MEASUREMENT OF HEIGHTS. 543 



that the temperature of the air observed on a plain or on a height 

 is always affected by the temperature of the surface of the earth. 

 Hence we see, were it from this cause only, how little fitted ba- 

 rometrical measurements of height are to determine questions, 

 the answers to w'hich depend on small differences between theory 

 and experiment. Possibly obsei-vations made late at night might 

 agree better together than those made in the day when the 

 surface of the earth is heated by the sun. 



9. 

 It is known that Gay-Lussac found the value here denoted 

 by k = 0'00375, by experiments agreeing almost perfectly with 

 each other; and that Dalton found exactly the same result 

 from his experiments. The object of both these great phy- 

 sicists was to determine directly the increase which an unit of 

 volume of dry air undergoes, when, the pressure remaining equal, 

 the temperature is increased from freezing to boiling water. The 

 accordance, not only of the several experiments in each series, 

 but also of the results of the two series, has caused the determi- 

 nation of ^ = 0*00375 to be generally regarded as one of the most 

 certain that we possess : and there would be no reason for doubt 

 respecting it at this period*, had it not been for recent experi- 

 ments of Rudberg's, distinguished by the great care with which 

 they were conducted, particularly in drying the air employed, 

 and which give a considerably smaller value for A:, i.e. 0*003648. 

 Any later determination, contradicting an older one which has 

 become in a degree classic by its intrinsic weight and by 

 its general acceptance and use, ought to be accompanied by 

 a strict examination of the older determination ; and it is only 

 when such criticism shows grounds for distrusting the older, 

 that the more recent should be deemed deserving of preference. 

 Rudberg has not entered into such a criticism. As the dif- 

 ference between the two values of k cannot be explained by 

 the accidental errors of the experiments, as is shown by the 



* I have myself determined, from my own observation, the value to be em- 

 ployed instead of Ic in computing astronomical refractions, and have found it 

 0-0036438 ; but this value must be different from that of k, and ought to be 

 less, as shown in the 7th part of my observations, page xi. The research might 

 have been spared had I possessed observations of the quantity of aqueous va- 

 pour in the air at the time of each observed refraction. It remedied the diffi- 

 culty as far as could be done in the absence of a knowledge of the actual acci- 

 dental state of the atmosphere on each occasion. But it is to be regarded as a 

 contribution to the knowledge of astronomical refraction, and not as a determi- 

 nation of the value of k. 



