On the Restricted Ordinal Theorem

Author(s): R. L. Goodstein
Source: The Journal of Symbolic Logic, Vol. 9, No. 2 (Jun., 1944), pp. 33-41
Published by: Association for Symbolic Logic
Stable URL: http://www.jstor.org/stable/2268019 .
Accessed: 16/06/2014 13:49

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

Association for Symbolic Logic is collaborating with JSTOR to digitize, preserve and extend access to The
Journal of Symbolic Logic.

http://www.jstor.org

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
 Tim, JOURNAL OF STMBOLIC Loaic
Volume 9, Number 2, June 1944

ON THE RESTRICTED ORDINAL THEOREM

R. L. GOODSTEIN

The proposition that a decreasing sequence of ordinals necessarily terminates

has been given a new, and perhaps unexpected, importance by the r6le which it
plays in Gentzen's proof' of the freedom from contradiction of the "reine Zahlen-
theorie." Godel's construction2 of non-demonstrable propositions and the
establishment of the impossibility of a proof of freedom from contradiction,
within the framework of a certain type of formal system, showed that a proof
of freedom from contradiction could be found only by transcending the axioms
andlproof processes of that formal system. Gentzen's proof succeeds by utilizing
transfinite induction to prove that certain sequences of reduction processes,
enumerated by ordinals less than e (the first ordinal to satisfy e =-') are finite.
Were it possible to-prove the restrictedordinal theorem,that a descending sequence
of ordinals, less than e, is finite, in Gentzen's "reine Zahlentheorie," then it would
be possible to determine a contradiction in that number system. In his paper,
Gentzen proves the theorem of transfinite induction, which he requires, by an
intuitive argument. There is also a method of reducing transfinite induction,
for ordinals less than e, to a number-theoretic principle given by Hilbert and
Bernays,3 and a similar method by Ackermann.4 None of these proofs of trans-
finite induction is finitist.
As the restricted ordinal theorem is a suggested minimum deviation from the
previously accepted field of finitist processes, it becomes highly important to
examine to what extent this theorem fulfils general finitist requirements. For
this purpose it is necessary to give an account of the ordinal signs which does not
presuppose any part of the Cantor theory of infinite classes, and in fact such an
account is given in Gentzen's paper, but it is more convenient for our purpose
to present the construction of ordinal signs differently from Gentzen.
By means of additions, multiplications, and exponentiations we can express
any numeral n uniquely in the form
cksak + csk-1 + ... + C2sa2 + CiSal + co

wheres > 2,0 co < s,0c < ,c,c3,C3, <

,ck < s,0 a, < a2< a3 < .. < ak,
and each a1 is itself of this form. We shall call this the representation of n with
digits 0, 1, 2, ***, s - 1 and scale symbol s. If +8(X) is an abbreviationfor the
representation of nt with scale symbol s, then this expression may be defined
recursively as being the same as CS(a) + 0,(n - cs), where a is the exponent

Received November 26, 1943.

X G. Gentzen, Die Widerspruchefreiheit der reinen Zahlentheorie, Mathematische Anna-
len, vol. 112 (1936), pp. 493-565.
2 K. Godel, Uber formal unentscheidbare Sdtze der Principia Mathematica und verwandter

Systeme I, Monatshefte fur Mathematik und Physik, vol. 38 (1931), pp. 173-198.
3D. Hilbert and P. Bernays, Grundlagen der Mathematik, vol. II, Berlin 1939, pi). 360-
372.
4 W. Ackermann, Zur Widerspruchsfreiheit der Zahie-ntheorie, Mathematische Annalen,
vol. 117 (1940), pp. 162-194.
33

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
34 R. L. GOODSTEIN

of the greatest power of s which does not exceed n, and CS'is the greatest multiple
of s not exceeding n.
Denoting by Sbva the expression obtained by replacing x, at each point of
its occurrencein an expression a, by y, we define T.(n) = Sb7Am(n), the operator
Sb' applying to the expression for which 0.(n) stands, not just to the sign 0b(n)
itself'(so that Sb74)m(n) is not +$(n)), and define the ordinals (less than e) to be
the expressions T(n) for any m and n, m > 2. Thus for instance TV(106) is
the ordinal wc' + 2 2 + 2w + 1 (express 106 in the scale of 3 with digits 0, 1, 2
and then replace "3" by "w"). Every ordinal a, less than e, in the Cantor theory,
is expressible in the form T. (n), m being any natural number greater than each
of the natural numbers which occur as coefficients or exponents in the expression
of a by powers of w and sums of such powers with numerical coefficients, and n
being uniquely determined by a and m.
We shall also use the sign S'(n), with natural numbers x, y, n, where y _ x > 1,
to denote the number obtained by substituting "y" for "x" in the expression
representing n in the scale of x; i.e., S:(n) is the number which is represented by
Sbo (n) in the scale of y. For example S'(34) = 265, since 34 = 38'+ 2 . 3 + 1
and 4' + 2.4 + 1 = 265; and S2(16) = 4256, since 16 = 22' and 444 .= 4256.
The formulae T7(n) are not all distinct, for we can show that corresponding
to any m' > m we can find n' such that T' (n') and T' (n) are the same formula;
in fact if n' = S.,(n) then T' (n') = T7(n), for by definition Sb ,+(n) =
0.,(SM,(n)) and therefore T. (S',(n)) = Sb''"4.,(SM,(n)) = Sbm'Sb,0Mf(n) =
Sb":'kn(n) = T1(n).
For any ni, n2, ml _ m2 > 1 we say that T71(nn)is greater than, equal to, or
less than T"2(ng) according as ni is greater than, equal to, or less than S'2(n2);
this definition is in accordance with the usual definition of inequalities between
ordinals. A decreasing sequence of ordinals takes the form
T'T.(ni), Tm'(Q2, T',7(713)} X r(r

where, for each value of r, mr+ _? mr and nr+l < S;r+(n7). For every con-
structively given sequence of ordinals the sequence mr is general recursive
though not perhaps primitive recursive in every case. For a given function mr
we obtain the 'longest' sequence by taking nrl = S;r+1(nr) 1, for T'(n) = 0
if and only if n = 0, and if n < nr then S;r+1(n) < Slr+1(n,)..
The restrictedordinal theoremmay now be expressed by saying that tor any
non-decreasing function Pr, pO > 2. and for nr defined by the recursive equation
nr+1= SP;+r(nr) * 1, we can find a value of r for which TW(nr) = ;0.
We observe first that the restricted ordinal theorem is equivalent to the
following number-theoreticproposition:
Given any non-decreasing function Pr, Po. 2, a number no, and the function
_

nr defined by the recursive equation nrI, = S'rJ(nr) * 1, then there is a value

of r for which it, = 0.
We shall call this proposition.P*. It makes no essential difference in forming
the sequence nr+,'n, n2, whetherin formingflrfJ from nr we first reduce n,
...

by unity and then change the scale, or as we have done above, first change the
scale in the representation of n,. and then reduce the resulting number by unity.

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
 ON THE RESTRICTED ORDINAL THEOREM 35

In fact if we form a sequence m, by the recursive equation m,+1 = SP"+,(m, A 1),

then the proposition P* above is proved if we can prove the proposition P that
there is a value of r for whichm, = 0. For if m, > 0, for r s, and m.+1 = 0,
then taking no = mo 1, from nk = mk - 1 we deriveni+i = s;(l(nk) 1-a =
SPt+,(mk * 1) I = k -;- 1, and therefore n, = m, -a- 1, for all r, whence
-

ne+ = 0.
We shall give a completely finitist proof of the proposition P (constructing an
explicit formula determining a value of r for which m, = 0) for values of mo
not greater than poPoP'. This is equivalent to proving the restricted ordinal
theorem for ordinals not greater than c,,(.
It will make the demonstration easier to follow if we consider first the case
mO ? Pol.
Let u(n) be a non-decreasing sequence, a(O) x 2, and let a sequence
7y(x, n, p, r) be defined by the equations:
'Y.(x,p, n, 0) = x Ia (n)
,,(x, p, n, r + 1) -S(f+,.)
= 17,(x,p, nr 11.
Define the function f,(x, p, n) by the equations:
*6,(a,b, c, n) = c + f,,(a, b, n + c),

f,(1, 0, n) = 1, (i)
f,,(x + 2, p, n) = ,,(x + 1, p, f,(1, p, n), n), (ii)
f(l, p + 1, n) = 4(o(n) * 1, p, f6(1, p, n), n), (iii)
wherex 0 > 0,n k 0. Thenforallx + 1,p < a(n), k 2 f,(x + l,p,n),
O,p
'y,(x+1,p,n,k) = 0.
For x, p, n ? 0 let Po(x + 1, p, n) denote the proposition, "If
k = f,,(x + 1, p, n) and x + 1, p < a(n) then -y.(x + 1, p, n, k) = 0.," Equation
(i) proves P,(1, 0, n). And equation (ii) proves P,(1, p, n) & P,(x + 1, p, n +
f,(1, p, n)) -4 P,(x + 2, p, n); for starting from (x + 2) fl(n))I, with x + 2 <
a(n), p < a (n), we reach in turn (x + 1){o(n + 1)) + Sl (:) )[|r(n) I 11,
and (x + 1){a(n + 2)1j + Sf (+:)[S (,%l)[{cr(n))P 1] -* 11, and so on up
to (x + 1) Ia(. + g~,p, n)) 1 in f,(1, p, n) steps, and (x + 1) ICf(n+ f(,p, n))J
is reduced to zero in a further f.(x + 1, p, n + f,(1, p, n)) steps. Furthermore,
starting from {G(n) '+', where p + 1 < a(n), the next term is fo(n)
11 a(n + 1)IP + So(n+)[TIa(n)I * 11, and so on, so that equation (iii) proves
P,(1, p, n) & PO(a(n) -A-1, p,n +f,(1, p, n)) P,(1,p + 1, n).
From the proved propositions,
P,(1, 0, n), (a)
Po(1,p, n) & P,(x + 1, p, n + f,,(1, p, n)) -- P,,(x + 2, p, n), (b)
P.(1, p, n) & P.(a(n) -*- 1, p, n + f.(1, p, n)) - P,,(I, p + 1, n), (c)
we can derive P,(x + 1, p, n) by an application of the generalised schema of
induction II described in Th. Skolern's paper Fine Bemerikungiiber die Induk-

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
36 R. L. GOODSTEIN

tionsschematain der rekursivenZahlentheorie,5which, as Skolem shows, if we take

into account the observation which Miss R. P6ter makes in her review of Skolem's
paper,6 can be reduced to an ordinary induction. For by generalized induction
the formula P,(x + 1, p, n), with variables x, n and some definite numeral p,
is derived from Po(1, p, n), with the same numeral p, by means of formula (b)
above; in particular Pq(a(n) . 1, p, n + fo(1, p, n)) is derivable from Po(1,p, n)
and hence by (c) we derive Po(1, p + 1, n) from Po(1, p, n). From tWis, in
conjunction with (a), we then derive P,(1, p, n) by induction over p, from which
we conclude that P,(x + 1, p, n) holds for arbitrary values of x, p, n _ 0.
This is a finite constructiveproof of the restricted ordinal theorem for ordinals
less than cw@.
Next we observe that, writing R for a(n) 1, we have {a(n) V(n" =
RtIa(n) r + Ia(n) and therefore the sequence S.(n, r), with S.(n, 0) = to(n) yr(n)
and Oq(n,r + 1) = S'(t'+)l) I6S(n, r) * 1), reaches zero in
fr(1l R. n) + fa(R, R. n + fq(1, R. n)) = fq(1, a(n), n)
steps, for fo(1, R, n) steps take us from RIa(n)IR + Ia(n)IR to Rta(n +
fr(1, R. n)) R, and therefore a further f,(R, R, n + f?(1, R, n)) steps are needed
to reach zero. Thus the restricted ordinal theorem is proved for ordinals less
than or equal to co@. (Notice that the formula fo(1, a(n), n) for the number of
terms in a sequence commencing with 0a(n)v(n) is the same as the formula for
a sequence commencing with 0a(n) ', p < n, with p replaced by a(n); this is
to be expected since the relation of t0(n)J'( ) to Ja(n) R is the same as the
relation of ta(n)}R to a(n) IR 1.)
Consider next the sequence eo(x, Yo,yl, ... y, n, r) with
Y, *,
ec(x, yoyo, y. , n, 0) = (x + 1){If(n)IyO+yIo(n)+Y2[o(n)12+.-+yif[(nRJi

and
(X,yo y1* X* y,* nr+ 1) = S'('+'l) yOI(X **o yl
.X , ,
n.r) } *
1) .

The function fo,j(x, yo, y, * . y j, n) is defined by the equations (recursive for a

definite value of j):
faj(O, 0, *.., 0, n) = 1 (iv)

fo, (x + 1, yo, y , *... y1, n)

= 4o,,(x, Yo,y,* , yj fr,j(Oyo,yi, *...*, yj, n), n), (v)
fa,,i(0 0, . * 0, Yr + 1, Yr+li Yr+2, . * yi, n)
= j(R 1,1xR R,. , R, Yr, Yr+l, ... Yjy

fcrj(O R, R, * I R, Yr, YJr+li *. yi, n), n), (vi)

where jx O
0, < r _ j, R = a(n) 1, and qj(x, yo.Y1, **.,yCl)=
c + fq,j(x,Yo,yi, *-** yi, n + c).
5 Monatshefte fUr Mathematik und Physik, vol. 48 (1939), pp. 268-276.
6 In this JOURNAL, vol. 5 (1940), pp. 34-3.5.

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
 ON THE RESTRICTED ORDINAL THEOREM 37

Let Pe, j(x, yo) Y', *.* , y, n) say, "If k = f1,,j(x,yo,Yi, *.. , y,, n) and x, io, YI,
*.. yj < a(n) then e.(x, yo, yi, *, yj, n, k) = O."
Equation (iv) proves P.,,(O, 0, * 0, n). Since (x + 2) t (n) 1' =
(x + 1)ta(n)1' + {a(n)}', equation (v) proves
tPr,i(O, Yo,yi, *... yy, n) & Poj(x, yo, yi, * , Yj,
n + fir,j(O,yo, yJ, *,, yi, n)) -+ P,,,(x + 1, yo, yi, *,, yi, n). (g)
And since
{a(n) I(Yr+1) (n)JI r+Yr+1[k(n)'rl+.- +vsqk(n)Ii

= R Ia(n) IR+Ru(n) +* +R[o(n)I r

I+yr lo(n)lr+- *+i[o(n).

+ I (n) IR+Rw(n)+...+R [(n)IJ-l+Vr (n)Ir+- +vi [(n)JI

therefore equation (vi) proves

Rx Rx Xv R, Yr, Yr+li *s,
(Pr~jJO, Yj, n)
& Pcr,j(R * 1, R, R. , R, Yr, Yr+i, , Yii
* **
j(, R. Rx * * Rx
n + fao, a YraYr+bx * y , n))}
Pa,(O, 0, X 0, Yr + 1, Yr+, Yr+2, * .
*, yj, n). (h)
If for given values of x, yo, y', *., yj we can derive, for any assigned
m, Paj(X, Yo, Y', *, yj, m) from P.,j(0, yo, y', *., yj, n) utilising only the
elementary propositional calculus and the operations of substituting for variables
and replacing computable functional expressions by their values, then in par-
ticular we can derive P., j( yo, Y * - y j, m) and P j(xYo,Yo , ,
yj, m + f?,j(O, Yo,y', ,yj,
*-- m)) by these means, and hence by (g) we derive
Po, j(x + 1, yo, y', ***, y ji m), whence it follows, by induction over x, that we can
derive P,,j(x, yo, y', yIj, m) from P.,j(0, yo, y', * ., yj, n), for any x. Fur-
***,

thermore, if for given values of x, r, Yr, Yr+1, X'', yj we can derive P.,j(x, yo,
Y*, y* , nm) for any assigned m, from P.,j(0, 0, * 0, Yr, Yr+b, *.*, yi, n), then
we can derive both PO,j(0, R, R, ***, R, Yr, Yr+l, * , yj, m) and P., j(R 1, R, R,
* , R, Yr, Yr+l, *'*, y,j m + fr,j(0, R, R, * * *, R, Yr, Yr+l, **', yj, m)) and hence
by (h), P,j(O, 0, , 0, Yr + 1, Yr+l, Yr+2, yj, m). By induction over Yr
0*.,

it follows that from PO,j(O, 0, ***, 0, 0, Yr+i, Yr+2, *.., yj, m) we can derive
Po, j(0, 0, , 0,
*X r, Yr+i, * * *, y , m) for any assigned m. Accordingly if we can
derive P., j(x, yo, y', * * , yj, m) from P, j(0, 0, ***, 0, Yr, Yr+i, ** X yi, n) then that
formula can also be derived from P,,j(0, 0, * 0, 0, *Y
1+1 yi, n). But we
have seen that Poj(x, yo, y', **., yj, m) can be derived from P.,j(0, Yo,Yi, ,
yj, n), and therefore Poj(x, yo, Yi, .., y,, m) can be derived from the proved
proposition P, ,j(O,0, *. , 0, n). This derivation is completely finitist, and in
.

fact it can readily be seen that, starting with the proved proposition P.,,(O, 0,
*.I, 0, n) and substituting repeatedly in this and in the formulae (g) and (h)
definite numerals for the variables x, yo, yi, *., yi, n, we derive the formula
P., (a, j0o, A3, **, IP, JA), for assigned numerals j, a, io, pi, *., p,
j, and an
assigned a(n), after exactly NO,i(a, Po, PI, * * , pi, 1A)applications of the formulae

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
38 R. L. GOODSTEIN

0, **.. 0, n), (g), and (h), where N, j(x,yo, yl, **.. yj, n) is defined by the
PU,,(O,
recursive equations:

Not,(O,0 * ,0, n) = 1,
N.,j(x + 1,yo, y/', .. yj, n)
= Nq,j(x, ye, yii ** yj, n + fr. (O Yo,y', *.*., yj, n)) + 1,
Ne, Ao; ; O, Yr+ 1 I.. yj, n) 1, R, R* RY r+i

=No j.(R -- 1; R; R; .. ; Ry yr; yr+,; -; yi,

n + fo,,j(0,. ;R) .. * R; Yr,
yryr+, * ., yj, n)) + 1.

Thus the restricted ordinal theorem is proved for ordinals less than w'*.
Since
I
a0 (w(n) = R Io(n) R+Ro(n) +R o( n)]2+- *+R.r(n))
R

+ Ia(n) jR+Ro,(n) +R[o(n)12+- .*+R[,(n)J R

where R = a(n) * 1, therefore the sequence r,(n, r), with

(n,) =

r,(n, r + 1) = S.(++i)Il[ , r) *1]

reaches zero in

R. Ry.. R*,R n) + fost(R -* 1, R. R.y*

fO,rt,RO R.
n + f(,RO; R) R, R, n)) = f,(O, 0, ,0 R + 1, n)
steps, which completes the proof for ordinals less than or equal to d"'.
We shall now show that the formula for the number of terms in the sequence
commencing with (a(n)}I?(')' ( ) is the same as the formula fo j(0, 0, *.*. , 0 1, n)
for the sequence commencing with (a(n) 1 I" j < 0(n), with j replaced by
0(n); i.e., that

fr,R(0, 0, *- , 0, R + 1, n) = futo(n)(0O0 ***, 0, 1, n).

First we prove the identity:

fw.,stx;yol yi; - , yj, n) =-- fo -j+1$ Yo, Yb ..

,j O n).
0,
Let E,(x, yo, yI/, ., yj, n) assert this identity. By equation (iv), E,(O, 0, *.*.
0, n) holds, and by equation (v) and induction we derive E,(x, yo, y,, ** *, y;, n)
from E,(O, yo, yi, *.., Furthermore by equation (vi) we derive E,(O, 0,
yj, n).
yan s 1t Yr+1o
, n)), + oEyj,n) from EI/O, R; R;.).fl R; Yr Yr+alas.the proon) and
-*
Eot(R 1, R; R; *-.- R; Yr; Yr+l; * - yj, n + fwst?, R; R; *-.*, R; yr; Yr+l; * *;
*
yj, n)), and so the proof of Eotx, yo, yi, - -*, y*, n) follows exactly as the proof of

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
 ON THE RESTRICTED ORDINAL THEOREM 39

P, ,(z, /Yo,I/, , Y , n) above. Hence

f,-i+(0 .0 *.*, O. 1, n) =
Of f.,+i(0, R. RI * , R. 0, n)
+ f.,,+1(R 1, R, R, - - R, 0, n + fui+i(O, R, R. **, R. 0, n))

=frj(O,R. R. -, R. n) + frj(R -z 1, R. R., RI

n + frj(O, R. R. , n))
=fvj(O, 0, **,O,, R + 1,'n),
and from this the required result follows by taking j = R.
The method of proof readily extends to ordinals beyond w', but to reach

by these means seems hardly to be worth the labour involved. On the other
hand it seems likely that a more subtle approach would enable the theorems to
be proved, by finitist methods, for ordinals up to any assigned v., where vo =
Vn+1 = w"'. The important point revealed by the foregoing proofs is that if a
function g(k, n) specifies the number of terms in a decreasing sequence com-
mencing with some F(k, u(n)), k < a(n), then F(a(n), u(n)) is followed by a
decreasing sequence of at most g(o(n), n) terms, so that from a proof of the'
restricted ordinal theorem for otdinals less than or equal to 11(k) we derive a
proof of the theorem for ordinals less than or equal to f(w). The position ap-
pears to be, therefore, that if P(n) expresses the restricted ordinal theorem for
ordinals less than or equal to v., then P(n) is capable of a finite constructive
proof for any assigned n, but (n)P(n) is not so provable-which of course in-
volves that in the "reine Zahlentheorie," there can be no general formula G(k, n)
with a free variable k, specifying-the number of terms in a decreasing sequence
commencing with the ordinal Vk, but only specific formulae for particular values
of k.
The formula P6(x + 1, p, n) above can be derived from the formulae (a),
(b), (c) in a purely formal manner by means of recursive number theory. The
following derivation was communicated to me by Professor Bernays.
Let f(x, p, n), g(p, n), h(n) be recursive functions, and let P(x, p, n) be an
abbreviation for the equation f(x, p, n) = 0. Then P(x, p, n) will be derived
from the formulae:
P(O, 0, n), (a*)
P(0, p, n) & P(x, p, g(p, n)) -_ P(x + 1, p, n) (b*)
P(0, p, n) & P(h(n), p, g(p, n)) -+ P(0, p + 1, n). (c*)

Define + (x, p, n) by the equations,

f(0,p, n) = n,
#(x + 1, p, n) = 4(x, p, g(p, n)),

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
40 R. L. GOODSTEIN

and let Q(x, p, n) be an abbreviation for

E f(O1 p, I(z, p, n)) = 0.

From the demonstrable formula,

f(O, p, #(z, p, n)) = f(O, p, n) + E f(o, p, 4P(z+ 1, p, n)),

and the definitions of O(x, p, n) and Q(x, p, n) we derive

Q(x + l, p, n) -P(O, p, n) & Q(x, p, g(p, n)), (d*)
which together with (b*) gives

IQ(x, p, g(p, n)) -+P(x, p, g(p, n)) I -- IQ(x + l,p, n) -*P(x + l,p, n) }
and this formula, in conjunction with the demonstrable formula Q(O, p, n)
P(O, p, n), gives Q(x, p, n) -- P(x, p, n), by means of that schema of generalised
induction referred to above. (The application of Skolem's schema II is not quite
immediate, since the two parameters p, n must first be reduced to a single param-
eter by the method explained by Hilbert and Bernays for the case of primitive
recursion.7) From Q(x, p, n) -- P(x, p, n) and the formulae (c*) and (d*) we
derive:
Q(h(n) + 1, p, n) -) P(O, p + 1, n). (e*)

From (e*) and (a*) we derive P(O, p, n) by the schema of generalised induction
briefly discussed, at the end of the paper of Skolem's to which we have already
referred,8 as being reducible to schema II and ordinary induction. To carry
out this reduction we make the following definitions. K(p, n) is an abbreviation
for
a
uWs, ssh(u)+l
Fa f(0,pd'(zpu))=0,
and
js(p, n) = Max Max ^,6(z,p, u).
uWi s<h(u)+l
From the definitions of K(p, n) and #(x, p, n) there follows
K(p, n) -+ P(O, p, n), (ko)
and using (a*) and the definition of Q(x, p, n) we obtain
K(0, n). (ki)

Furthermore the definitions of K(p, n), Q(x, p, n) and formula (e*) yield
K(p, k)-+ (E f(0, p + 1, x) = 0),
xz5k

7D. Hilbert and P. Bernays, Grundlagen der Mathematik, vol. I, Berlin 1934, p. 322.
8 In footnote 5. Vide pp. 276-276.

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions
 ON THE RESTRICTED ORDINAL THEOREM 41

whence

K(p, ;(p + ln)) f(O, p+ C,#(zp + 1,u)) =O,

WS% s~h(u)+l

i.e.,
K(p, is(p + 1, n)) -+ K(p + 1, n). (k2)
Formulae (k1) and (k2) give K(p, n), by the generalised induction schema II,
whence by (ko) we derive P(O, p, n), i.e., f(O, p, n) 0, whence Q(x, p, n).
Finally P(x, p, n) is derived from the proved formulae Q(x, p, n), Q(x, p, n)
P(x, p, n). Taking
(ao(n) (x + 1)) *(a(n) .y(x + 1, p, n, f.(x + 1, p, n))
*- p))

for f(x, p, n), n + f,(1, p, n) for g(p, n), and a(n) - 1 for h(n), it follows that
(a(n) *(x 1)).(a(n) * p)).y,(x + 1, p, n,f.(x + 1, p, n)) = 0
is proved, and this equation is a formal transform of the formula P,(x + 1, p, n).

The author acknowledges with most grateful thanks his deep indebtedness to
Professor P. Bernays for much valuable advice and most generous assistance.

THE UNIVERSITY, READING, ENGLAND

This content downloaded from 185.44.78.129 on Mon, 16 Jun 2014 13:49:02 PM

All use subject to JSTOR Terms and Conditions