BIRTH AND AFFINITIES OF CRYSTALS 125 
the supersolubility curves for triphenylmethane in organic 
liquids, and have shown how they explain the course of 
crystallisation even when that substance forms crystalline 
compounds with the solvent. These authors suggest that the 
metastable condition is to be explained by the increased solubility 
which is known to belong to very small crystals. The first 
products of spontaneous crystallisation must accordingly be 
more soluble than the larger crystals for which the solubility 
curve is determined. They cannot therefore persist through 
the early stages of growth until the degree of supersaturation 
becomes sufficient to counteract the solubility. 
Many peculiarities or apparent anomalies in the crystallisa- 
tion of solutions can be explained when account is taken of 
their passage from the metastable to the labile condition. 
An experiment familiar to every schoolboy is the scratching 
of the inside of a test tube containing a strong solution with a 
glass rod, and the appearance of the first crystals along the line 
of scratch. This is often explained as due to a tendency of 
crystals to grow on rough surfaces and protuberances rather 
than on a flat, smooth surface. 
If this explanation were correct, it ought to make no 
difference whether the scratch be made before or after the tube 
is filled with the solution, and the crystals ought to appear 
along the lines of old scratches. In reality, it is necessary 
that the scratch should be made within the liquid; moreover, 
we have found that no effect is produced until the solution has 
been supersaturated to the exact point at which it is becoming 
labile. The process of scratching merely supplies the mechanical 
disturbance sufficient to make the crystals grow in the labile 
solution. 
It is true that scratching may produce a line of crystals in a 
metastable solution; but this only occurs if the rod is intro- 
duced from outside, and has caught up crystalline germs in 
the air or at the surface of the liquid; no effect is produced 
by scratching with the end of a rod which has been kept in 
the metastable liquid throughout the cooling. 
Now a strong aqueous solution, being merely a mixture of 
the dissolved substance and of ice, both in the liquid condition, 
a similar supersolubility curve ought to be obtained for the 
spontaneous freezing of ice in solutions which contain excess 
of water. We have not conducted any experiments at sufficiently 
