“…The exact sequence (3) is sequentially pure if, for every increasing sequence of ordinals and ~'s _v=(~r,),~N, the following sequence is exact: (6) is exact for every v such that a,,<), for every nEN. The converse is not generally true, as the theory of Warfield groups and balanced projective groups shows ( [3], [4], [9]); but according to the fact that torsion Warfield groups are totally projective, the converse holds if all groups in (3) ; it follows that the Ulm indicator of b" is -~(a0, co, ~, ...)=>v. Suppose now that the claim is true for n-1 (n>l); then, if b+A has exponent n, pb+A has exponent n-1 and pb+AEC (al, a2, 9 .... a ..... ); we can suppose pbEp*lB with bEp~ for the second equality in (5) and for al-_>o'0+1; by induction hypothesis, there exists xEB(al, a2, ..., a~, .,.) ).-pure), then it is also balanced (resp.…”