The calculation of the torque T on a permanent ellipsoidal magnet immersed in a magnetic liquid is more delicate than one might be inclined to expect. This problem, for instance, has a bearing on fundamental discussions carried out in the past about the choice between different formulations of electromagnetic theory. The present paper analyses the calculation of T anew, emphasizing the relation to, and the explanation of, the important experiment carried out by Whitworth and Stopes-Roe in 1971. The outcome of the experiment is explained in terms of standard magnetostatics, based upon use of the scalar potential phi . The relation to the so-called Kennelly-Sommerfeld controversy is commented upon. The physical interpretation of the situation is elucidated by carrying out in detail an examination of the total energy balance in the special case of spherical magnet geometry.