The empirical likelihood approach was proposed by Owen to construct nonparametric confidence regions for finite dimensional parameters under a fixed number of known constrains. In this paper, an Wilks type theorem is derived where the number of constraints is allowed to grow at the rate o(n^(2/5)).