REMARKS ON PREDICATE LOGIC WITH
INFINITELY LONG EXPRESSIONS
Colloquium Mathematicun, vol. 6 (1958), pp. 171-176.REMARKS ON PREDICATE LOGIC
WITH INFINITELY LONG EXPRESSIONS
By
A. PARSKI (BERKELEY)
As extensions of ordinary first orde1 predicate logic P, various systems
of predicate logie with infinitely long expressions ean be considered(*),
To fix the ideas we restrict ourselves to the discussion of predicate
logic P, with denumerably long expressions. Atomic formulas in P, are
expressions like
(Po +++ Pa)
consisting of a predicate g and a finite sequence of variables (0), ..., t1)+
‘The set of all predicates is assumed to be at most denumerable (though
this restriction is not essential) and to contain the binary identity predi-
eate =; instead of =: (v%, 0%) we write (r = v,). Compound well-formed
formulas are obtained from simpler ones by means of the following ope-
rations: (i) the formation of the negation ~F of a formula F; (ii) the
formation of the implication [F,—> F,] of two formulas F, and Fy; (iii)
the formation of the disjunetion
VEFy ee Pred
and the conjunetion
ALPy ee Pyoved
of a finite or denumerable sequence of formulas (Py)... Py +.d3 (iv)
the universal quantification
(We -
(}) Tais note contains the text of the remarks made by tho author at the
Summer Lustituto of Symbolic Logic in 1957 at Cornoll University; it first appoarod
(undor tho saine title, though in a more eoneiso form) in Summaries of talks presented
at the Summer Institut: of Symbolio Logic in 1957 at Cornell University, vol. 1
P. 160-163 (mimoographrd). Tae rusulls of this note ware obtained and th» noto was
prpared for publication while the author was working ou a resoaroh project in the
foundations of mathomatics sponsored by tho Na‘ional Seionce Foundation and carried
‘through in the Univursity of California, Berkeley.
inand the existential quantification
(Fe. F
of a formula F over a finite or denumerable sequence of variables
0qy's++4 Mp) ++.» The notion of @ free (or bound) occurrence of a variable
ina formula is defined in the usual way. A formula without free occurrences
of variables is called a sentence. A universal sentence is a sentence of the
form
(Weg oo. Me dP
where F is a formula without quantifiers.
In this note we shall not attempt to define for P, such fundamental
syntactical notions as provability or derivability; we shall be concerned
exclusively with some semantical and, specifically, model-theoretical
problems,
There is no difficulty in extending basic semantical notions to the
logic P,; in particular, it is elear under what conditions a relational system
WA, Roy very Beye
is regarded as a model of a sentence & in P, or of a set 2 of such sentences.
Bach of the relational systems 2% involved here is formed by a non-empty
set A and by @ finite or denumerable sequence of finitary relations
Rg; +++) Ryy «..) among elements of this set. A class K of relational systems
is called an arithmetical P-class or, simply, a P-class it it eoincides with
the class of all models of some set of sentences in the logic P; it is called
8 universal P-class if it coincides with the class of all models of some set
of universal sentences in P, For P we can take here Po, P,, or any other
logical system which may be mentioned in our further discussion (*),
Various notions applying to arbitrary relational systems (such as
similar systems, at most denumerable system, isomorphic image, sub:
system, and extension of a system, union of a class of systems) are assu-
med to be known. A non-empty class L of relational systems is ealled
directed (denumerably directed) if, for every finite (at most denumerable)
subclass M of L, all systems in Mhave a common extension which belongs
to L, See here [6], part I, p. 573 ff. (A bibliography is given at the end
of this note),
Several known results and observations in the theory of models can
be extended in an appropriate form from the logic P, to the logic P,.
As an example we state the following
(@) The notion of a Py-class (or « universal P,-elase) can also bs dofined in purely
mathomatical ternis, without involving tho logic P, itsolf or any other logical forma
ism. Compare {0] for an analogous dofinition of Py classes (arithmetical classes)TrmoREM 1. For a class K of (similar) relational systems to be a uni-
versal Py-class it is necessary and sufficient that I satisfy the following three
conditions:
(i) sf @ system belongs to KK, then all its isomorphic images belong to K;
(ii) if @ aystem delongs to Ky, then all its subsystems belong to Ki;
(iii) if @ denumeradly directed clase of systems is included in K, then
the wnion of this class belongs to K.
Condition (iii) can be equivalenty replaced by:
(il) af every at most denumeradle subsystem of a sysiem belongs to K,
then the system itself belongs to K.
The proof of this theorem follows the lines of the proof of an analo-
gous theorem for the logic P,, in fact, of Theorem 1,2 in [6]. The argument
can be conveniently based upon the following simple
Leama. For every at most denumeradle relational system & a univer-
sal sentence 8 in P, can be constructed such that a system B is a model of 8
if and only if no subsystem of B is an isomorphic image of U.
In connection with this lemma compare Theorem 1.1 in [6] and its
proof.
For illustration consider the class W of all well-ordered systems
formed by the set « of all natural
numbers and the ordinary relation <). It has been shown that W is not
a Py-class; ef, [7], part IT, p. 301, or else (8], p. 382. On the other hand
it is well known that the class K = W satisfies conditions (i) and (ii)
of Theorem 1, and it is easily seen that it also satisfies condition (iii) (*),
Consequently, by Theorem 1, W must be a universal P,-class; and, in
fact, W proves to coincide with the class of models of the following three
universal sentences (in which the binary predicates » and = occur as the
only non-logical constants):
mw (Vo90s)[p (0904) > [p (01%) > (v, = »4)]]5
(2) (Ve,010.)[P(t%) > [9(002) > ploovs)]},
(3) (Wy «++ Bae) V [P(t 0) «+ (On Mag)»
Here