Instrument inertia can produce spurious effects in controlled‐stress rheometry whenever the imposed torque changes with time, as in step‐jump or ramp‐function programming of the torque. These errors are especially serious with samples of low to moderate viscosity, where spurious time‐dependent viscosities are found even with Newtonian samples. The usual way of dealing with this problem, that of slowing the speed change program sufficiently to eliminate errors caused by inertia, both lengthens the time of measurement and obscures any rapid thixotropic transformations. This paper analyzes errors due to instrument inertia for both step‐jump and ramp‐function programming in terms of an inertial response time β=KI/η, whereKis the usual instrument constant for the particular rotational geometry,Ithe moment of inertia of the rotating part of the instrument, and η the viscosity of the sample. It then shows that ramp times of the order of 400β are needed to reduce inertial errors to acceptable levels. Finally, the paper shows that these errors can be eliminated by subtracting the inertial torque from the imposed torque, thereby eliminating spurious effects even for ramp times less than 40β. The proposed method of correction is applicable to non‐Newtonian and viscoelastic samples in all rotational geometries, and to any programming of the imposed torque.