412 Mr. L. R. Wilberforce on the Calculation of the 



whose diameters there was in consequence little uncertainty, 

 were quite irreconcilable. 



In order to compare the different experiments, a curve in 



which v was the abscissa and - the ordinate was plotted for 



each tube. Then, considering the curves for two tubes of dif- 

 ferent lengths but the same diameter, and assuming that no 

 appreciable effect was produced by the ends occupying differ- 

 ent positions with regard to the vessels into which they 

 opened, the ordinates for the same abscissa should have dif- 

 fered by a constant quantity ; and this quantity, multiplied 

 by the square of the diameter of the tubes and divided by the 



32 a 

 difference in length, should have been equal to , i. e. to a 



. . . . w 



constant multiple of the coefficient of viscosity. 



The values thus deduced for the coefficient were very dis- 

 cordant ; for example, those deduced from Poiseuille's tube 



(A) varied from -000403 x ^ to -000445 x ^ 



These differences were very large indeed when compared 

 with the concordance which seemed to exist between the 

 experiments of the first series ; but upon examination of these, 

 it appeared that their discrepancies were in reality almost as 

 o-reat, and that PoiseuihVs final comparison was only one be- 

 tween a comparatively small number of results whose agreement 

 was quite fortuitous. It was in fact found that the coefficient 

 of viscosity deduced from the experiments of the first series 



by the ordinary uncorrected formula varied from '000409 x |^ 



to -000437 x ^z ; and as here these differences occurred even 



when the velocity was so low that the correction must have 

 been quite insensible, it became evident that the errors of 

 experiment were too large to allow the results to be available 

 for my purpose. 



I had reached this stage of the investigation at the beginning 

 of the present year, and was hoping to carry out some expe- 

 riments of the same general nature as Poiseuille's in order, if 

 possible, to determine in some simple cases numerical values 



for the function </> ( -^— ), when my attention was called to a 



statement by Gartenmeister, in the number of the Zeitschrift 



fur Physikalische Chemie published on December 31st, that 



Prof. Finkener of Berlin bad suggested to him as preferable 



to Hagenbach's correction the same greater value which I 



