300 Transactions of the Boyal Microscojpical Society. 
every measurement, so that on the average the same number of 
threads probably represented nearly the same distance. In the 
case where a spring is more and more compressed and the number 
of threads embraced by the nut is constantly increasing (although 
some compensating action may arise), the screw and parts are 
certainly submitted to very varying and uncertain conditions. 
Another source of error arose — shahe of rotation. Having aban- 
doned the usual form of micrometer construction (a revolving nut 
with a constant change in the number of threads embraced — a 
plan, one would think, fatal to all delicate accuracy), the next diffi- 
culty was to ensure to the steel screw absolute advance and retreat 
without rotatory shake or motion. For this purpose slides were 
also abandoned. This action, the most important part of the in- 
strument, should now be described. 
A lever is affixed to the cylinder of steel and bent at right 
angles ; it carries an ad.justable weight. This weight slides upon 
a flat edge formed parallel to the axis of the steel screw by a most 
careful process, tested by a carefully prepared spirit-level for 
parallelism with the axis of the screw. 
On lifting the weight slightly, the lever rotates the screw 
through a small angle ; and this lever forms a constant test of the 
efficiency of the screw action of the greatest sensitiveness. 
A further action put into motion by a fine screw gives to the 
advance of the film-forming surface, or prism-lens, a movement of 
the millionth of an inch. 
Supposing that the recording wheels have advanced several 
turns, the weight and lever also advance on the smooth edge 
already said to be formed parallel with the axis of the screw. 
The constancy of the weight preserves the screw in one normal 
fiducial position, as regards its liability to rotate on its axis. An 
error of one hundredth of an inch in the sliding edge would pro- 
duce an error of the reading of less amount than the hundred- 
thousandth of an inch.* But the lever advances so very slowly, 
as the wheels rotate the nut upon the screw, that this error appears 
to be almost destroyed. 
An arrow-head shows upon the face of the differential wheels 
the number of turns taken by the nut. The instrument is self- 
recording, and reads to four places of decimals, from the hundredth 
of an inch to the hundred- thousandth. Two wheels, divided into 
98 and 100 teeth respectively, run in gear at will with a long 
pinion of ten leaves, carrying a wheel showing the hundred- 
thousandths of an inch. 
The prism end of the screw passes through the ground socket ; 
* The path of the weight on the lever would be for a complete rotation about 
20 inches ; the proportion of to this is 2000, and the two-thousandth part of a 
revolution is two-thousandth of = a o^oo- 
