In this paper it will be shown that any conditional moment test of functional form of nonlinear regression models can be converted into a chi-square test that is consistent against all deviations from the null hypothesis that the model represents the conditional expectation of the dependent variable relative to the vector of regressors.