276 DR CARGILL G. KNOTT ON 
These considerations led me to abandon my first method of experimenting ; 
and in the modified method finally adopted, the temperature of the upper plate 
of the condenser was kept constant, while the temperature of the lower was 
made to vary. This required the contact to be instantaneous, so that only one 
reading could be taken between each preparation of the surfaces and observa- 
tion of the variable temperature. During this interval the upper plate was 
laid upon an iron slab, and thus kept at the temperature of the room; and 
just before the apparatus was reset for observation the temperature of the 
lower surface was noted, and both surfaces were polished and dusted as usual. 
The first experiments were made with two iron surfaces, which, after sufficient. 
polishing at the ordinary temperature of the air, gave no deflection on separa- 
tion after contact. The lower surface was then heated up to 70° or 80° C. in 
the manner formerly described, and then allowed to cool, while at rapid 
intervals instantaneous contacts were made with the upper surface, each 
contact being made as soon after polishing as possible. In this way I found 
that iron hot was strongly negative to iron cold, and apparently more negative 
the higher its temperature—in other words, the difference of potential between 
iron and iron increases with the difference of temperature, being zero when 
the temperature difference is zero. A glance at the representative curve 
(Diagram, fig. 2) shows the nature of this change. The six different symbols 
represent six different curves, five of which give the results of as many 
independent experiments, while the sixth (represented by the circle and dot) 
is the average curve formed by the combination of the others. Each point on 
any one of the five primary curves, is, as far as possible, the mean of five 
consecutive readings—a method of reduction which recommended itself as 
giving the most probable value for the mean contact. Each point of the 
final mean curve is obtained by taking the average of all such points as lie in 
the same temperature decade. Subjoined are the tabulated results of these 
experiments, the upper row of each of which gives the temperature of the 
lower condenser plate, and the lower the corresponding deflection on the 
electrometer scale. 
EXPERIMENT I, (February 27, 1879). 
(Curve symbol -), 
Temperature (in degrees C.),. . : 53°8 49 +4 45:4 31 23°1 
Deflection, . . : ° ' ; 21'8 21 17-4 10:8 72 
EXPERIMENT IT. (February 28). 
(Curve symbol +). 
Temperature, : : : ; : t aa 3 40°7 31°6 224 
Deflection, . : : é : : Ly 252 14:7 fia’ 
