644 PROCEEDINGS: philosophical society 



steel balance wheels and the watches were constructed from new de- 

 signs in only a few months. Those submitted on the other order (for 

 250 watches) had brass-invar balance wheels and had been carefully 

 developed during about five years or more. The mean of the middle 

 temperature errors (difference of daily rate at 20 °C. from the mean 

 of the daily rates at 35 °C. and 5°C.) for the 200 accepted watches 

 with brass-steel balance wheels was 2.25 seconds, while for the 250 

 accepted watches with brass-invar balance wheels the mean was 0.24 

 second, thereby showing that invar will permit much better tem- 

 perature compensation than will steel. 



Following the end of the acceptance tests of these watches they were 

 checked up each week until they were sent to the ships ; the mean weekly 

 temperatures were also observed. The rates of 30 watches with brass- 

 invar balance wheels and 32 watches with brass-steel balance wheels 

 were carefull}^ corrected for the observed changes of temperature. 

 Based upon observations for 18 weeks the mean of the probable errors 

 of forecasting the mean daily rate for the ensuing week was 0.277 

 second for the brass-invar watches and 0.260 second for the brass- 

 steel. 



It was noticed that occasionally the watches with brass-invar balance 

 wheels showed sudden permanent changes of rate corresponding to 

 permanent changes in the inertia of the balance wheels. These 

 occasionally amounted to 10 seconds in the mean daily rate. Follow- 

 ing a jump of this character, the rate of a watch usually would remain 

 practically constant at its new value. Most of the watches did not 

 show any pronounced changes. In computing the above probable 

 errors, there was not used any watch which showed an important 

 change of rate. These erratic performances were not observed for the 

 watches with brass-steel balance wheels. 



The conclusion is that invar, when substituted for steel in balance 

 wheels, will permit practically perfect temperature compensation, 

 and will show good stability of rate; occasionally, however, sudden 

 permanent changes of rate throw a shadow of doubt over the ad- 

 visability of this substitution. This conclusion is in agreement with 

 statements made by employees of certain watch manufacturers who 

 have experimented on this subject, and with the reports on the stability 

 of invar for other purposes. 



The paper was illustrated by lantern slides. It was discussed by 

 Messrs. R. Y. Ferner, C. A. Briggs, and W. P. White. 



Adjournment took place at 10.10 and was followed by a social hour. 



S. J. Mauchly, Recording Secretary. 



