38 



Mr. C. Chree. 



quantity was poured into the dish. With bisulphide of carbon of 

 course a much lower temperature was employed, but even then its 

 heated top layer evaporated so fast as to affect the contact of the 

 liquid with the dish. For these reasons in all experiments on the 

 bisulphide the water was siphoned out of the dish. 



In the earlier part of each experiment the heat was of course con- 

 centrated chiefly in the upper layers. Still as about an inch inter- 

 vened between the dish and the platinum, the variation of tempera- 

 ture in by far the greater portion of the liquid layer was comparatively 

 small. Thus the error due to treating the conductivity as inde- 

 pendent of the temperature caunot be great. 



There are two possible disturbing agencies which require comment. 

 Any difference of temperature between the two copper-platinum 

 junctions in the liquid would produce a thermoelectric current. Care 

 was taken, however, that the junctions should be as nearly as 

 possible in the same horizontal plane. At the depth in question the 

 greatest possible difference between the temperatures at two points 

 differing a few millimetres in depth could not exceed a small fraction 

 of one degree. Thus the thermoelectric current, if existing, must 

 have been very small, and necessarily its variation, from which alone 

 any error could arise, must have been very trifling. Further, the 

 neutral point of copper and platinum is only about 70°, so they would 

 under the circumstances form a very weak couple. 



An attempt was in fact made to employ a thermoelectric couple of 

 iron and lead, whose neutral point is over 350°, one junction being in 

 the liquid and the other maintained at a constant temperature. This, 

 however, failed completely, owing to want of sensitiveness. Thus 

 there are various d priori grounds for neglecting the thermoelectric 

 effect in the actual experiment. This view was further justified by 

 actual trial, first by finding the deflection that followed when one of 

 the junctions was suddenly heated to a considerable temperature, 

 second by cutting out the battery during the usual experiment, and 

 observing whether shunting the platinum wire affected the galvano- 

 meter. Finally, in the various experiments on any one liquid the 

 battery current traversed the platinum wire sometimes in one direc- 

 tion sometimes in the other; and thus any possible thermoelectric 

 effect must have tended sometimes to increase and sometimes to 

 diminish the rate of variation of the galvanometer reading. The 

 small variation in the observed times of most rapid variation is thus 

 sufficient proof of the small disturbing action of the thermoelectric 

 effect, and the variation in the direction of the battery current would 

 further tend to eliminate any such small effect if existent. 



Another disturbing cause existed in the case of the sulphuric acid 

 solutions. These conducted electrolytically, and also attacked the 

 copper wires. By covering these wires with shellac varnish this was 



