The measure of non-normal sets

Summary It is well-known that almost every number in 0, 1] is normal in base 2, in the sense of Lebesgue measure. Kahane and Salem asked whether the same is true with respect to any Borel measure whose Fourier-Stieltjes coefficients vanish at infinity — in other words, whether the set of non-normal numbers is a set of uniqueness in the wide sense. We show that this is not the case. In fact, we give best-possible conditions on the rate of decay of in order that -almost every number be normal. The techniques include, on the one hand, probability measures with respect to which the binary digits in 0, 1] are independent only by blocks, rather than individually, and on the other hand, the strong law of large numbers for weakly correlated random variables.This work was partially supported by an NSF Graduate Fellowship, NSF Grant MCS-82-01602, and an AMS Research Fellowship.