516 Mr. J. S. G. Thomas on 



1*5 amp., but an examination of fig. 5 shows that such 

 inversions do occur in this case also *. 



In Plate X. fig. 7 the ordinates are proportional to the 

 square root of the velocity of the air-current. It is seen 

 that in every case there is a region over which the graph is 

 a straight line, the extent of this straight portion being- 

 greater the larger the current employed. In the case 

 employing a current of 1*5 amp., this straight-line portion is 

 seen to extend over the region included between the points 

 P and Q. The curves become concave towards the axis of 

 deflexions in the region of the origin, the value of the de- 

 flexion at which this occurs being greater in proportion as 

 the electric current employed is increased. The point at 

 which this occurs has not been attained in the case where 

 a current of 1*1 amp. was employed. The straight portion 

 of the several graphs is succeeded by a portion convex to 

 the axis of deflexion. With the use of still larger electric 

 currents, within the limit of destruction of the filament, the 

 straight portion can be considerably extended. 



The resistance of the exposed wire was determined in 

 every case when exposed to the various air-currents, the 

 electric current in the bridge being adjusted to its appro- 

 priate value when balance of (he bridge was restored by 

 unplugging resistances from the box. The resistance of 

 the protected arm is immediately calculated from a know- 

 ledge of the drop of potential occurring across it, and the 

 resistance of the exposed wire calculated therefrom in the 

 usual manner. The results obtained are shown in Plate XL 

 fig. 8. It will be observed that the effect of the free con- 

 vection current is very pronounced at the lower velocities, 

 and that, moreover, the velocity at which the change of 

 curvature of the respective graphs occurs is greater the larger 

 the heating current employed. In Plate XI. tig. 9 the abscissae 

 represent resistances, and the ordinates are the logarithms 

 to base 10 of the volumes of air passing through the pipe 

 measured in cubic feet per hour. It is seen that within the 

 range of mean gas velocities employed, and where the effect 

 of the free convection is small, if V denote the volume of 

 air passing through the pipe in cubic feet per hour, and 

 R the resistance of the wire, the relation between V and R 

 is, outside the region of low velocities where the effect of 

 the free convection current is of importance, of the form : — 

 R=« — Mog ]0 V. 



* Although the curve for c=l"4, as shown, cuts the curve for c=l'5 

 before that for 1*3 cuts the latter, it is evident that such order of cutting 

 may be attributable to experimental error. 



