92 Proceedings of Royal Society of Edinburgh. [sess. 
which was placed upon the surface of the cylinder, and its position 
accurately determined. 
For the microscopical examination of the impressions a magni- 
fication of about fifty proved sufficient. This was obtained by 
using Zeiss 5 objective A. A. and ocular 2, provided with an ocular 
micrometer, which was divided in two directions perpendicular to 
each other, into fifty parts. Further magnifying proved inadvis- 
able, as the light failed, and the curved surface of the cylinder 
hindered accurate results. 
When a phonograph cylinder, the contents of which were exactly 
known, was to be subjected to measurement, the place of the 
cylinder on the mandrill was once for all fixed by means of a cross 
mark made in the surface, the position of which was accurately 
determined by means of the drum, P, and the scale, SS. Then the 
position of each vowel sound, and, if necessary, the number of its 
periods, were determined by turning the cylinder in the same 
direction in which it had been turned when taking them up. 
At the same time, by the microscopical examination of the periods 
of each vowel, their fitness to be measured could be judged of. 
After this “grazing off 55 — so to call it — of the whole cylinder 
the measuring of the transverse diameter of the fit periods could 
be undertaken. As the latter only gives the ordinate of the curve, 
it was necessary to obtain the abscissa corresponding to it. 
The value of the abscissae was determined in two ways. 
I. First by measuring the diameter of a period only on that 
spot where it attained either a maximum or a minimum, and by 
measuring the distance from maximum to minimum by means of 
the perpendicular part of the ocular micrometer, the horizontal 
part of which had served to determine the transverse diameter. 
But as the curved surface of the cylinder proved a great hin- 
drance for accurate measurements, in my later experiments the 
distance of maximum to minimum, i.e., the value of the abscissa, 
was determined by means of the drum, Q, which should indicate 
this distance on a tenfold magnified scale. 
By-and-by, however, I came to the disagreeable conclusion that 
the drum, Q, did not do what it should do. 
By measuring out the length of some subsequent periods of 
sounds, such as that of a cornet-a-piston, or that of a telephone 
