88 Mendenhall — Free Pendulum as a Time /Standa?'d. 



second and only an inexcusably careless observer would ever 

 be a second in error. After having ascertained the " coinci- 

 dence interval " and observed the first coincidence, the happen- 

 ing of any one in the future can be quite closely predicted. 

 It is desirable to allow the pendulum to swing through a period 

 equal to ten coincidence intervals and although only the first 

 and last observations may be used it is well to observe two or 

 three at the beginning and end of the period as a check. An 

 idea of the accuracy of the method may be formed by assum- 

 ing that an error of one second is made in observing the coin- 

 cidence either at the beginning or end of the swing. It will 

 be seen that the error of the result will be less than one part 

 in two and a half millions. As a sample of the uniformity of 

 results which is sometimes attained under favorable conditions 

 the periods of a pendulum known as A 3 as obtained from a 

 Negus break-circuit sidereal chronometer in three separate sets 

 of vibrations extending through about an hour each, made on 

 the afternoon of October 23, are here given : 



1st -5006279 



2d -5006281 



3d -5006280 



If such a pendulum be properly cared for there appears to 

 be no reason why under the same conditions, its oscillation 

 period may not remain sensibly constant for an indefinite 

 period. As an example of constancy during a short period, a 

 little over half a year, but under unfavorable conditions the 

 following results snowing the oscillation period of three pendu- 

 lums as measured in March last and in October. In both cases 

 time is referred to the standard sidereal clock of the U. S. 

 Naval Observatory. Between the first determination and the 

 last, the pendulums traveled about fifteen thousand miles, 

 were several times landed from a vessel in the surf and with 

 difficulty, and were exposed to great changes of temperature 

 and conditions of weather. The remarkable agreement in the 

 case of all justifies the assumption that a sufficient degree of 

 permanency is attainable. 



Mean Periods. Washington. 



A-l. A-2. A-3. 



March, 1891 -5008779 -5007667 '5006702 



October, 1891 -5008759 -5007668 -5006696 



When one of these pendulums is compared by means of the 

 flash apparatus above described with a good break-circuit chro- 



