A recapitulation of the main features of the moisture distribution in an ideal soil is given in order to emphasise a point previously neglected, namely, that the changes are not in the main strictly reversible, but fall into two series corresponding to the two directions of moisture change. The cellular nature of the soil pore space imposes a quantum character on the moisture changes over a great part of the higher moisture range. The individual cell does not fill or empty by smooth reversible changes but shows two unstable stages at which filling or emptying is completed at a bound. For falling moisture the suction level is that at which a meniscus can invade a cell through one of its narrow entries, which gives a value in the neighbourhood of 12 T/r for close packing. For rising moisture the suction level is that for which water returns to the cell by the collapse of a bubble in it, giving a lower value in the neighbourhood of 6·9 T/r. A still lower value of 4·5 T/r can be reached for a particular type of distribution confined to the lower moisture range.

All the above suction values are closely verified by measurements made with bronze balls and paraffin oil. A detailed exploration of the case for glistening dew and water has been made, which verifies the theory while illustrating more nearly the behaviour of an irregular soil. The two values which rule the two halves of the “hysteresis loop” for this case of natural or common packing are 6·0 T/r for falling moisture and 4·0 T/r for rising.

The new considerations resolve the differences of opinion raised by R. A. Fisher. It appears that the author has throughout supported a case which belongs correctly to the falling moisture series. Fisher considered a very limited series for the rising moisture case and a hypothetical reversible series, and has been in error in so far as his treatment was regarded as exclusive.