( 415 ) 



perature is greatest at the highest temperature. At first 1 intended 

 to calculate this disturbance from the measurements of the fall of 

 temperature mentioned in the preceding part. But I have changed 

 my mind because it appeared that in spite of accurate placing of 

 the thermo-elements, deviating falls of temperature occurred with 

 different measurements. For the most part, these deviations must be 

 ascribed to the existence of an other slope of temperature. There- 

 fore of late I have determined the fall of temperature on both these 

 places before and after the measurement of the longitudinal effect. 

 If we knew exactly the difference of temperature between the two 

 places of contact we might have expressed the correction in percen- 

 tages. Although the temperature of the places of contact between 

 German-silver and copper is exactly known, there is however no 

 sufficient certainty that this is at the same time the temperature of 

 the place of contact of bismuth and copper. The same uncertainty 

 is the reason that the thermo-electric power of bismuth can also 

 not be exactly deduced from these measurements; the values of the 

 difference bismuth-copper, calculated from the results of different 

 experiments, range from 8400 to 10500 per 1° for temperatures 

 ranging from 20° to 50°. It should be remembered that in the 

 place B we find a slope of temperature from 2,5° to 5° per ww., 

 while the whole difference of temperature was 20° to 30°. For 

 these reasons I have for the time being given up a calculation 

 of the longitudinal effect in absolute measure and have calcu- 

 lated the correction by taking for the thermoelectric difference 

 bismuth-copper 10000 per 1°; for the thermo-element German-silver 

 and copper was found 1590. In the correction there remains an 

 uncertainty of probably not more than 10°/^ ; the correction itself 

 amounted at most to 25% of the total effect, so that for tliis there 

 remains an uncertainty of 2, 5%; as moreover the fall of temperature, 

 as afterwards appeared, shows almost the same variation with the 

 magnetic force as the longitudinal-effect, the proportion of the phe- 

 nomenon in different magnetic fields will not be changed by the 

 error. 



A great number of measurements had been made already before 

 the method of observation described here was applied. In order to 

 use these in the calculation, I interpolated from all the sets of 

 observations the percentage variation for the fields 2000 to 6000 ; 

 then I always assumed 10 as the value for the field 6000 which 

 was the result according to the last method, and changed the other 

 figures proportionally. The following results were obtained. 



