86 BULLETIN OF THE 



ernor being a vibrating spring. The observations of 510 stars 

 over nine transit tlireads were examined from each chronograph. 

 These stars were observed in the last six months of 1872. 



The following record of an observation of a Canis Minoris 

 will illustrate the manner of recording the work in the observing 

 book : — 



40^2 



42. T 



44.3 



50.5 



1^ 32-^ 52.5 



54.5 



0.8 



2.3 



4.8 



These records are read from the sheets by means of scales care- 

 fully graduated lo tenths of the second's interval. 



The number of times that each figure from to 9, inclusive, 

 occurred in the tenths of seconds in the one thousand and twenty 

 observations, were carefully counted, and the results are shown 

 in the following table : — 



From Cheonogkaph A. 



01234567 8 9 



729 258 438 434 450 494 459 446 432 450 



Fkom Chronograph B. 



0123456789 



962 110 349 436 480 464 463 431 471 424 



A slight inspection of the results from chronograph A shows 

 a striking dift'erence between the Os and Is, and a similar differ- 

 ence occurs in the work from chronograph B between the Os, Is, 

 2s, and 9s. 



Omitting those numbers and using the results from both chro- 

 nographs for the figures 3 to 8, inclusive, it appears that the 

 probable error of a single sum is less than -g'^- of the mean. As 

 this probable error decreases with the increase of the number of 

 observations examined, it seems perfectly safe to conclude that 

 there is no tendency in my work to record any particular figures. 



The discrepancies in the results for 0, 1, 2, and 9, can be ex- 

 plained by the action of the chronograph pen and by the pecu- 

 liarities of the two chronographs. 



If the armature of the magnet which carries the pen be so 

 adjusted as to make the length of the clock-mark more than ^^^ 

 of the interval corresponding to one second, all marks of the 

 observer's key v.'hich might happen to fall in the time while the 

 clock keeps the circuit closed will be at least uncertain, and can 



