412 Proceedings of the Royal Society 
A little later after the water had acquired nearly the temperature of 
its maximum density a change took place. In addition to the two 
currents described, another current began to set in, flowing on the 
surface and over the other two currents already described. Its 
direction being from the cold, that is in the opposite direction to the 
previous surface current. Its return current coming back with the 
previous surface current. This new surface current was produced by 
the water being cooled below its maximum density, and rising, instead 
of sinking as before. The effect of this new surface current was also 
indicated by the upper thermometer, which began to fall much 
faster than the other two. These two sets of currents kept flowing 
for some hours, gradually getting slower and slower, particularly the 
lower one, which seemed to stop after about four hours from the be- 
ginning of the experiment, by which time the water at the bottom 
had arrived at a temperature of about 40 a . After eight hours almost 
all circulation had ceased even by the surface. 
While the temperature of the water was above 39*2°, the heat was 
distributed by convection currents, causing all the water to circulate, 
and the temperature fell at about the same rate at cdl depths , and 
further it fell from 52’5° to 41*5°, or 11° in two hours. After the 
maximum density had been attained, and the cold current, instead of 
sinking to the bottom, flowed over the surface, the upper ther- 
mometer fell 7° in two hours, while the bottom thermometer during 
eight hours only fell 2°, being only 39*5 at the end of that time. 
This experiment proves that the downward convection of heat 
ceased about the temperature of maximum density, as the lower 
thermometers ceased at about that temperature to fall at the same 
rate as the upper one. It is true the lower thermometer did fall 
below 39*2°, but it did so only very slowly, part of the fall being 
probably due to convection currents caused by the heat of the sides 
of the trough taking up the dense water, its place being supplied by 
the overlying colder water. It must also be remembered that the 
lower thermometer was only 10*5 cm. from the surface, so that the 
bottom water would lose part of its heat by radiation and conduc- 
tion. 
On removing the refrigerator and applying heat a reverse series of 
phenomena was observed. The hot water sunk and gave rise to 
a circulation, the surface current being towards the heat. The 
