This paper is about a nonparametric structural model that is not additive in disturbances. We use independence and monotonicity conditions for identification. The average structural function is obtained from a nonparametaric regression on the endogenous variables and a control function which is the conditional distribution function for the regressors given the instruments. A two step series estimator is used, the first step estimating the conditional distribution and the second using this estimator as a regressor. Convergence rates are given.