448 0. N. Rood—Application of the Horizontal Pendulum to the 
as to render the _ 7 eben invisible. Still, in spite of 
ficient number of observations. To illustrate the cae of 
the apparatus, 1 may remark that children playing on an 
iron bridge, 360 distant, caused temporary deflections of one or 
two ae and that similar deviations were caused by the 
lower notes of an organ in a neighboring church, the —— 
to produce permanent effects; this evil, however, practically 
only cpa bates the observer to temporary interruption. 
Changes of temperature produce two different kinds of effect 
on the Sibattibad: in the first class we may place those which 
would compensated by turning the leveling screw Q, fig. 4; 
in the second, those that would be neutralized by turning 
either of the other leveling screws. It is evident that changes 
of the first class alter the sensitiveness of the apparatus or the 
value of a scale-division, but on account of the compara- 
tively large distance between IK and Q, these effects are small, 
and may be rendered almost insensible by attention to con- 
sate of bone? rir thus far they have not proved a source 
of annoyance 
e case is different with the second class; the leveling 
screws I and K being separated only by a distance of 50 mm., 
it is necessary, on their account mainly, to preserve the temper- 
ature of the apparatus as constant as possi le. The means of 
changes of the kind now under pouanioration will communi- 
cate to the pendulum a continuous motion toward the right or 
left, and will finally drive it to the stops, and that all the read- 
ings will be affected with a constant or changing error. If the 
motion is so slow that the operation of reaching the stops con- 
sumes an hour or even less, it is not a source of serious inconven- 
ience, and the errors thus introduced can easily be eliminated. 
eneral mode of experimenting.—In all cases the micrometer 
screw rests directly or indirectly on the body the change in 
whose dimensions is the subject ‘of study, and the first pro- 
ceeding will naturally be to ascertain whether the different por- 
tions of the apparatus are at rest relatively to each other, or 
approximately so. Afterward the value of a scale-division can 
