18 ERRETT A. BISHOP because M and No, for instance, are fixed positive integers, defined explic- itly by the proof of the inequality a b. Of course we could decide to substi- tute other values for the original values of M and N if we desired, so 0 ' some choice is possible should we wish to exercise it. If we do not explicitly state what choice we wish to make, it will be assumed that the values of M and No given by the proof of a b are chosen. The number H, which is constructively a well-defined real number, is classically rational, because if the Goldbach conjecture is true then H = 0, and if the conjecture is false then H = 2- n t l , where n is the first even integer for which it fails. We are not entitled to assert constructively that H is rational: if it is rational, then either H = 0 or H # 0, meaning that either Goldbachls conjecture is true or else it is false and we are not entitled to assert this constructively, until we have a method for deciding which. We are not entitled to assert H is irrational either, because if H is irrational, then H f 0, therefore Goldbachls conjecture is false, therefore H is the rational -nt1 number 2 , a contradiction! Thus H cannot be asserted to be rational, although its irrationality is contradictory. (I am indebted to Halsey Royden for this amusing observation. ) It is easy to prove the existence of many irrational numbers, by proving the uncountability of the real numbers, as a corollary of the Baire category theorem. For the present, let us merely remark that h/Z is irrational. Of course, fi can be defined by constructing successive decimal approxima- tions. It is therefore constructively well-defined. The classical proof of the 2 2 irrationality of fi shows that if p/q is any rational number then p /q f 2. Since both p2/q2 and 2 can be written with denominator q2 , it follows that Since clearly p/q f fi in case p/q 0 or p/q 7 2 , to show that p/q # fi we may assume 0 s p/q 5 2 . Then Therefore, h/ ?T # p/q. Thus is (constructively) irrational. The failure of the usual form of trichotomy means that we must be care- ful in defining absolute values and maxima and minima of real numbers. For

Purchased from American Mathematical Society for the exclusive use of nofirst nolast (email unknown) Copyright 1985 American Mathematical Society. Duplication prohibited. Please report unauthorized use to cust-serv@ams.org. Thank You! Your purchase supports the AMS' mission, programs, and services for the mathematical community.