In general this problem would be insoluble, since there is no difference in the nature of the acceleration of gravity and that due to centrifugal force. A way out is offered, however, by the fact that true gravity acts continuously and at a small angle to the axis of the gyro, while the components which cause the pseudo-gravity are of short duration, liable to rapid changes of direction, and, on a turn, act at a large angle. What we require, therefore, is an erector which will respond slowly but surely to the _average_ acceleration, which is downward, but too sluggishly to be affected by the shorter period accelerations due to turns or rolls. Slowness of response is a matter of the erecting forces being small and of the mass and angular velocity of the gyro disc being large. The success of the compromise called for depends on the relative times taken for the gyroscope to tilt seriously from the true vertical, due to the causes above mentioned, and for the average turn or roll. Fortunately the former is a matter of minutes, the latter of seconds or at the worst of fractions of a minute. More than this, since the roll or turn is apt to be of much greater angle than any normal deviation of the gyroscopic axis from the vertical in the same time, we are offered the possibility of some device for filtering out the deviations which alone are to effect the erector. For instance, by shunting the restoring force whenever it is called upon to act through more than a predetermined small angle. As to the method of erecting the gyroscope, its characteristic property must be kept in mind. This is that the axis does not tilt under an applied force in the direction it would if the gyro were not rotating, but around an axis at right angles to that of the applied couple. Thus in Fig. 79, if a weight is attached as shown, the disc does not incline downward toward the weight, around the axis _Y_, _Y´_, but _precesses_ about the vertical axis Z, Z´. Some means is therefore needed to translate the pull which any gravitational control, such as a freely swinging pendulum, would give, into a pull with _at least a component_ at a finite angle to this.