1867.] made with a Rigid Spectroscope. 15 



I ventured to request his opinion thereon, and with his permission beg 

 to annex his reply with Mr. Stewart's remarks. 



" Cambridge, 29th May, 1867. 



" My dear Sir, — I have read and carefully considered your Spectro- 

 scope paper, and send you my remarks. 



" On examining Captain Mayne's diagram, we are struck with the 

 paramount influence of change of temperature. I say temperature, 

 without specifying whether of air or prism, for the two are nearly 

 equal, so that we have not data to decide. It appears therefore that 

 the small variation for temperature, previously known to exist, forms 

 the leading variation observed ; and the variation due to any other 

 cause must be sought for in the residue left on eliminating this. 

 Hence accuracy in the temperature-correction applied is of much 

 importance. 



" To what degree of accuracy then can we trust the temperature cor- 

 rection? To form a notion of this, I took* the mean temperatures and 

 mean readings for the five groups of Kew observations, and the mean 

 of the means. Taking the mean of the mean readings to correspond 

 with the mean of the mean temperatures, and applying Mr. Stewart's 

 temperature-correction (Kew observations), namely, -f 1*44 reading for 

 + 30° temperature, I calculated the reading for each of the five groups, 

 subtracted the results from the observed readings, and regarded the 

 differences as errors. The mean error is "19; and the difference be- 

 tween two quantities subject both to errors having a mean value of 

 •19 would be giv^en to within a mean error of -27, which being for 25° 

 would correspond to "32 for 30°. No doubt the mean of several 

 comparisons would come closer than this, but "20 for 30° or '13 for 20° 

 may be taken to be a very probable uncertainty. 



" Mr. Stewart's corrected numbers are plotted in fig. 2. The curve 

 pretty plainly exhibits two features, (1) a general descent, (2) a con- 

 cavity turned upwards. 



" The general descent Mr. Stewart attributes, most probably correctly, 

 to a progressive change in the instrument. On the assumption that the 

 progressive change is uniform, the readings would be represented, so 



Mean 

 temperature. 



Mean reading. 



Calculated 

 reading. 



Obs"^ — calci. 



49-8 



0-71 



0-98 



-•27 



75-5 



2-01 



2-21 



-•20 



51-5 



113 



105 



+ •08 



76-7 



2-48 



2-25 



+ •23 



53-6 



1-31 



M6 



+ •15 



61-4 1-53 

 Mean of means. 





•19 



Mean error. 



