of Edinburgh, Session 1885-86. 593 
ice with ice-cold water for the same time. But the first of these 
two operations was found to melt the ice on its surface; while 
during the second, the ice-cold water froze on the surface of the 
previously cooled ice. To obviate this difficulty, a thin sheet of 
tinfoil was placed on the suface of the ice; on this was laid for five 
minutes the freezing mixture of pounded ice and salt; and during 
the next five minutes the surface of the ice was simply exposed to 
the air. This alternation of cooling and heating was then regularly 
carried on until the thermal changes at the thermo-electric junctions 
were observed to be periodic ; after which readings were taken at 
equal intervals of time, viz., 20 seconds. 
The metal- box, containing the cube of ice, was placed in a bath 
of ice and water, which provided a steady temperature for one of 
the junctions in each of the two thermo-electric circuits. By means 
of a commutator, one junction (in the ice) at a time was placed in 
circuit with a Thomson’s dead-beat galvanometer (resistance 1 *5 
ohms at 9° *4 cent.) ; the circuit was so changed as to include the 
two junctions alternately. In this manner it was found possible to 
take readings every ten seconds. 
It was found difficult to attain a strictly periodic change in the 
temperature at each of the junctions. This may be due partly to 
the fact that freezing mixtures themselves undergo sudden changes 
of temperature, and partly to the circumstance that the completeness 
VOL. xm. 2 R 
