Bayesian estimation for the two parameters of the Gumbel distribution are obtained based on record values. Prediction, either point or interval, for future lower record values are also presented from a Bayesian view point. Some of the results of Ahsanullah (1991) can be obtained as limiting cases from our results. Based on a recurrence relation of conditional moments of nonadjacent record values, a characterization for the Gumbel distribution is also given. Numerical computations are given to illustrate these procedures.