ences (cf. second and third tables given below). Successive readings are taken till the graduated circle is rotated as near as possible through 360°.
As has been said before, there are two semi-cylinders P and Q. In the first set of experiments P is turned towards the radiator, Q acting as a focussing lens. The circle at each successive operation moves in the right-handed direction.
In the second set of experiments Q is turned towards the radiator, P acting as the converging lens. Successive readings are taken as before, the circle now rotating in the left-handed direction.
It will be observed that the final results obtained from the two sets are freed from many of the outstanding errors.
I give below the results of two sets of experiments each extending through eleven observations. The receiver was in an unusually good condition for nearly an hour, and during that time I took six observations with P to the front and six more with Q in the same position. As the receiver continued to remain in a fairly responsive condition I took five more for each set. As already said, greater weight should be given to the first two sets of six readings, as being taken under the most favourable conditions.
Successive differences of the mean readings, taken with the cylinders P and Q are given below.
| P | .... | 71 | 70 | 70·5 | 70·5 | 70·5 | 70 | 69·5 | 69·5 | 69·5 | 70 | = | 701 |
| Q | .... | 70 | 70 | 71 | 71 | 70 | 71 | 71 | 70 | 70 | 69·5 | = | 703·5 |
| Mean | .... | 702·25 |
.
