This paper studies testing for the presence of smooth transition nonlinearity in adjustment parameters of the vector error correction model. We specify the generalized model with multiple cointegrating vectors and di§erent transition functions across equations. Given that the nonlinear model is highly complex, this paper proposes an optimal LM test based only on estimation of the linear model. The null asymptotic distribution is derived using empirical process theory and since the transition parameters of the model cannot be identiÖed under the null hypothesis bootstrap procedures are used to approximate the limit. Monte Carlo simulations indicate a good performance of the test.