fi DETERMINATION OF THE TEMPERATURE CO-EFFICIENT 



Observations between February and June, 1843, 2.50 scale divisions 



" July and December, 1843, 2.28 " " 



" January and June, 1844, 1.94 " 



July and December, 1844, 2.00 " 



for 1 Fahr. It may also be stated that no reasonable supposition in regard to 

 differences of temperature between the indications of the thermometer and mag- 

 netic bar, or to changes in the co-efficient varying with the temperature, will explain 

 all the cases of discrepancies. In these comparisons, always near each other in 

 time, small differences in intensity, as shown by the subsidiary instrument, were 

 allowed for, but the corrections for temperature of this latter instrument were 

 neglected, as the changes of temperature in the building where it was placed were 

 small. 



Another method, not quite so unobjectionable as the preceding one, was tried ; 

 it consisted in taking the results corresponding to the highest temperatures during 

 each winter, and comparing them with those corresponding to the lowest tempera- 

 tures, a correction being made to reduce the changes of force by means of the 

 secondary instrument. These comparisons were liable to be affected by the unequal 

 distribution of the results used over the different parts of the month. The result 

 was : for combinations and comparisons, from 



January, 1844, to June, 1844, 2.03 

 July, 1844, to December, 1844, 2.29 

 scale divisions for each degree of Fahrenheit's scale. 



The mean value of all the results obtained by the various processes explained, is 

 2.6 scale divisions, and as a preliminary measure, it was supposed that the co-effi- 

 cient was changeable, and hence a correction for change of temperature was applied, 

 varying from 3.2 scale divisions, in 1840, to 2.0 scale divisions, in 1844. 



On resuming the discussion it was thought desirable to deduce a value for this 

 co-efficient directly from the entire mass of observations, as this could not fail to 

 satisfy the whole series. For this purpose it was indispensable to make the series 

 of observations continuous, or, in other words, to refer the readings, extending over 

 five consecutive years, to the same initial division of the scale. This is, therefore, 

 a proper place for stating all cases when the instrument suffered any disturbance 

 and the amount of scale correction required. All necessary explanations are given 

 in the record. 



The first break in the series occurred August 27, 1840, at 12 h 22 m (Philadelphia 

 time), when the mirror was accidentally deranged. The observed numbers from 

 this date to September 22, at 12 h 22 m have been brought to comparison with former 

 numbers by the mean position of the bar for six previous days (in some cases 

 seven) and by the hours, from O h 22 m to 22 h 22 m inclusive. This correction is 

 already applied in the record, its probable error is given as 3.3 scale divisions. 



On September 22, 1840, the instrument was readjusted. 



An interruption of eleven days occurred, in January, 1841, owing to the intro- 

 duction of a reflecting vertical force magnetometer, and requiring a new arrange- 

 ment of the instruments. The horizontal force magnetometer was left in its place. 

 The mean values for January, viz : 944.6 divisions for the bifilar, and 36.5 for the 



