444 



No. 



Anderson. 



Kamtz. 



Lubbock. 



1 



•0122 



•0108 



•0110 



2 



•0116 



•0107 



•0109 



3 



•0118 



•0107 



•0110 



4 



•0115 



•0104 



•0107 



5 



•0108 



•0104 



•0107. 



6 



•0118 



•0107 



•0110 



7 



•0112 



•0102 



•0105 



8 



•0110 



•0105 



•0108 



9 



•0113 



•0104 



•0107 



10 



•0114 



•0104 



•0107 



11 



•0113 



•0104 



•0107 



12 



•0112 



•0103 



•0107 



13 



•0117 



•0107 



•0111 



14 



•0113 



0104 



•0107 



15 



•0117 



•0108 



•0110 



16 



•0117 



•0107 



•0110 



17 



•0117 



•0108 



•0111 



18 



•0116 



•0106 



•0109 



19 

 Mean 



•0116 



•0107 



■0109 



= •01150 



•01055 



•01084 



It will be remarked at once, on the inspection of these 

 numbers, that the differences of the corresponding results 

 for the same experiment, as well as those of the means, are 

 considerably greater than those of different results, as cal- 

 culated by the same table: plainly proving that the error 

 due to the imperfection of the tables is greater than the 

 error arising from observation. If we now take the diffe- 

 rences between the mean values of m according to each 

 table, and the final mean already obtained, we find that the 

 error in the value of m deduced from the first table is only 

 + .00010. The same error, in the case of the second table, 

 is — .00085 ; and in that of the third, — .00056. The pro- 

 bable difference, supposing the partial means to be affected 

 only by the errors of observation, is less than .00008. We 

 have reason to conclude, therefore, that the second and 

 third of these tables are not so correct as the first — at least 

 for temperatures corresponding to those of the thermometer 



