542 THE TEMPERATURE CORRECTION OE BAROMETERS. 



7° '2, the greatest observed difference was 2°*8 at ten minutes after the change of place ; 

 and for a change of 19° '9, the greatest difference was 8°'l at eleven minutes. 



Every degree Fahr. above the freezing-point corresponds to not quite 0*003 inch of 

 expansion in the mercury column of a barometer, so that a reading taken at 5.11 p.m. 

 under the conditions of the first experiment would introduce an error of fully O'OIO 

 into the barometric pressure, 



Further experiments were made to ascertain the effect of opening the window of a 

 warm room, and lighting a fire in a cold room. It was found that in either case 

 differences of 2° or 3° between the two thermometers could be caused, which would 

 give rise to errors of nearly a hundredth of an inch. The latter case is especially 

 interesting. There must be many instances of the 9 a.m. reading of a barometer being 

 taken in a room in which a fire has been recently lit, and these ought therefore to show 

 the 9 a.m. reading during winter lower relatively than in summer when fires are not 

 used. 



These experiments indicate that to get a satisfactory reading of barometers as at 

 present constructed, it is necessary that during the hour previous they should not be 

 exposed to a change of temperature of more than 3°. If this condition be fulfilled, the 

 temperature of the mercury will approximate so closely to that of the attached ther- 

 mometer, as not to introduce an error greater than the unavoidable errors in setting 

 and reading the instrument. 



