of the Barometer near Edinburgh. 



177 



part of his elaborate paper upon Astronomical Refractions, in- 

 serted in the second volume of the Memoirs of the Astronomical 

 Society. 



14. I assumed the form of the function of the latitude- which 

 expresses the oscillation, to be some unknown power of the co- 

 sine, affected by an unknown coefficient, and constant. These 

 quantities are therefore to be eliminated from three or more 

 equations, derived from observation in different parts of the 

 globe. Let the barometric oscillation be represented by z fl z /f , 

 z,,,, in latitudes 6 fl &, 6 /// respectively : then, denoting by n the 

 unknown power of the cosine, and by a and the other con- 

 stants, we may separate the first of these unknown quantities 

 from three equations of the form z ~ a cos 72 6 + , by putting 



LI j j/i f %i /// cos0. cos t... inn -.1 



them under that of ^_ g " = COS , M)< _COSM/ Whence the va- 

 lue of n may be deduced by approximation. For this purpose, 

 I selected nine values of z from the list of observed oscillations, 

 and, for greater convenience of calculation and tabulation, I shall 

 employ these during the rest of this paper in the form of milli- 

 metres, which in fact, since HUMBOLDT wrote, have become the 

 most common units of measure for these results. These num- 

 bers were not in general single results, but were combined from 

 several in such a way as appeared best to correct accidental irre- 

 gularities. Three observations nearly equatorial were taken, 

 three towards the tropics, and three in high latitudes, as follow : 



