ON THE MECHANICAL EQUIVALENT OF HEAT 467 



looking over the observations again, I see that, had I taken account of 

 the small irregularities, it would have made the observations agree more 

 nearly with the other thermometers. Hence the objection vanishes. 

 However, I intend working up some observations which I have with the 

 Kew standard at a higher temperature, and shall publish them at a 

 future time. 



There is one other error that might produce an apparent decrease in 

 the specific heat, and that is the slight decrease in the torsion weight 

 from the beginning to the end of most of the experiments, probably due 

 to the slowing of the engine. By this means the torsion circle might 

 lag behind. I made quite an investigation to see if this source of error 

 existed, and came to the conclusion that it produced no perceptible 

 effect. An examination of the different experiments shows this also, 

 for in some of them the weight increases instead of decreasing. See 

 Tables XXXVII to L. 



The error from the formation of dew might also cause an apparent 

 decrease; but I have convinced myself by experiment, and others can 

 convince themselves from the tables, that this error is also inappre- 

 ciable. 



The observations seem to settle the point with regard to the specific 

 heat at the 4 point within reasonable limits. There does not seem 

 to be a change to any great extent at that point, but the specific heat 

 decreases continuously through that point. It would hardly be possible 

 to arrive at this so accurately as I have done by any method of mixture, 

 for Pfaundler and Platter, who examined this point, could not obtain 

 results within one per cent, while mine show the fact within a fraction 

 of one per cent. 



The point of minimum cannot be said to be known, though I have 

 placed it provisionally between 30 and 35 C., but it may vary much 

 from that. 



The method of obtaining the specific heat of the calorimeter seems 

 to be good. The use of solder introduces an uncertainty, but it is too 

 small to affect the result appreciably. The different determinations of 

 the specific heat of the calorimeter do not agree so well as they might, 

 but the error in the equivalent resulting from this error is very small, 

 and, besides, the mean result agrees well with the calculated result. It 

 may be regarded as satisfactory. 



The apparatus for determining the equivalent could scarcely be im- 

 proved much, although perhaps the record of the torsion might be made 

 automatic and continuous. The experiment, however, might be im- 



