122 G. H. KNIBBS. 
the volume method gave for the diameter of a circle of equal area : 
00376 c.m., while direct measurement and the assumption of an 
elliptical contour gave ‘00377. The difference 1/376 is small but 
by no means insensible, for since the diameter is raised to the — 
fourth power this difference will lead to an error of 1/94 in the — 
deduced value of the viscosity. 
Computation of mean radius of efflux.—Poiseuille used for the 
mean radius the half sum of the geometrical means of the terminal 
radii! But it has been shewn in § 11—formule (18) and (J8a)—_ 
that the geometrical mean is too great, and in § 12 and § 15—_ 
formulz (23a) and (29) that the mean of the terminal means is | 
also too great. I have computed the ‘mean radius of efflux’ by : 
the principle of formula (24) using formula (26) and occasionally 
checking by (29). With regard to the neglect of small terms in 
the computation dc., the condition has always been secured that — 
the error shall be small in relation to the probable uncertainty 
the data. 
Dimensions of the tubes.—Poiseuille measured the lengths ds 
the tubes used by him by means of a beam compass*—compas 4 
verges—reading by a vernier to ‘005 em. and permitting, by 
estimation, a record.to half this amount. 
When a series of efflux experiments with a given length of vabe 
was completed, a portion of the free end was cut off by means 
of a file? a procedure that was several times repeated. This 
permitted the measurement of transverse sections at different 
distances along the axis, and the regularity of these gave & 
general indication of the symmetry of the tube. The measure : 
ments of each section are taken in account by me in the comp 
tation of the ‘mean radius of efflux.’ ; 
Effect of temperature on ‘dimensions. —When, as was the cast ‘ 
in Poiseuille’s experiments, the whole of the reservoir of supply— 
the bulb V Fig. 4—the flowing liquid, and the capillary are mail 
tained at the one temperature by immersion in a bath, or otherwi 
. ‘Thia. § § 114, p.513. 2 Ibid. § 83, p.497. 3 Ibid. 
