Experimental Researches on Gravitation. 501 



Observed effect +mg. 0-00208+0-00007 



Newtonian effect of the Hg on the tare — mg. - 000S5 



„ receptacle of Hg + mg. 0*00007 



„ floats K and K' -nig. 000034 



„ Hg on the beam mg. 0-00000 



Correction for the displacement of the zero... + nig. 0*00001 



Greatest error admissible for asymmetry mg- +0-00009 



Net effect e=+mg. 0-00098+0-00016 



The four corrections given here for the Newtonian effects- 

 of the mercury and float K with the K' counterpoise have 

 been rigorously calculated. Their probable error is very 

 much inferior to the probable error of my observations. 

 The greatest error admissible for asymmetry in the mercury's 

 position relatively to the leaden ball, estimated at +0*00009 

 mg., is certainly superior to the true one: it has been my 

 purpose to exaggerate in its admission, to show that it cannot 

 cover the phenomenon that has been discovered. 



Hence ice have a net diminution of weight undergone by the 

 leaden ball weighing 1274: </?■., and, this diminution is equal to 

 0-00098 mg. (that is to say, 7*7 . 10" 10 of its value) due to the 

 fact of the ball being surrounded ly the mercury. 



Possibility of other causes of error. — In the detailed relation 

 of these experiments that will be published by the Accademia 

 dei Lincei in Rome, I discuss minutely the possibility of 

 other causes of error. Here I shall simply enumerate 

 them : — 



I. Perturbations of a mechanical character, such as the 

 effects of the weight of the mercury on the balance 

 of the projection lamp, the position of the scale, or 

 the deformation of the recipient U, augmentation 

 by compression of the density of the mercury, &c. ; 

 II. Perturbations of a calorific character; 



III. Radiometric actions ; 



IV. Electrostatic actions ; 

 V. Magnetic actions ; 



VI. Electromagnetic actions. 



I shall add only that such causes of error, if intervening, 

 cannot modify sensibly the result that has been obtained. 



Determination of tlte h constant. — The already verified 

 variation of weight allows the valuation of the universal 

 quenching constant h at least between certain limits of 

 approximation. I make use of the relations (9) and (10). 

 I must, however, introduce an hypothetical simplification in 

 the experiment as performed if I want to avoid a very great 



