The Rev. Dr. Robinson on the Constant of Refraction. 195 



window of the same tower, till both agree, which generally is the case an hour 

 after sunset. The internal temperature is also in most cases recorded, from a 

 third standard thermometer attached to the telescope near its centre ; but in this 

 observatory it is not to be used in computing refraction. If any error were 

 produced by preferring the external, its amount should be greatest when the 

 difference is greatest, which I do not find to be the case. For instance, among 

 39 refractions of a Cygni, I find, 



9 with I — E from 0° to 3°, mean 2°.37, give diff. from mean — 0".22. 

 10 from 3° to 4° difF., mean 3°.39, give - 0".17 

 10 from 4° to 5° diff., mean 4°.45, give + 0".58 

 10 from 5° to 7°, mean 6°.01, give — 0".21 



In this star, 1° would change the refraction 0".72. 

 Among southern stars, 23 of \ Sagittarii. 



8 from 0° to 3° mean 2°.l6 give — 0".22 

 8 from 3° to 5° mean 3°.78 give — 0".ll 

 7 from 5° to 7° mean 5°.66 give + 0".33 



Here 1" gives a change of 0".65. In these the discordances obviously have no 

 connexion with the state of the internal thermometer ; and the case is the same 

 with other stars. 



The barometer used was, till December 4, 1835, a portable one, by Ramsden. 

 It was then replaced by a standard one of Newman, similar to that described by 

 Mr. Baily in the Philosophical Transactions for 1837, p. 431. Mr. Newman 

 states, that the specific gravity of its mercury is 13.545 at 60°, and that the 

 diameter of its tube is 0'.570. In such a tube the correction for capillary action 

 is nearly insensible ; but it happens to be unnecessary here, for a reason given 

 by Laplace, Conn, des Tems, 1829, but not, that I am aware, noticed in any 

 English work. In barometers like this, the scale is terminated at its lower 

 extremity with a point which is brought into contact with the mercury of the 

 cistern ; but the surface of the latter is also curved, so that the contact, if near 

 the edge, is made at a surface lower than the real zero. K the distance from 

 the edge be properly assumed, this may be made to counteract the depression 

 above : it is rather too great here, giving only 0'.003, but the rest is neutralized 

 by the fact, that the contact (if estimated, as I do it, by the meeting of the point 



2 c2 



