Abstract.For an integer n , let G(n) denote the smallest x such that the primes < x generate the multiplicative group modulo n . We offer heuristic arguments and numerical data supporting the idea that G(n) < (log2)_1 lognloglog« asymptotically. We believe that the coefficient 1 / log 2 is optimal. Finally, we show the average value of G(n) for n < N is at least(1 + o( 1 )) log log N log log log N, and give a heuristic argument that this is also an upper bound. This work gives additional evidence, independent of the ERH, that primality testing can be done in deterministic polynomial time; if our bound on G(n) is correct, there is a deterministic primality test using O(logn)2 multiplications modulo n.