390 



Messrs. E. H. Scott and R. H. Curtis. [May 20, 



serve as a standard with which the coefficients obtained by means of 

 the machine might be compared, was established by calculating them 

 from the odd and even hours, quite independently, for all the seven 

 observatories. 



The outcome of this experiment was thoroughly satisfactory, and 

 the entire series of results obtained both by calculation and from 

 the machine was published as Appendix IV to the Quarterly Weather 

 Report for 1876, together with a Report prepared by Prof. Stokes, 

 the concluding paragraphs of which may be quoted here, since they 

 sum up in a few words the conclusions arrived at. 



" Disregarding now the systematic character of some of the errors, 

 and treating them as purely casual, we get as the average difference 

 between the constants as got by the machine and by calculation from 

 the twenty-four hourly means 0*065°. It may be noticed, however, 

 that the numbers are unusually large (and at the same time very 

 decidedly systematic) in the case of the second cylinder of the first 

 order for which the average is as much as 0*150°, the seventh 

 of a degree. 



" If \ be omitted, the average for the remaining cylinders of the 

 machine is reduced to 0*047°. 



" We see, therefore, that with the exception perhaps of b lt the 

 constants got by the machine for the mean of the days constituting 

 the month are as accurate as those got by calculation, which requires 

 considerably more time, inasmuch as the hourly lines have to be 

 drawn on the photograms, then measured, then meaned, and the 

 constants deduced from the means by a numerical process by no 

 means very short." 



The curves for the twelve years 1871 to 1882 inclusive have now 

 been passed through the machine, and the results obtained have been 

 carefully checked so far as the arithmetical work involved is concerned, 

 upon a plan approved by the Council. No direct check, short of 

 passing the curves a second time through the machine can however 

 at present be put on any portion of the results except as regards the 

 means, which have been compared with the means calculated from 

 the hourly readings obtained by measurement from the curves. The 

 results of this work will be published as an appendix to the " Hourly 

 Readings from the Self-Recording Instruments," for 1883, but the 

 general results may here be stated. 



As a rule, the monthly means yielded by the harmonic analyser 

 agree well within a tenth of a degree with those obtained by calcu- 

 lation from the hourly measurements of the curves ; and although in 

 some exceptional cases larger differences have been found, amounting 

 in rare instances to as much as half a degree, it is probable that 

 generally these are less due to defects in the working of the instrument 

 than to other causes. In some cases large breaks in the curves, due 



