kariib.J VISCOSITY OF GASES. 275 
temperature registered by the thermo-couple and the mean temperature 
of the helix of capillary tube are not necessarily the same. Each of 
these pyrometers furnishes its own thermal datum correctly ; but these 
data have reference to environments which are not thermally identical. 
In short, the degree of constant temperature throughout the space en- 
veloping the helix is as yet far from satisfactory; and the residual 
differences of temperature between the inside and the outside of the 
helix or between its top and bottom surfaces have not been rigorously 
allowed for. To this must be added the fact that the thermo-couple 
was not compared with the air thermometer at temperatures above 
1,300°, and that the calibration in question loses in accuracy when 
these high temperatures are approached. (Cf., Figs. 41 and 42.) Hence, 
in view of the great difficulties in the way of correct temperature meas- 
urement, it is hardly profitable to enter into more than a cursory con- 
sideration of the minor sources of error. 
In addition to the extraneous causes for incorrect temperature meas- 
urement, there is also an internal cause, due to the expansion of the 
transpiring gas from the pressure P to the pressure p. Meyer has 
elaborately discussed this phenomenon. Again, the purely convective 
effect due to the introduction of cold gases into the capillary tube is not 
to be lost sight of. In experiments of the next section I found a cool- 
ing effect as high as 20°. In the present experiments, where thin tubes 
and slow currents alone occur, the convection error is nil. 
To avoid incidental complications the radius of the capillary tubes 
here in question was chosen small (R <0.01 cm ), so that the Meyer for- 
mula fully applies. Again, preference is given to absolute methods of 
experiment; differential methods, inasmuch as they compare two mag- 
nitudes without fully characterizing either, would have encumbered the 
present research with an additional element of uncertainty. 
Diffusion. — The observed circu in flexion of the curve which repre- 
sents the mean distribution of the points in the diagram, Fig. 48, at a 
temperature near 1,000°, together with the fact that this change of the 
sign of curvature is much more pronounced for hydrogen than for air, 
points, I think significantly, to the occurrence of diffusion of gases 
through the walls of the platinum capillary tube. In the case of trans- 
piration experiments with air, the hydrogen gases of the buruer passing 
through the platinum septum combine with the oxygen of the capillary 
current within. There results an increase of the volume of the oxygen 
combined as 1:2; but as the water formed is absorbed in the pneu- 
matic apparatus, it follows that the volume V n actually measured by 
the burette is too small. Again, in the case of transpiration experi- 
ments with hydrogen the prevailing diffusion is from within outward, 
so that hydrogen simply leaks out of the tube. Thus the volume of V 
actually measured is again too small. Hence in the case both of air 
and of hydrogen, since F is negatively in error, the error of r/', which 
varies inversely as F , will be positive. This accounts for the circum- 
(929) 
