
m
: N, c : N] and we can dene
k
= k[x
1
: N, . . . , x
m
: N, c : N].
By lemma 6.1 we obtain that x
i
, x
j
, c E(k
). Moreover E(k
) also holds.
For each
(k
) #(k
, ,
, x
i
,
)
and #(k
, c,
).
By lemma 3.12 we have (=)(x
j
, (x
i
c)) S(k
).
By lemma 3.1
(c : N, (=)(x
j
, (x
i
c))) E(k);
() ( (c : N, (=)(x
j
, (x
i
c)))) S(k);
for each (k)
#(k, () ( (c : N, (=)(x
j
, (x
i
c)))) , ) =
= P
(#(k
, (=)(x
j
, (x
i
c)),
)[
(k
), _
) .
Lemma 6.1 also tells us that x
i
, x
j
E(k) and for each E(k) #(k, x
i
, ) N,
#(k, x
j
, ) N. Moreover [ E(k) also holds.
For each (k) #(k, [, ) = #([) is a function with two arguments and
(#(k, x
i
, ), #(k, x
j
, )) is a member of its domain, therefore ([)(x
i
, x
j
) E(k).
Moreover for each (k)
#(k, ([)(x
i
, x
j
), ) = #([)(#(k, x
i
, ), #(k, x
j
, ))
= N : #(k, x
j
, ) = #(k, x
i
, ) ;
182 M. Avon
so #(k, ([)(x
i
, x
j
), ) is true or false and ([)(x
i
, x
j
) S(k).
From there follows that
() (([)(x
i
, x
j
), () ( (c : N, (=)(x
j
, (x
i
c))))) S(k) ;
[x
1
: N, . . . , x
m
: N, () (([)(x
i
, x
j
), () ( (c : N, (=)(x
j
, (x
i
c)))))] S() .
By theorem 3.6 we can rewrite
#( [x
1
: N, . . . , x
m
: N, () (([)(x
i
, x
j
), () ( (c : N, (=)(x
j
, (x
i
c)))))])
as follows
P
(#(k, () (([)(x
i
, x
j
), () ( (c : N, (=)(x
j
, (x
i
c))))) , )[ (k))
and this can be further rewritten
P
(P
(#(k, ([)(x
i
, x
j
), ), #(k, () ( (c : N, (=)(x
j
, (x
i
c)))) , ))[ (k))
P
(P
#([)(#(k, x
i
, ), #(k, x
j
, )),
#(k, () ( (c : N, (=)(x
j
, (x
i
c)))) , )
[ (k))
P
(P
#([)(#(k, x
i
, ), #(k, x
j
, )),
P
(#(k
, (=)(x
j
, (x
i
c)),
)[
(k
), _
[ (k))
P
(P
#([)(#(k, x
i
, ), #(k, x
j
, )),
P
(P
=
(#(k
, x
j
,
), #(k
, (x
i
c),
))[
(k
), _
[ (k))
P
({P
#(|)(#(k, x
i
, ), #(k, x
j
, )),
P
({P
=
(#(k
, x
j
,
), #(k
, x
i
,
) #(k
, c,
))|
(k
),
})
| (k)}) .
The nal statement can also be expressed (more textually) as follows:
for each (k)
#([)(#(k, x
i
, ), #(k, x
j
, )) if and only if
there exists
(k
) such that _
and #(k
, x
j
,
) = #(k
, x
i
,
) #(k
, c,
).
By denition we have
#([)(#(k, x
i
, ), #(k, x
j
, )) = N : (#(k, x
j
, ) = #(k, x
i
, ) ) .
Suppose #([)(#(k, x
i
, ), #(k, x
j
, )) holds.
There exists N such that #(k, x
j
, ) = #(k, x
i
, ) .
We dene
= + (c, ). We have
(k
) and _
.
Moreover since x
j
E(k), V
b
(x
j
) = and then c / V
b
(x
j
) we can apply lemma 4.12
and obtain that #(k
, x
j
,
) = #(k, x
j
, ). Similarly #(k
, x
i
,
) = #(k, x
i
, ). Using
lemma 4.13 we obtain #(k
, c,
) = . Therefore
#(k
, x
j
,
) = #(k, x
j
, ) = #(k, x
i
, ) = #(k
, x
i
,
) #(k
, c,
) .
Conversely suppose there exists
(k
) such that _
and
#(k
, x
j
,
) = #(k
, x
i
,
) #(k
, c,
).
A dierent approach to logic 183
There exists a positive integer n such that k K(n), N E
s
(n, k). Since
c 1 var(k) we have k
= k + (c, N) K(n)
+
, so there exist (k),
s #(k, N, ) = N such that + (c, s) =
.
Now we have to consider that _
, so =
/dom()
, and similarly _
, so
=
/dom()
. Moreover dom() = dom(k) = dom() so
=
/dom()
=
/dom()
= .
Since x
j
E(k), V
b
(x
j
) = and then c / V
b
(x
j
) we can apply lemma 4.12 and obtain
that #(k
, x
j
,
) = #(k, x
j
, ) = #(k, x
j
, ). Similarly #(k
, x
i
,
) = #(k, x
i
, ). So we
have
#(k, x
j
, ) = #(k, x
i
, ) #(k
, c,
).
By lemma 6.1 we have #(k
, c,
) N, so #([)(#(k, x
i
, ), #(k, x
j
, )) is proved.
This lemma allows us to create an axiom which is the set A
6.2
of all expressions
[x
1
: N, . . . , x
m
: N, () (([)(x
i
, x
j
), () ( (c : N, (=)(x
j
, (x
i
c)))))]
such that m is a positive integer, x
1
, . . . , x
m
1, with x
,= x
for ,= ,
i, j = 1 . . . m, i ,= j, c 1 var(k[x
1
: N, . . . , x
m
: N]).
Lemma 6.3. Let m be a positive integer, x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j.
We have H[x
1
: N, . . . , x
m
: N] and we dene k = k[x
1
: N, . . . , x
m
: N].
Suppose i
1
, i
2
, i
3
are distinct members of 1, . . . , m. Then
[x
1
: N, . . . , x
m
: N, (=) (() (()(x
i
1
, x
i
2
), x
i
3
) , () (x
i
1
, ()(x
i
2
, x
i
3
)))] S();
#( [x
1
: N, . . . , x
m
: N, (=) (() (()(x
i
1
, x
i
2
), x
i
3
) , () (x
i
1
, ()(x
i
2
, x
i
3
)))]) is true.
Proof.
By lemma 6.1 we obtain that
for each j = 1 . . . 3 x
i
j
E(k);
for each (k) #(k, x
i
j
, ) N.
Also belongs to E(k). For each (k) #(k, , ) = #() is a function with two ar-
guments and (#(k, x
i
1
, ), #(k, x
i
2
, )) is a member of its domain, so
()(x
i
1
, x
i
2
) E(k) and for each (k)
#(k, ()(x
i
1
, x
i
2
), ) = #()(#(k, x
i
1
, ), #(k, x
i
2
, )) = #(k, x
i
1
, ) #(k, x
i
2
, ) N.
Clearly we have also () (()(x
i
1
, x
i
2
), x
i
3
) E(k).
Similarly () (x
i
1
, ()(x
i
2
, x
i
3
)) E(k) so by lemma 3.12
(=) (() (()(x
i
1
, x
i
2
), x
i
3
) , () (x
i
1
, ()(x
i
2
, x
i
3
))) S(k) ,
and
[x
1
: N, . . . , x
m
: N, (=) (() (()(x
i
1
, x
i
2
), x
i
3
) , () (x
i
1
, ()(x
i
2
, x
i
3
)))] S().
184 M. Avon
By theorem 3.6 we can rewrite
#( [x
1
: N, . . . , x
m
: N, (=) (() (()(x
i
1
, x
i
2
), x
i
3
) , () (x
i
1
, ()(x
i
2
, x
i
3
)))])
as follows
P
(P
=
#(k, () (()(x
i
1
, x
i
2
), x
i
3
) , ),
#(k, () (x
i
1
, ()(x
i
2
, x
i
3
)) , )
[ (k))
P
(P
=
[ (k))
P
(P
=
[ (k))
P
(P
=
(#(k, x
i
1
, ) #(k, x
i
2
, )) #(k, x
i
3
, ),
#(k, x
i
1
, ) (#(k, x
i
2
, ) #(k, x
i
3
, ))
[ (k))
The last condition is clearly true, since for each (k)
(#(k, x
i
1
, ) #(k, x
i
2
, )) #(k, x
i
3
, ) and #(k, x
i
1
, ) (#(k, x
i
2
, ) #(k, x
i
3
, )) are
the same.
Lemma 6.3 allows us to create an axiom which is the set A
6.3
of all expressions
[x
1
: N, . . . , x
m
: N, (=) (() (()(x
i
1
, x
i
2
), x
i
3
) , () (x
i
1
, ()(x
i
2
, x
i
3
)))]
such that m is a positive integer, x
1
, . . . , x
m
1, with x
,= x
for ,= ,
i
1
, i
2
, i
3
are distinct members of 1, . . . , m.
6.1.1. The proof. We have already dened the sets ( and T in our language, as follows:
( = N, [, ;
T = , , , , , , , , =.
Moreover we dene 1 = x, y, z, c, d, e.
Our deductive system includes the axioms and rules weve listed in chapter 5 and in
this section 6.1.
The rst step in our proof of statement Th
1
uses axiom A
6.2
:
[x : N, y : N, z : N, () (([)(x, y), () ( (c : N, (=)(y, (xc)))))] . (6.1.1)
Then we can use R
5.3
to derive a new statement from 6.1.1:
[x : N, y : N, z : N, () (([)(x, y), () ( (c : N, (=)(y, (xc)))))] . (6.1.2)
In the next step we use axiom A
5.4
:
[x : N, y : N, z : N, () (()(([)(x, y), ([)(y, z)), ([)(x, y))]. (6.1.3)
At this point we can apply rule R
5.5
to 6.1.3 and 6.1.2 and obtain
[x : N, y : N, z : N, () (()(([)(x, y), ([)(y, z)), () ( (c : N, (=)(y, (xc)))))]. (6.1.4)
A dierent approach to logic 185
In much the same way we can obtain
[x : N, y : N, z : N, () (()(([)(x, y), ([)(y, z)), () ( (d : N, (=)(z, (yd)))))]. (6.1.5)
The next two statements are instances of axiom A
5.4
:
x : N, y : N, z : N, c : N, d : N, ()
()
(=)(y, (xc)),
(=)(z, (yd))
, (=)(y, (xc))
, (6.1.6)
x : N, y : N, z : N, c : N, d : N, ()
()
(=)(y, (xc)),
(=)(z, (yd))
, (=)(z, (yd))
. (6.1.7)
In fact if we dene h = k[x : N, y : N, z : N, c : N, d : N] then x, y, z, c, d E(h) and
for each (h) #(h, x, ), #(h, y, ), #(h, z, ), #(h, c, ), #(h, d, ) N.
Moreover E(h) and for each (h) #(h, , ) = #() is a function with two
arguments and (#(h, x, ), #(h, c, )) is a member of its domain. Therefore (xc) E(k)
and similarly (yd) E(h).
By lemma 3.12 we get (=)(y, (xc)) S(h) and (=)(z, (yd)) S(h) and the two
statements are instances of A
5.4
.
To proceed with our proof, our idea is to apply rule R
5.6
to 6.1.7 and 6.1.6.
We have x, y, z, c, d, e 1, N E, H[x : N, y : N, z : N, c : N, d : N, e : N].
We have already dened h = k[x : N, y : N, z : N, c : N, d : N] and we dene
k = k[x : N, y : N, z : N, c : N, d : N, e : N].
We want to apply rule R
5.6
with
() ((=)(y, (xc)), (=)(z, (yd))) in the role of ,
(=)(z, (ed)) in the role of ,
y in the role of t,
(xc) in the role of t
.
It has been shown above that () ((=)(y, (xc)), (=)(z, (yd))) S(h).
Its easy to see that (=)(z, (ed)) S(k). In fact e, d, E(k), for each (k)
#(k, , ) = #() is a function with two arguments, and #(k, d, ), #(k, e, ) N. This
implies that (ed) E(k). Since z E(k) we obtain (=)(z, (ed)) S(k).
Clearly y E(h) and for each (h) #(h, y, ) N = #(h, N, ).
Moreover x, c, E(h), for each E(h) #(h, , ) = #() is a function with two
arguments and #(h, x, ), #(h, c, ) N. This implies that (xc) E(h) and for each
(h)
#(h, (xc), ) = #()(#(h, x, ), #(h, c, )) = #(h, x, ) #(h, c, ) N = #(h, N, ).
We can use assumption 2.1.10 to evaluate V
b
(y) and V
b
((xc)). That assumption tells
us that V
b
(y) = and V
b
((xc)) = V
b
() V
b
(x) V
b
(c) = . Therefore, clearly,
V
b
(y) V
b
((=)(z, (ed))) = ; V
b
((xc)) V
b
((=)(z, (ed))) = .
186 M. Avon
In order to calculate (=)(z, (ed))
k
e/y and (=)(z, (ed))
k
e/(xc) we can exploit def-
inition 4.16. In one part of it we established that one of ve conditions holds true and a
consequent calculation of
k
x
i
/t.
By lemma 5.6 we know there exists a positive integer n
such that
/(n
, k) and we can
dene (=)(z, (ed))
k
e/y and (=)(z, (ed))
k
e/(xc).
There also exists a positive integer n
, k). If we set
n = maxn
, n
, n
; k; x : N, y : N, z : N, c : N, d : N, e : N)
holds, (ed) E(n
, k). If we set
n = maxn
, n
x : N, y : N, z : N, c : N, d : N, ()
()
(=)(y, (xc)),
(=)(z, (yd))
. (6.1.8)
The next statement is an instance of axiom A
6.3
:
[x : N, y : N, z : N, c : N, d : N, (=) (()((xc), d), ()(x, (cd)))] . (6.1.9)
Using rule R
5.7
we obtain
x : N, y : N, z : N, c : N, d : N, ()
()
(=)(y, (xc)),
(=)(z, (yd))
.
(6.1.10)
We can apply rule R
5.8
to 6.1.8 and 6.1.10 to obtain
x : N, y : N, z : N, c : N, d : N, ()
()
(=)(y, (xc)),
(=)(z, (yd))
. (6.1.11)
To proceed, our idea is now to apply rule R
5.9
to 6.1.11.
188 M. Avon
We have x, y, z, c, d, e 1, N E, H[x : N, y : N, z : N, c : N, d : N, e : N].
We want to apply our rule with the following assumptions:
k[x : N, y : N, z : N, c : N, d : N, e : N] takes the role of k;
h[x : N, y : N, z : N, c : N, d : N] takes the role of h;
() ((=)(y, (xc)), (=)(z, (yd))) takes the role of ;
(cd) takes the role of t;
(=)(z, ()(x, e)) takes the role of .
It has been shown above that () ((=)(y, (xc)), (=)(z, (yd))) S(h).
We observe c, d, E(h), for each E(h) #(h, , ) = #() is a function with
two arguments and #(h, c, ), #(h, d, ) N. This implies that (cd) E(h) and for each
(h)
#(h, (cd), ) = #()(#(h, c, ), #(h, d, )) = #(h, c, ) #(h, d, ) N = #(h, N, ).
We can add that V
b
((cd)) = V
b
() V
b
(c) V
b
(d) = .
Its also easy to see that (=)(z, (xe)) S(k). In fact x, e, E(k), for each (k)
#(k, , ) = #() is a function with two arguments, and #(k, x, ), #(k, e, ) N. This
implies that (xe) E(k). Since z E(k) we obtain (=)(z, (xe)) S(k).
In order to calculate (=)(z, (xe))
k
e/(cd) we can exploit denition 4.16. In one part
of it we established that one of ve conditions holds true and a consequent calculation of
k
x
i
/t.
By the proof of lemma 5.6 we know there exists a positive integer n
such that
/(n
, k) and we can
dene (=)(z, (xe))
k
e/(cd) at step n
, k). If we set
n = maxn
, n
, n
; k; x : N, y : N, z : N, c : N, d : N, e : N)
holds, (xe) E(n
.
There also exists a positive integer n
, k). If we set
n = maxn
, n
x : N, y : N, z : N, c : N, d : N, ()
()
(=)(y, (xc)),
(=)(z, (yd))
.
(6.1.12)
190 M. Avon
We can use the following instance of axiom A
6.2
:
[x : N, y : N, z : N, c : N, d : N, () (([)(x, z), () ( (e : N, (=)(z, (xe)))))] . (6.1.13)
Using rule R
5.3
we can derive
[x : N, y : N, z : N, c : N, d : N, () (() ( (e : N, (=)(z, (xe)))) , ([)(x, z))] . (6.1.14)
We can apply rule R
5.5
to 6.1.12 and 6.1.14 to obtain
x : N, y : N, z : N, c : N, d : N, ()
()
(=)(y, (xc)),
(=)(z, (yd))
, ([)(x, z)
. (6.1.15)
We can now apply rule R
5.10
. With the denition h = k[x : N, y : N, z : N, c : N, d : N]
we have ([)(x, z), (=)(y, (xc)), (=)(z, (yd)) S(h). So by R
5.10
we obtain
[x : N, y : N, z : N, c : N, d : N, () ((=)(y, (xc)), ()((=)(z, (yd)), ([)(x, z)))] .
(6.1.16)
By lemma 3.5 this can be rewritten
[x : N, y : N, z : N, c : N, [d : N, () ((=)(y, (xc)), ()((=)(z, (yd)), ([)(x, z)))]] .
(6.1.17)
We can apply rule R
5.11
using k = k[x : N, y : N, z : N, c : N, d : N],
h = k[x : N, y : N, z : N, c : N]. We consider that ()((=)(z, (yd)), ([)(x, z)) S(k),
(=)(y, (xc)) S(k).
Moreover, x, c, E(h) and for each (h) #(h, , ) = #() is a function
with two arguments and (#(h, x, ), #(h, c, )) is a member of its domain. Therefore
(xc) E(h). We also observe that y E(h) and therefore (=)(y, (xc)) S(h). Thus we
derive
[x : N, y : N, z : N, c : N, () ((=)(y, (xc)), [d : N, ()((=)(z, (yd)), ([)(x, z))])] .
(6.1.18)
This can be rewritten
[x : N, y : N, z : N, [c : N, () ((=)(y, (xc)), [d : N, ()((=)(z, (yd)), ([)(x, z))])]] .
(6.1.19)
We intend to apply rule R
5.12
using
k = k[x : N, y : N, z : N, c : N],
h = k[x : N, y : N, z : N],
= (=)(y, (xc)) S(k),
= [d : N, ()((=)(z, (yd)), ([)(x, z))] S(k).
To be able to apply that rule we need to show that
[d : N, ()((=)(z, (yd)), ([)(x, z))] S(h) .
Let = k[x : N, y : N, z : N, d : N]. By lemma 6.1 x, y, z, d E(), for each ()
#(, x, ), #(, y, ), #(, z, ), #(, d, ) N.
A dierent approach to logic 191
Therefore (yd) E(), (=)(z, (yd)) S(), ([)(x, z) S(),
()((=)(z, (yd)), ([)(x, z)) S() and [d : N, ()((=)(z, (yd)), ([)(x, z))] S(h).
We obtain
[x : N, y : N, z : N, () (() ({}(c : N, (=)(y, (xc)))) , [d : N, ()((=)(z, (yd)), (|)(x, z))])] .
(6.1.20)
We can apply rule R
5.5
to 6.1.4 and 6.1.20 and obtain
[x : N, y : N, z : N, () (()(([)(x, y), ([)(y, z)), [d : N, ()((=)(z, (yd)), ([)(x, z))])] .
(6.1.21)
At this point we need to apply rule R
5.13
using
k = k[x : N, y : N, z : N, d : N],
h = k[x : N, y : N, z : N],
= ()(([)(x, y), ([)(y, z)) S(h),
= (=)(z, (yd)) S(k),
= ([)(x, z) S(h) S(k).
We obtain
[x : N, y : N, z : N, () (()((|)(x, y), (|)(y, z)), () (() ({}(d : N, (=)(z, (yd)))) , (|)(x, z)))] .
(6.1.22)
The nal step in our proof consists in applying the modus ponens rule R
5.14
to 6.1.5
and 6.1.22. We get
[x : N, y : N, z : N, () (()(([)(x, y), ([)(y, z)), ([)(x, z))] . (6.1.23)
192 M. Avon
6.2. Second example. In this other example we want to prove a form of the Bocardo
syllogism. In Ferreiros referenced paper ([3]), on paragraph 3.1, the syllogism is expressed
as follows:
Some A are not B. All C are B. Therefore, some A are not C.
Suppose A, B and C represent sets, the statement we actually want to prove is the
following:
If ( (there exists x A such that x / B) and (for each y C y B) ) then
(there exists z A such that z / C).
In order to formalize this, our language must be as follows
( = A, B, C,
T = , , , , , , , , =,
1 = x, y, z,
where A, B, C are constants each representing a set.
At this point we suppose we can formalize the statement as
()
()
() ( (x : A, () (()(x, B)))) ,
() ( (y : C, ()(y, B)))
, () ( (z : A, () (()(z, C))))
.
(Th
2
)
Well soon see a proof of this statement and within the proof well also prove Th
2
is
a sentence in our language.
First of all we need the following lemma, that can be applied to the general language
of chapter 5.
Lemma 6.4. Let m be a positive integer, x
1
, . . . , x
m
1, with x
i
,= x
j
for i ,= j. Let
A
1
, . . . , A
m
( such that for each i = 1 . . . m #(A
i
) is a set. Let D ( such that #(D)
is a set. We have H[x
1
: A
1
, . . . , x
m
: A
m
]. If we dene k = k[x
1
: A
1
, . . . , x
m
: A
m
] then
for each i = 1 . . . m ()(x
i
, D) S(k).
Proof.
We rst consider that A
1
E() and #(A
1
) is a set, so A
1
E
s
() and H[x : A
1
].
Let k
1
= k[x : A
1
].
If m > 1 then for each i = 1 . . . m 1 we suppose H[x
1
: A
1
, . . . , x
i
: A
i
] holds and
we dene k
i
= k[x
1
: A
1
, . . . , x
i
: A
i
].
Clearly A
i+1
E(k
i
) and for each (k
i
) #(k
i
, A
i+1
, ) = #(A
i+1
) is a set.
So A
i+1
E
s
(k
i
), which implies H[x
1
: A
1
, . . . , x
i+1
: A
i+1
] (and we can dene
k
i+1
= k[x
1
: A
1
, . . . , x
i+1
: A
i+1
]).
A dierent approach to logic 193
This proves that H[x
1
: A
1
, . . . , x
m
: A
m
] holds.
Let i = 1 . . . m. We have A
i
E
s
(k
i1
) and k
i
= k
i1
+ (x
i
, A
i
). So we can apply
lemma 4.13 and obtain that x
i
E(k
i
). If i = m this implies x
i
E(k).
If i < m we consider that for each j = i + 1 . . . m x
j
/ V
b
(x
i
). So we can apply
lemma 3.14 and prove x
i
E(k).
Moreover D E(k) and for each (k) #(k, D, ) = #(D) is a set. By lemma 3.13
we have ()(x
i
, D) S(k).
6.2.1. The proof. To provide a proof of statement Th
2
well make use of a deductive
system which includes all the axioms and rules listed in chapter 5.
If we go back to the language we have introduced for our proof, using the former lemma
we can derive H[x : A] and we can dene h = k[x : A]. Moreover ()(x, B) S(h), so
()(()(x, B)) S(h).
We also have H[x : A, y : C] and we dene k
y
= k[x : A, y : C].
We have ()(y, B) S(k
y
) and by lemma 3.1 ()((y : C, ()(y, B))) S(h).
Thus () (()(()(x, B)), ()((y : C, ()(y, B)))) also belongs to S(h).
Moreover H[x : A, z : A] and we dene k
z
= k[x : A, z : A].
We have ()(z, C) S(k
z
) and by lemma 3.1 ()((z : A, ()(z, C))) S(h).
The rst sentence in our proof is an instance of axiom A
5.4
.
x : A, ()
()
()
()(()(x, B)),
()({}(y : C, ()(y, B)))
,
()({}(z : A, ()(z, C)))
, ()
()(()(x, B)),
()({}(y : C, ()(y, B)))
.
(6.2.1)
By A
5.4
we also obtain
x : A, ()
()
()(()(x, B)),
()((y : C, ()(y, B)))
, ()(()(x, B))
. (6.2.2)
By 6.2.1, 6.2.2 and rule R
5.5
x : A, ()
()
()
()(()(x, B)),
()((y : C, ()(y, B)))
,
()((z : A, ()(z, C)))
, ()(()(x, B))
.
(6.2.3)
Another instance of A
5.4
is the following
x : A, ()
()
()
()(()(x, B)),
()({}(y : C, ()(y, B)))
,
()({}(z : A, ()(z, C)))
.
(6.2.4)
By axiom A
5.16
we obtain
[x : A, ()(x, A)]. (6.2.5)
194 M. Avon
By 6.2.5 and rule R
5.7
we also get
x : A, ()
()
()
()(()(x, B)),
()((y : C, ()(y, B)))
,
()((z : A, ()(z, C)))
, ()(x, A)
. (6.2.6)
Since x E(h), C E
s
(h) etc. we can apply rule R
5.17
to 6.2.4 and 6.2.6 and obtain
x : A, ()
()
()
()(()(x, B)),
()((y : C, ()(y, B)))
,
()((z : A, ()(z, C)))
, ()(x, C)
. (6.2.7)
By axiom A
5.4
x : A, ()
()
()(()(x, B)),
()((y : C, ()(y, B)))
. (6.2.8)
By 6.2.1, 6.2.8 and rule R
5.5
x : A, ()
()
()
()(()(x, B)),
()({}(y : C, ()(y, B)))
,
()({}(z : A, ()(z, C)))
.
(6.2.9)
Since x E(h), B E
s
(h) etc. we can apply rule R
5.17
to 6.2.7 and 6.2.9 and obtain
x : A, ()
()
()
()(()(x, B)),
()((y : C, ()(y, B)))
,
()((z : A, ()(z, C)))
, ()(x, B)
. (6.2.10)
By 6.2.10, 6.2.3 and R
3.8
x : A, ()
()
()
()(()(x, B)),
()({}(y : C, ()(y, B)))
,
()({}(z : A, ()(z, C)))
, ()
()(x, B),
()(()(x, B))
.
(6.2.11)
By R
5.18
x : A, ()
()
()
()(()(x, B)),
()((y : C, ()(y, B)))
,
()((z : A, ()(z, C)))
. (6.2.12)
By R
5.19
x : A, ()
()
()(()(x, B)),
()((y : C, ()(y, B)))
.
(6.2.13)
By R
5.20
x : A, ()
()
()(()(x, B)),
()((y : C, ()(y, B)))
.
(6.2.14)
A dierent approach to logic 195
By R
5.10
x : A, ()
()(()(x, B)), ()
. (6.2.15)
Using lemma 6.4 we obtain that ()(y, B) S(k[y : C]) and ()(z, C) S(k[z : A]).
By lemma 3.1 we obtain that ()((y : C, ()(y, B))) S() and similarly
()((z : A, ()(()(z, C)))) S().
We can apply rule R
5.21
to 6.2.15 and obtain
()
(6.2.16)
Finally, by R
5.22
, we obtain
()
()
(6.2.17)
6.2.2. Additional notes. We have proved statement Th
2
, this also means that Th
2
is
a sentence in our language. It seems quite obvious that the statements meaning is as
expected, anyway to complete the argument we also want to prove this.
We need the following lemma, that can be applied to the general language of chapter 5.
Lemma 6.5. Let u 1, D ( such that #(D) is a set. We have H[u : D] and we can
dene h = k[u : D]. Then, for each (h) #(h, u, ) #(D). Moreover, for each
#(D), if we dene = + (u, ) then (h) and #(h, u, ) = .
Proof.
We have D E(1, ) and #(, D, ) is a set, so D E
s
(1, ), h = + (u, D) K(1)
+
and
(h) = + (u, s)[ s #(, D, ) = + (u, s)[ s #(D) .
Therefore, for each (h) there exists s #(D) such that = + (u, s).
It follows by lemma 4.13 that #(h, u, ) = s #(D).
Now let #(D) and = +(u, ), clearly (h) and by 4.13 #(h, u, ) = .
We now examine the meaning of () ((x : A, ()(()(x, B)))).
We can rewrite #(() ((x : A, ()(()(x, B))))) as
P
(P
(P
(P
(P
(P
(
x
, #(B))),
there exists
x
#(A) such that
x
doesnt belong to #(B).
Similarly we can rewrite #(()((y : C, ()(y, B)))) as
P
(P
(P
(
y
, #(B)),
for each
y
#(C)
y
belongs to #(B).
Similarly we can also rewrite #(() ((z : A, ()(()(z, C))))) as
P
(P
(P
(P
(P
(P
(
z
, #(C))),
there exists
z
#(A) such that
z
doesnt belong to #(C).
At this point we can rewrite
#
()
()
as
P
()
and then
P