at Leith Fort every Hour of the Day in 1824 and 1825. 383 



ter gives, by applying, according to its sign, the correction oppo- 

 site to the given hour. In place of taking the mean of the two 

 years, it might be preferable to take the results for 1824 in cold 

 years, and those for 1825 in warm years. 



Before concluding this part of the subject, it may be inte- 

 resting to ascertain, from the preceding results, the relation be- 

 tween the mean temperature of the day, and the results obtain- 

 ed from the hours used at Paris, Halle, and Abo, where the ther- 

 mometer is observed more than three times : 



Hours used : 



Deviation from the 

 Morning. Noon. Afternoon. Evening. Mean Temp. 



Paris*, - 9 12 3 9 +1.282 



Halle, 8 12 2 6 10 +1.103 



Abo, aW . 8 11 2 510+ 1.053 



These deviations are very great, and shew how little is gain- 

 ed by multiplying observations, as in the preceding journals. 

 Any two pair of similar hours would have given deviations less 

 than one-third of those in the preceding table. Indeed, it is ob- 

 vious, that any number of observations made during the day, can 

 never give a correct mean, without corresponding observations 

 made late at night and early in the morning. In the register at 

 Paris, Halle, and Abo, it would require the addition of the mi- 

 nimum to obtain the proper mean, as appears from the following 

 Table : 



* The hours of 9 and 9 being nearly those that give the mean temperature of 

 the day, it is obvious, that the mean of 12 and 3, must give a result considerably 

 above the mean temperature of the day, and consequently, that the mean of ail 

 the^/owr observations must err considerably in excess. 



VOL. X. P. II. 3 c 



