786 professor g. g. stokes ok the coefficient of viscosity* of air. 
Addendum. 
Note on the preceding Paper , by Professor G. G. Stokes, P.R.S. 
(Received January 14, 1886.) 
The consistency of Mr. Tomlinson’s different determinations of the coefficient of 
viscosity of air, notwithstanding the great variation in the circumstances of the experi¬ 
ments, and the consistency with one another of the numbers got by a different 
process by Maxwell, led me to endeavour to make out the real cause of the differ¬ 
ence, and I think the main part, at any rate, of it can be explained by a very natural 
supposition. 
The fact that Mr. Tomlinson worked with air in its ordinary state, whereas 
Maxwell’s air was dry, even if it tends in the right direction, would evidently not 
go nearly far enough. But it occurred to me that the effect of any error of level in 
the movable disks employed by Maxwell must have been much greater than might 
at first sight appear. For suppose a very small error 8 to exist, and suppose the fixed 
disks adjusted to be parallel to the movable ones in the position of equilibrium of the 
latter. Then the two systems must be, very nearly indeed, parallel throughout the 
motion, since the angle of oscillation of the movable disks to one side or other of the 
position of equilibrium is very small. If 2a be the whole amplitude, the greatest 
error of parallelism will be of the order Sa, and it would naturally appear at first sight 
that the effect of so small an error of parallelism must be insignificant for any such 
error of level as we can reasonably suppose to have existed. But a little consideration 
will show that this need not be the case when the distance between the fixed and 
movable disks is very small compared with the diameter of the latter. For suppose 
the disk to have been rotated through a small angle p round a vertical axis; the 
rotation p may be decomposed into a rotation p cos 8 round the axis of figure, and a 
rotation p sin 8 round a horizontal axis in the plane of the disk. As regards the 
former, the motion takes place as supposed in the investigation. But as regards the 
latter the disk oscillates about a horizontal axis in its own plane. Now, when the 
disks are very near one another this oscillation entails a squeezing thinner of the 
stratum of air opposite to one half of the disk, and a widening of the stratum opposite 
the other half, the two halves being alternately squeezed thinner and widened j and, 
since for such slow motions the air is practically incompressible, this transfer of air 
cannot be effected without a motion of the air along the surface of the disk far larger 
than what would be produced by an equal rotation about the axis of figure. Accord¬ 
ingly a very slight error of'horizontality in the movable disk might produce a sensible 
error in the result, though an error of direction of similar amount in the orientation 
of the fixed disk would be quite insignificant in its influence on the final result. 
This conclusion is fully borne out by the result of mathematical calculation founded 
on the equations of motion of a viscous uncompressed fluid. The calculation becomes 
