MATHEMATICAL LOGIC

A PHILOSOPHICALLY PLAUSIBLE FORMAL INTERPRETATION

OF INTUITIONISTIC LOGIC
BY

ANDRZEJ GRZEGORCZYK

(Communicated by Prof. A. HEYTING at the meeting of September 26, 1964}

The notion of forcing introduced by P. J. CoHEN [2] in order to prove

the independence of the axiom of choice, may be considered as a modi-
fication of the notion of fulfilment of a formula by a tree introduced
by E. W. BETH [l] p. 447. But the word "forcing" used by Cohen
suggests an interesting philosophical interpretation of intuitionistic logic.
According to this interpretation, which shall be described in detail, the
intuitionistic logic can be understood as the logic of scientific research
(rather positivistically conceived) while on the other hand the classical
logic is the logic of ontological thought.

l. The formal description of scientific research

Scientific research (e.g. an experimental investigation) consists of the
successive enrichment of the set of data by new established facts obtained
by means of our method of inquiry. When making inquiries we question
Nature and offer her a set of possible answers. Nature chooses one of
them. Hence scientific research may be conceived as a triple
R=(J, o, P).
The elements of R will be called: J - the information set, i.e. the set
of all possible experimental data, o - the initial information, P - the
function of possible prolongations of the informations.
The set J may be finite or infinite. The elements of J are finite ordered
collections of atomic sentences: Pt1 (aJ), Pt 2 (aJ> ak), ... where Ptn are
n-place predicates and a1, ak, ... are object names. (The compound
sentences are not a product of experiment, they arise from reasoning.
This concerns also negations: we see that the lemon is yellow, we do
not see that it is not blue.)
The element o may be often considered as the empty collection.
The function P is defined over the set J and assumes as values non-
empty subsets of J. For IX E J, P(1X) C J. Moreover if IX= (A1, ... ,An)
where At are atomic sentences, then either P(1X) ={IX} or for every f3 E P(1X)
there exist atomic sentences An+b ... , An+k+l (k;;;.O) such that
{3 =(AI. ... , An, An+b ... , An+k+l)·
 597

That is, the next state of scientific information contains the given one
and a finite sequence of new atomic sentences.
Hence research may be described as a tree: (Fig. 1}.

Fig. 1.

In performing an experimental trial we proceed from the point o along

a branch gaining new data. If a branch is finite, the last information
cannot be enlarged: P(y) = {y }.

2. Inducing to assertion
Now we define the relation: {3 >-R IX ({3 is an extension of IX in the researchR).
0 n+l n
{3 >-IX= {J=IX;
R .
{3 >-IX-
R
vy (y >-IX A
R
{3 E P(y))

>-R = V {3 >-R
1'1

{3 IX IX.
n

The fundamental notion: IX [> ([J (the information IX ~n the research R

R
forces (or induces) us to assert the formula ([J) is defined inductively. We
shall consider two definitions; the simple one is adequate and exact
for the sentential intuitionistic logic, the other is exact also for the
quantification logic.
Definition 1:
1. If ([J is an atomic formula without variables then
IX [> ([J = ([J E IX.

2. For compound formulas: 1)

IX [> r ([J V lJ'l = (IX [> ([J V IX [> lJ')
IX [> r ([J A lJ'l - (IX [> ([J A IX [> lJ')

1) The index R by the symbols >- and [> are omitted if there is no danger of
confusion.
 598
>- =* ,...._, (f3 [> (/)))
1\ (f3 1X
fJ R

~ 1\ (f3 >- ex=* ((3 [> W =? f3 [> P))

fJ R

ex [> 1 V Xt W(xt) l ~ there exists a constant object name at such

x; that ex [> W(xt(at)
ex [> 1 1\ x; W(xt) l ~ for every object name at : ex [> W(xi(at).
x;

One can easily recognize here the essential points of Beth's definition.
Cohen's definition differs from the above one in the case of general
quantifier.) The positivistic methodological attitude is expressed for
example in the case of negation: we are forced by our research to assert
a negation --, (/)if this research never will force us to assert the statement (/).

3. The interpretation of intuitionistic sentential logic

We shall say that a triple R=(Jn, on, Pn) is a research if it satisfies
the formal conditions of section l and all information is an extensions
of the initial information o:

One can prove easily the lemma:

Lemma 1. If R=(Jn,on,Pn) ts a research, cxEJ, f3>-cx and
ex [> W, then f3 [> W.

Proof. By induction with respect to the length of the formula W.

The proof of the following is sketched thus:
Theorem l. A formula (/) (without quantifiers) is logically true in
the formal intuitionistic logic if and only if each information ex of every
research R forces us to assert the formula W.
Proof. Using Lemma l we verify Heyting's axioms. For the converse
inclusion let us recall Jaskowski's matrices for intuitionistic propositional
calculus 2). They can be presented as the algebras of open sets on trees.
Let T=(ST, oT, PT) be a tree: ST is the set of all ramification-points,
oT is the top-point and PT is a function mapping ST into the class of
subsets of ST. PT(x) CST for x EST. By means of PT we can define the
relation >- as in section 2.
T
Then for subsets XC ST we define the interior function:
(l) y E Int X Df 1\ x E ST(x-< y =* x EX).
T

Hence with every tree T we let AT be the corresponding algebra of

open sets on T. If T contains only one point: S={o}, then AT is the

2) See also G. F. RosE [5], and D. ScoTT [6J.

 599
two element Boolean algebra. If T1 and T 2 are two disjoint trees, then
T 1 +T 2 will be the tree obtained from T1 and T2 by adding one new
common top-point above T1 and T2. Hence every open set X of T1 +T2,
except the whole space BT1 +T2 , is the union of two disjoint parts
X= (X n BT1 ) u (X n BT2 ) and may be considered as a pair: (X n BT 1 ,
X n BT2 ). Thus the algebra AT1 +T2 may be formally obtained from
AT1 and AT2 by taking first the direct cartesian product of AT1 and AT2
and then by adding one new open set containing all others. These two
operations are according to JAsKOWSKI [4] sufficient to generate a
characteristic set of matrices for the HEYTING's [3] formal intuitionistic
calculus of propositions.
Now we prove that for every finite set of atomic sentences A1, ... ,AN
and for every finite tree T and valuation g of A1, ... , AN (by open sets
in the algebra AT), there exists a research R isomorphic with T and such,
that for every formula f/J (without quantifiers) built from A1, ... ,AN,
we have the equality:
(2) valg f/J = {x E BT: rp(x) 1> f/J}
R
where rp is the isomorphism of T onto R.
The set BT being finite may be considered as ordered in such a way
that if y E P(x) andy =1= x then x precedes y. Then we define the function
rp by induction on the tree T:

rp(oT) Df (Ai0 , ••• , Aik) where Ai0 , ••• , Aik are all atomic sentences
At ( 1 < i < N) such that oT E g(At), if they exist;

rp(oT) Df (AN+l) if there is no At (l <;i<;N) such that oT E g(At).

Suppose that rp(x)=(Bb ... , Bs) and y E P(x), y =1= x, we put:

rp(y) Df (B1, ... , Bs, A10 , ••• , AJn) where AJ0 , ••• , A1n are all atomic
sentences A 1 (l <i <;N) such that x ¢ g(A 1) and y E g(AJ), if
they exist; if not we put:
rp(x) Df (B1, ... , Bs, Au), where Au is the first atomic sentence
different from A 1, ... , AN and such that for every z E BT if
z precedes y Au ¢ rp(z).
The construction of the research R = ( J R, oR, P R) is the following:
IX E J R Df v X E BT IX=tp(X),
Df
OR= tp(oT),
IX E PR(fJ) nr V x, y E BT (x E PT(Y) A IX= rp(x) A {J = rp(y)).
It is easy to see that R is a research, rp is an isomorphism, and for
l <i <N the definitions imply the equality:
g(At) = {x : "rp(x) 1> At},
R
 600
which involves (2) for atomic sentences accordingly to the definition of
valuation: valg(Ad=g(At), for At atomic. Suppose (2) for f!J and 'P we
shall prove it for f!J ~ 'P. According to Tarski's interpretation method
for intuitionistic connectives, and by (1) and (2) we have:
valg(f!J ~ 'P) =Int((ST-valgf!J) u valg'P) = {x : (\ y(y >-- x =*
T
y E ((ST-valgf!J) u valg'P))}= {x : 1\ y(y >-- x =*
T '
y E ((ST- {z : cp(z) [> f!J}) U {z : cp(z) [> 'P}))}= {x : A y(y >-- x =*
· R R T
(cp(y)·[> f!J=* cp(y) [> 'P))}={x: 1\ {J({J >-- cp(x)'*
R R R
({J [> @'*_ f3 [> 'P))}={x: cp(x) [> '@ ~ 'Pl}.
R R R

In the last two steps we profit by the fact that cp is an isomorphism

and we use Definition l.
For the other connectives the proof is similar.
From (2) we obtain that:
(3) valgf!J =ST = 1\ eX E J ReX [> f!J.
R

Hence if f!J is forced by· every information-state of every research R,

then f!J is topologically true by every valuation in every finite tree. (The
first inclusion may be also proved in this way instead of verifying the
axioms.)
This interpretation is not exact for the intuitionistic predicate calculus,
e.g. the following rule is forced by every information state of every
research: 1\ x(f!J v 'P(x)) ~ (f!J v 1\ x 'P(x)). (By Cohen's original definition
even the rule 1\ xf!J(x) ~ f!J(a) is not satisfied.)

4. The interpretation of the intp,itionistic predicate calculus

In order to obtain an exact interpretation for the whole intuitionistic
logic we must modify the definition of forcing in the following way:
Definition 2 (informal version):
'l. For atomic formulas: c:x [> At Df, irrespectively of how we continue
our research R from the state c:x, we obtain an information containing the
statement At 3).
2. For compound formulas we modify Definition 1 only for disjunction
and existential quantifier:
c:x [> 'f!J v 'Pl Dr irrespectively of how we continue our research from
the state c:x, we obtain an information f3 such that fJ [> f!J or
fJ [> 'P;

3) In the case of Definition 2 it may be better to say that "o.: [> <P" means: the
R
information state o.: of the research R potentially forces us to assert the statement <P.
 601
IX [> rv Xi IP(xt) 1 Dr irrespective of how we continue our research from
the state IX, we obtain an informaLion fJ such that for some
a : fJ [> IP(a).
To formalize Definition 2 we need the notion of branch:
X is a branch of R=<J, o, P) Df XC J A o EX A 1\ IX (IX EX=:-
there exists one and only one fJ such that fJ E P(1X) and fJ EX).
Then the phrase "irrespectively of how we continue our research R from
the state IX we obtain an information fJ such that ... " may be formalized as
"1\ X((X is a branch of R A ex EX)=:- V {J({J EX A ••• ))". And it is easy
to see that, in the case of trees with P(1X) finite for every IX E J, this
definition coincides with Beth's.
Theorem 2 . A formula IP is provable in the formal intuitionistic
logic of quantifiers if and only if IP is forced by each information of every
research.
Proof. First we prove by induction the lemma 1 and the following one:
IX[> IP = 1\ X((X is a branch of R A IX EX)=:- V {J({J EX A fJ [> IP)).

Using the lemmas we verify that all rules of the formal intuitionistic
logic are forced by every information of every research. The converse
inclusion follows from Beth's theorem.
Instytut Matematyczne
Sniadeckich 8
Warszawa 10, Polen

REFE~ENCES

1. BETH, E. W., The foundations of Mathematics. Studies in Logic. Amsterdam,

1959.
2. CoHEN, P. J., The independence of the axiom of choice (mimeogr.) Stanford
1963.
3. HEYTING, Die formalen Regeln der intuitionistischen Logik, Sitzungsberichte
der Preussischen Akad. der Wiss. 1930, pp. 42-56.
4. JASKOWSKI, S., Recherches sur le systeme de la logique intuitioniste. Actes du
Congres Inter. de Philosophic Sci. Paris 1936, part VI, 58-61.
5. RosE, G. F., Propositional calculus and realizability. Trans. Amer. Math. Soc.
75, 1-18 (1953).
6. ScoTT, D., Completeness Proof for the Intuitionistic Sentential Calculus.
Summer Institute of Symbolic Logic in 1957 at Cornell University
(mimeogr.), 231-241.
7. TARSKI, A., Der Aussagenkalkiil und die Topologie. Fund. Math. 31, 103-134
(1938).

•lltJOTHEIK MATH!MATISCH C&NTIUJK

AMSTeRDAM