14 EXPERIMENTS ON 



tity of refraction, which seems to have affected the 

 observations of the different objects referred to in the 

 tables; and here, it is perhaps worthy of notice, that out 

 of 49 contemporaneous observations, of the top and foot 

 of the flag-staff, at the S. extremity of the line, the 

 refraction attending those of the foot are 36 repeated 

 times less than those of the top ; and that, in the 13 re- 

 maining ones, vi^here the contrary occurs, the excess is 

 seldom above 2' of refraction, and frequentl}'^ below 

 unity*. As this circumstance is in opposition to 

 the general theory, " tliat the lower the object, the 

 greater the refraction," should the same circumstance 

 occur again, in future experiments, it will be worth 

 while to inquire, whether the rays, when passing 

 through the atmosphere below the line of the level, 

 may not be refracted differently from what they 

 are when passing above it. This may perhaps be 

 thought better than a n^ere conjecture, if it be 

 recollected, that Mr. Bouguer, (whilst employed 

 inmeasurino' a des^ree of the meridian in South Ame- 

 rica, and observing on the summit of the Cordeliers) 

 noticed a sudden increase of refraction, when he 

 could view the stars below the line of the level. 



3d. With regard to any attempt towards estimating 

 the effects of terrestrial refraction, by an assumed 

 ratio to the contained arc, as has been hitherto the 

 practice ; without entering into any discussion of the 

 subject, I shall only observe, that if, in the foregoing 

 experiments, we go by the observations taken in the 

 day time, Ave shall have (considering the foot of 

 flag-staff, and preserving the same notation) r— \ - ^ ^ ; 



* The mean difference of refraction, between the top of tlie flag- 

 staff and tlie Miintapum, (on 38 contemporaneous observations) 

 is l6". 44; and that between the top of flag and Muntapum is 

 22". 51, where the order is inverted by 6". 07 : and if tlie top and 

 foot of flag-staff be compared, in an equal and contemporaneous num- 

 ber of observations, the mean of their difference is 6 ". 08 likewise in 

 the inverse order. 



