106 
Transactions of the Boyal Society of South Africa. 
sight. Selecting two turning-points (points of symmetry, of course, if 
such are available), we divide the difference between the corresponding 
points by the number of intervening major oscillations. The result will 
be an approximation to Tj, say T', ; but generally only an approximation, 
because the turning-points are displaced by the other seiches, mainly by 
the binodal. 
Now residuate with respect to T',. The result will be a curve in which 
the uninodal harmonic is very much reduced, and will show the binodal in 
predominance. From this determine an approximation to T2, say T^. 
Now return to the original limnogram, and residuate with respect to 
Tl. The result will be a curve from which the binodal harmonic is 
nearly absent. Hence the turning-points of the uninodal will be much 
less disturbed than before ; and we can now get a better approximation 
to T„ say T','. 
Eesiduate now the original limnogram with respect to T'/, and we get 
a curve in which the binodal is less disturbed than before. We can there- 
fore make a closer approximation, say T',', to T2 ; and so on. 
The process can, if necessary, be repeated until the accumulated errors 
incidental to the manipulation obliterate the essential features of the 
diagram with which we are dealing. 
In this way the periods of two or even three seiches can be deter- 
mined, one after the other, from a comparatively short large-scale limno- 
gram. We discovered in this way seiches the existence of which had not 
been suspected ; and without this process we should not have obtained a 
determination of the period of the trinodal seiche of Earn at all, which 
never occurred pure, and had always a comparatively small range. The 
process was also used in purifying compound limnograms with a view to 
determine nodes" (Trans. Eoy. Soc, Edinburgh, vol. xlv., part 2 (No. 14), 
p. 385). 
