D'UN PENDULE A TRÈS-LONGUE PÉRIODE. 207 
rom the inclination which the oscillations present, and 
regard what remains as the true dynamic inclinations, 
we cannot correctiv deduce from them the dynamic con- 
ditions by which the rolling is governed. In the latter 
case, we shall find ourselves equally obliged to select, 
and equally able to select, the true line as the zero of the 
oscillation. 
I do not mean that absolutely no error thus introduces 
itself; but in effect, I do not believe that the error exceeds 
0°,40 in any case in which the curves have been critically 
examined, excepting in one or {wo instances, in whichit 
has been evident that a sudden motion has been impres- 
sed on the wheel by some casual external force. In these 
instances, the zero line, «which we find ourselves seeking, 
presents manifestily a discontinuity of curvature which 
pecessarily renders 1t of no value as a zero line at that 
particular part, because we are not abie with even appro- 
ximate exaciness to measure the disturbance at that point, 
and this visible deformity warns us to disregard that part 
of the record. Bat if we omit these parts as exceptional, 
all the other parts of the record appear to be as good as 
can be desired ; and by comparing it, throughout, with 
the pendulum record, we obtain an irreproachable mea- 
sure of the inclination of the wave, continuously. 
£ have already mentioned that in the ship in which 
this automatic instrument was fixed, another pen at the 
same time was governed by an observer who constantiy 
directed an index arm atthe horizon in the same man- 
ner as you also operated, so that thus two records of 
the absolute angle were produced on the same paper in 
that ship, and these were carefully compared. For the 
most part, the agreement is admirable; but occasion- 
