The present paper describes a method whereby the field of an electromagnet can be kept constant automatically to at least one part in fifty thousand ( i. e. , to 0⋅2 gauss in 10,000 gauss) for long periods. The method has the great advantage of not merely ensuring that the current through the magnet (or, say, the voltage across the magnet windings) is kept constant, but that the field itself , as measured by a fluxmeter, is kept within one-fifth of a gauss (or less) of some desired value. Thus, it is not necessary to know whether a fluctuation of the field is caused by unsteadiness of the exciting battery voltage, variations of the magnet coil resistance with temperature, etc. ; in all cases, the stabilizer automatically readjusts the exciting current until the field is brought back to its original value. Further, as explained at the end of the paper, the stabilizer can be made to bring about an increase or decrease in the field of a definite number of gauss, and to continue to maintain the field at this new value. Also with very slight modifications, which are so obvious as to require no further explanation, the stabilizer could be made to maintain a current, or a potential difference, constant just as easily as a magnetic field. The method has been applied with success to the annular magnetic field used in the Cavendish Laboratory. The magnet is used for the analysis of α-particle groups. Under the influence of suitable fields of the order of 10,000 gauss, α-particles of various velocities are made to travel in a semicircle of 40 cm. radius, and are focussed on to the slit of a counting chamber. Groups of various velocities are focussed in turn by varying the field, their velocities being determined from the values of the fields. It is extremely important, therefore, to know the field strength with accuracy and to maintain it constant, at each of these values, during the experiment. In some experiments already described this was done manually. An observer watched the spot of light reflected from the mirror of a fluxmeter, the coil of which was connected to a stationary search coil situated in the magnetic field. A change of one-fifth of a gauss in the magnetic field produced a deflection of the fluxmeter spot of about 1 mm., and by suitable adjustment of the field current by means of a mercury rheostat it could be arranged that the spot was kept within half a millimetre of its zero portion. This, however, required one observer to spend his whole time in stabilizing the field during an experiment which might last several hours.