107 



36°. S6'. 46",52 + ,82 c = 56. 36. 46,71 f- 1,56 c. Tlie 

 co-efficients of c are o+)tained from the tangents of the re- 

 spective zenith distances. 



0" 19 



This equation gives c ^ ^T = r 0",26 



and therefore T~ -- = 67^56'. By which, the mean refrac- 



tion at 45° = 4=i- f 1 ~ ^Jl) = 57",42 



Now from the number of observations used, it cannot be 

 doubled that the above conclusion is free from the errors of 

 observation. The only error by which it can reasonably be 

 supposed affected, is that arising from errors of division. 



It is difficult to state the limit of error from hence arising, 

 but it wdl readily appear that much dependence cannot be 

 had on a correction so small as that which I have deduced. 

 For each star or each co-latitude, 12 points of the circle are 

 used so that the quantity 36°. 3&. 46",52, the mean of the 

 results of the three first stars is affected by the mean error of 

 36* points of divisions of the circle. This mean error must 

 certainly be very small. Yet it is not improbable that it may 

 amount at least to 0",15. , 



The error of the quantity S6°. 36'- 46",7l must be smaller, 

 being only affected by the mean error of 120 points, yet it 

 is not improbable it may amount to 0",04 and so the whole 

 quantity 0",19, the numerator of the value of c, will be ac- 

 counted for, 



Q 2 



