550' Mr. P. E. Shaw on 



tract, according as the short arm of L 3 is moved towards or 

 away from the observer. Suppose the rings are made to 

 contract by the working of the pulleys, and that while the 

 observer at M sees one ring absorbed, the observer at T sees 

 the graduated disk move by n divs. Then the value of the 

 disk units in terms of the wave-length of sodium light can 

 be found. 



The glass surfaces (a, b) must not touch during these 

 observations. If the surfaces be moved uniformly together 

 the ring system expands at a uniform rate until they touch, 

 when the system begins to expand at a very much slower 

 rate. Thus we shall know the position of contact. 



The burner was surrounded by an Argand chimney, and 

 glass plates (not shown in the figure) were placed between it 

 and the rest of the apparatus to cut off heat. It was found 

 well to leave the apparatus for 15 minutes after lighting up, 

 till temperature equilibrium was established ; then the rings 

 were fairly steady, and observations could be obtained. 



At first the plan was to have one observer at the te^scope 

 and another at the microscope, as above described. But this 

 was unsatisfactory, there being a want of cooperation between 

 hand and eye ; so a second pulley string (s 2 s 2 ) was fitted to 

 pulley Px and passed round pulley P 2 so that the microscope 

 observer could move the ring system himself, and when the 

 observer at the telescope called out a reading to him he put 

 it dowm. This method of working was retained as satisfactory. 



Table IV. shows the results ultimately obtained. The rings 

 moved to some extent under the temperature changes (though 

 this movement was small when precautions were taken). But 

 we know about the time taken to pass from ring to ring, and 

 by measuring the time taken for the creep of one ring, we 

 can make a correction. This is the use of column 1. 



It will be seen that the four means agree very well, but 

 the individual values are not nearly so close ; surprise may 

 be felt at the considerable discrepancies, for if the apparatus in 

 its general use were not more reliable than in this calibration 

 it would not be very efficient. But the conditions in the two 

 cases are very different, for in calibration (1) the levers are 

 uncovered ; (2) the burner being near causes considerable 

 disturbances ; (3) to avoid a large creep during an observa- 

 tion the levers are moved much faster than usual, which may 

 cause shaking. 



The value obtained from calibration agreed closely with 

 the number previously estimated by measurement of the lever- 

 arms. 



The band system was disturbed by the passing of a cart 



