24 PROCEEDINGS OF THE AMERICAN ACADEMY 



" D." (I was obliged to omit " I " from consideration, owing to an 

 apparent error. Wien. Ber., Ixxiii. 440.) Tiie separate results are 

 fairly accordant. This line agrees quite closely with my fifth series, 

 ST. The results obtained with a brass capillary are indicated by 

 the line d e, which is a continuation also of *S^, since the brass capil- 

 lary agreed substantially with the others at b. The curvature of this 

 line, as well as of a jS 5 c, is decidedly convex upward, much more so 

 than ^J". Obermayer's further series with capillary "D " is of much 

 value. It is represented in the line gJi, and the three marked points 

 indicate the points yielded by the experiments. This line is also 

 decidedly convex upward, and its curvature seems to be beyond the 

 range of variable erroi-s of measurement. I have connected the in- 

 dividual points with each other by a broken line, because, owing to 

 the small number of points, almost any form of curve could be 

 drawn through them, and to select any one equation would be wholly 

 arbitrary. This remark applies also to the line UV, representing 

 Wiedemann's results. This line is still more convex upward than 

 either of the preceding. The line WX, representing Warburg's re- 

 sults with oscillating plates, is in substantial accord with the results 

 by Obermayer, Wiedemann, and myself, as well as with the mean of 

 Meyer's results ; but as it is derived from observations including but 

 one temperature interval, it cannot serve to determine any beyond the 

 first power variation in the coefficient. 



To show the relation of the several lines to one from an equation 

 of the exponential form which has been so generally adopted, I have 

 drawn the line Z from the equation 



1L = (^1_^ 0.003670)°-^«. 



Vo 



The fifth series of my own results is shown in the line ST, the ex- 

 perimental means being denoted by crosses. The deviation of ^Z 

 from a straight line is quite slight, and I am confident that this is a 

 close approximation to the true result, and that the greater curvature 

 of the lines from my own earlier results and those of other observers 

 is due in part to impurities in the air, either in the form of carbonic 

 acid or of vapor of water. I regard those of my measurements on air 

 which precede the fifth series merely as checks upon the accuracy 

 of the process, and as possessing small weight as compared with that 

 series, for reasons assigned in the following Critique of the Method. 

 The same is true of all but the last series with carbonic acid. 



