1884.] of the Mean Density of the Earth. 157 



temperature. I was astonished to find in his numbers most decided 

 signs of some temperature effect, and although I have not been 

 able to discover the cause of this effect, yet it may be important to 

 put the facts on record. The description of the apparatus, calcula- 

 tions, and observations form Vol. xiv. of the Memoirs of the Royal 

 Astronomical Society, which also contains the mathematical theory 

 as developed by Mr (now Sir George) Airy. His observations 

 extended over sixteen months from January 24, 1841, to May 8, 

 1842, and amounted in all to about 2000 separate sets of observa- 

 tions. They were therefore made at all temperatures, varying from 

 a winter temperature of 30° F. to a summer temperature of 69° F. 

 They were made by employing balls of various materials for the 

 torsion balance — lead, platinum, zinc, brass, hollow brass, glass and 

 ivory — of different sizes, and suspended in different ways, — bifilar 

 lines of silk and metal at different distances, and single wires of 

 different metals and diameters. These were all arranged in sixty- 

 two different sets. To discover any temperature effects, it would 

 have been better had the experiments been carried out on one 

 uniform system; but as it is, those averages will give the most 

 reliability for comparison, which are based on the largest number 

 of different series, as thereby we eliminate effects due to particular 

 series. 



The method adopted in forming the table below has been first 

 to go through Baily's Table I. and make lists of those days when 

 the temperature lay within certain limits, thus for instance two 

 degrees above or below 50°. On any particular day the tempera- 

 ture in the case, which surrounded the box in which the torsion 

 rod was suspended, remained remarkably constant, only as a rule 

 varying a fraction of a degree. Within the torsion box it would 

 therefore be almost rigorously uniform. By means of the preliminary 

 lists, the deduced daily means for those chosen days were then ex- 

 tracted from his Table II. and entered in lists by themselves. 

 Then the means of the daily means in each series were taken, and 

 finally the means of these last taken. In this way, irregularities 

 due to special days and to different series were successively 

 smoothed down as far as possible. The results are given in the 

 table below. I have not discussed the whole of the observations, 

 but only those near the temperatures marked. For the lowest 

 temperatures of all, from 30° to 37° I have not taken account of 

 his second series of experiments. My reason for this is that the 

 method by which the observations were made was different from 

 all the others, and a bad one. Also that out of 12 means for these 

 temperatures, 5 would be from this series, which happened to be 

 made in cold weather. The consequence of admitting them would 

 therefore be to mask effects due to temperature by the errors of 

 observation of this particular series. The densities obtained by it 



