144 PROCEEDINGS OF THE AMERICAN ACADEMY. 



;>j = 0.00177 mm. k = 0.00450 

 p^ = 0.00024 mm. 4 = 0.00229 



Using these values, we find for m the value 0.00194. If now this 

 quantity m be inserted for K in the e(j[uation relating p and /, as given 

 above, viz. : 



C can be determined for the different pressures, or C having been de- 

 termined for the higher pressures, for example for those between 1 mm. 

 and 0.1mm., the values oi jj corresponding to the lower values of / 

 can be solved for. The value of A used in this process was 0.1655, and 

 the value of C used was the mean of the first seven values given in the 

 fourth column of Table I. The fourth and fifth columns of Table I 

 give these results. 



A comparison of the numbers given in the fifth column with those 

 in the first show that the values of p deduced irom the observed values 

 of / are in general greater than the observed values of p. 



It must be admitted at once that the foregoing method of getting 

 the value of ^ is a very imperfect one, and it is inserted only tenta- 

 tively. It is now proposed to remove the sulphur and silver tubes 

 and place a vessel containing liquid air so as to surround a portion 

 of the tubes connecting the pump with the apparatus, so that not only 

 may all vapor be removed, but also that the very highest possible 

 vacuum may be reached. The decrement will then be measured. This 

 should give the value of the part of the decrement due to friction in 

 the fibre. The liquid air will then be replaced by liquid carbon diox- 

 ide, which will remove the vapor but not the gas to be experimented 

 with. It is hoped that this method of procedure will settle the only 

 point that seems to remain in doubt in this part of the investigation. 

 The full discussion of the law relating I and p is reserved until this 

 step has been taken. 



With regard to the results given in Table I for the transpiration in- 

 strument it must be stated that the smaller numbers in the third col- 

 umn may quite easily have an error of ten per cent. Figure 7 shows 

 the results graphically. In this figure, a unit on the axis of abscis- 

 sas corresponds to 0.01 mm. of pressure, while on the other axis a unit 

 represents 10° of torsion. That portion of the curve which corresponds 

 to pressures below those for which the torsion is a maximum, ap- 

 proaches a straight line, and it is apparently a line which passes very 

 nearly through the origin. It is perhaps allowable to assume that it 

 does go through the origin ; for the force here, unlike the friction in the 



