In their article, Liu and Breen propose a new estimator of the large-sample asymptotic covariance matrix for the limited information maximum likelihood estimator in simultaneous equations, and question the interpretation of a statistic used in the past to test overidentifying restrictions. The present paper comments on this matter.