BRIDGMAN. — MEASUREMENT OF HYDROSTATIC PRESSURES, 333 



other before making readings. The following set of readings, picked 

 at random ft-om a great number, shows how nearly it was possible to 

 avoid hysteresis by the method. The difference of readings is seldom 

 more than the possible error of reading. The small extra load applied 

 or removed before making these readings was constant over the entire 

 range, being sufficient to produce a deflection 

 of about 2.0 divisions. It is evidently not 

 quite enough at the higher pressures and a 

 .little too much at the lower pressures. The 

 mean of the two readings nowhere differs from 

 either reading by more than the errors of 

 observation, however, and this simpler pro- 

 cedure was therefore adopted. 



The most inconvenient fact disclosed by the 

 calibration was that the constant of the springs 

 varies slowly from time to time. Over two or 

 three days no change whatever is to be noticed, 

 but in a week or a month there are likely to 

 be changes beyond the limits of error. The 

 change is irregular and has no apparent con- 

 nection with temperature changes. No tem- 

 perature effect within the limits of working 

 room temperatures was ever found. The 

 change is doubtless due to some slow process 

 of molecular accommodation going on within 

 the metal itself. At one time the springs 

 were permanently deformed by a violent explo- 

 sion, so that the deflection under the same 

 load was increased in the ratio 16:14. After 

 this deformation for a month or more the 

 change with time was more rapid than usual 

 magnitude of the variation with time. The first set of three, Jan- 

 uary-April, 1910, was made at intervals during constant use of the 

 gauge. During this time the gauge constant decreased and then in- 

 creased again. The explosion referred to above took place on De- 

 cember 24, 1909. The last two readings, September 6 and 24, 1910, 

 were made after the springs had been resting for about four months. 

 The constant is in general higher, but the greatest increase has come 

 at the middle of the range, so that the relation between stress and 

 strain is more nearly linear than before. After this prolonged period 

 of rest, the gauge has remained much more nearly constant than 

 before. 



Figure 5. Detail of a 

 hysteresis loop, showang 

 the method of avoiding 

 the effects of hysteresis in 

 making the readings with 

 the springs. 



The table shows the 



