G40 MR E. M. WEDDERBURN AND MR A. M. WILLIAMS. 



(the origin being taken at the vertex of the parabola). The teak block was given a 

 teak backing, and a front to the trough was made by clamping on a piece of plate- 

 glass, half an inch thick, by means of four strong clamps. A piece of rubber tubing was 

 clamped between the wood and the glass, and this made the trough sufficiently water- 

 tight. There was thus obtained a trough with a glass front, the longitudinal section of 

 which was accurately parabolic, the cross-section being rectangular. The glass front 

 enabled oscillations to be observed. It could be easily removed, and the shape of the 

 trough altered by fitting in blocks of wood of various shapes. In this manner the ends 

 of the trough could be truncated at any desired point. 



§ 47. Oscillations were started in two ways. The method first adopted was to move 

 a plate slowly backwards and forwards at the node in the lower liquid. The plate had 

 attached to it a long thin handle, so that the disturbance caused in the upper liquid by 

 the stirring was not great. It was found, however, that oscillations could often be more 

 easily started by slow to-and-fro stirring in the upper liquid. In fact, one to-and-fro 

 motion in the upper liquid in the large trough was sufficient to start an oscillation of 

 considerable amplitude. This is specially interesting because temperature oscillations 

 in lakes are caused by forces acting only in the upper layer ; and it indicates how very 

 easily temperature oscillations may be caused. The oscillations were timed with a stop- 

 watch reading to fifths of a second, and usually a series of twenty or more oscillations 

 was timed. Where there was any considerable divergence between the determinations 

 of the period, ten and sometimes more determinations were made. The tabulated 

 results give the greatest variation from the mean value. The " probable error" was not 

 calculated, but would, of course, be much smaller than the greatest variation. At times 

 the period of the oscillations did not appear to be very constant, especially when the 

 depth of the upper liquid was small. It was then difficult to start pure oscillations of 

 any considerable amplitude, and when the oscillations were small it was difficult to time 

 the beginning and end of an observation accurately. 



§ 48. The principal results are tabulated below, the contractions used being the 

 same as before. In cases where there are observations in the plain parabolic trough, 

 and in the truncated trough with the same depths of liquids, a temperature correction 

 has been applied to the latter, to make the results comparable. The last column gives 

 the ratio of the observed period to the calculated corrected period, taking means of the 

 periods for the plain and the truncated basin, where there are observations of both for 

 the same depths. 



