6 CHARLES G. ROGERS AND ELSIE M. LEWIS. 



practicable, or by reading to fractions of a millimeter upon the 

 galvanometer scale even smaller fractions could have been 

 estimated though such an estimate was made in only a few cases. 

 The scale used in these readings was placed at a meter distance 

 from the galvanometer and was mounted upon an arc of a circle 

 having a meter radius. This made it unnecessary to make cor- 

 rections in the readings as would have been necessary if a straight 

 scale had been used. It was possible to determine directly the 

 temperature difference between the two junctions from the 

 deflection, having determined the value of the parasitic current 

 at experiment. Our temperature differences were so small that 

 we did not find it at all necessary to make use of a potentiometer 

 in connection with the galvanometer for this work. 



ELIMINATION OF ERRORS. 



The elimination of possible sources of error has been touched 

 upon in the description of the methods, but the precautions 

 employed may be summed up briefly. Variations due to differ- 

 ences of temperature in the varying currents of the water bath 

 were checked by keeping the two junctions of a couple very close 

 together and at a constant distance from each other during the 

 course of an experiment. It was found that when this precaution 

 was taken that there was scarcely ever a noticeable difference in 

 the temperature of the water at the two points. 



Leakage to the system from stray electric currents was not 

 observed. Tests with the motor employed for stirring the water 

 in the thermostat, the greatest possible source of outside influ- 

 ence, showed no effect whatever upon the system. Variations 

 due to lack of uniformity in the wires used were practically nil. 

 One set of couples was used throughout the whole series of ex- 

 periments and the wires were carefully protected from strains. 

 Secondary couples were prevented through the careful insulation 

 from temperature changes of all wires and switch connections in 

 the system. Change of resistance due to alterations of tem- 

 perature of the lead wires would be practically negligible as the 

 wires were heavily insulated and the temperature of the room 

 showed no wide variations during the course of the work. Heat 

 production upon the part of the animals on account of injury or 



