376 



Scientific Proceedings, Royal DuUin Society. 



specific volumes against the temperatures and to draw curves to pass as 

 evenly as possible through the points. Now it is a matter of common 

 experience that, in drawing a curve through a series of points, even when the 

 errors of experiment are fairly regularly distributed, the deviation of the 

 drawn curve from its true position is likely to be smaller in the middle 

 portions than at the extremities. And when the trend of the curve— giving 



the values of is considered, it is found that the errors are likely to be 



much greater at the extremities than near the middle. 



In the case of the volumes of saturated vapour, liowever, tlie errors of 

 experiment are usually greatest at the lowest temperatures; and tlierefore the 

 deviation of the drawn curve from its true position, and, to a still greater 

 extent, the error in the trend of the curve, is liable to be very much more 

 pronounced at its lower extremity than in any other part. 



Having been asked by M. Chappuis, in 1907, to furnish a table of 

 densities of saturated vapour under normal pressure, and of the rate of 

 change of density per millimetre rise of pressure, for the " reciieil des 

 constmites 'physiques," it became necessary to consider the data at the lowest 

 temperatures more carefully than had hitherto been done. 



An examination of the whole of the data leads to the conclusion that the 



