The effects of random errors in array weight and sensor positions on the performance of a Linearly constrained linear sensor array is analyzed in a weight vector space. It is observed that a nonorthogonality exists between an optimum weight vector and the steering vector of an interference direction du e to random errors. A novel approach to improving the nulling performance by compensating for the nonorthogonality is proposed. Computer simulation results are presented.