We briefly discuss the so called pseudo-GLS estimator in a standard linear regression model with nonsperical disturbances, and conclude that the potentiality for applications is higher than originally assumed by Fiebig Bartels and Kramer (1996).