f 



On the Horary Oscillations of the Barometer at Rome, 219 



might be at different times, that the most erroneous results 

 would have been obtained. For instance, suppose the mean of 

 six observations at nine o'clock to be 30.000 on days when no 

 observations at ten o'clock had been made; and again, that 

 after a month I find the mean of six observations at ten to be 

 30.500, I would note half an inch for the horary variation, be- 

 cause the radical height was less at the former than the latter 

 period. I therefore resolved to take the daily difference from 

 some standard point, which, though liable in some degree to the 

 same objection from the inequable fluctuations of the barome- 

 ter, yet I expected that the combination of a sufficient number 

 of observations would neutralize these irregularities. I resolved 

 to employ 10 m. as the standard, the instrument being most 

 regularly observed at that hour. I have since found that Cap- 

 tain Sabine uses the maximum for that purpose, but I rather 

 prefer my own scheme, as it may often happen that the maxi- 

 mum of the day is not observed. Such is the ground-work of 

 my third table. Part first contains the differences of each hour 

 from 10 M., arranged in columns from 7 M. to 12 a. or eighteen 

 hoursoutof the twenty-four, in which the observations have been 

 most frequently made. As to part second, it is merely to give a 

 wide guess of the state of the barometer during the night, as it 

 only contains three sets of observations, and these not the most 

 favourable. The first supplement contains the most useful ob- 

 servations selected from my ordinary diary in January, to find 

 the difference between 10 m. and 10 a., since the reductions of 

 the second supplement depends in a great degree upon that 

 point, as it contains observations on those days on which 10 m.. 

 has not been observed and 10 a. is therefore substituted. 



