24 PROCEEDINGS OF THE AMERICAN ACADEMY 



our experience indicates that, with an intermediate amount, the loss and 

 gain resulting from the causes we have mentioned are more or less 

 closely balanced. By preserving as nearly as possible identical condi- 

 tions, and taking, for example, to every 2 grammes of antimony 30 

 cubic centimetres of hydrochloric acid and 15 grammes of tartaric 

 acid, it was found possible to obtain results closely agreeing not only 

 with each other, but also with what we finally concluded was the most 

 probable value of the atomic weight of antimony. But, during the 

 investigation by which these facts were developed, we made many 

 determinations with varying proportions both of hydrochloric and tar- 

 taric acid, in which the error arising from an excess of one or the other 

 must have had its fidl effect. All these determinations not obviously 

 defective are recorded below ; and although a closer agreement would 

 appear, were only those determinations selected which were made after 

 the more accurate knowledge was obtained, and therefore under more 

 favorable conditions, yet we feel much greater confidence in the result 

 obtained by taking the mean of all, since in tliis mean the errors must 

 be to a great extent, if not perfectly, balanced. 



We were for some time in doubt in what condition the sulphide of anti- 

 mony ought to be weighed, in order to ol)taiu the most accurate results. 

 Our final judgment was that the errors already referred to would best 

 be balanced, while others would be avoided, by w'eighing the sulphide, 

 after it had been dried, at from 180'^ to 200"^, but before it was actually 

 converted into the gray sulphide. This conversion takes place between 

 210'^ and 220°, varying to that extent in different eases. The change, 

 as we infer, is attended with a sudden evolution of heat, and the action 

 is quite violent. Small particles of the material are frequently projected 

 from the vessel, and we sometimes noticed that the surface of the pla- 

 tinum nacelle became coated with the familiar sublimate of sulphide 

 of antimony. If there was oxichloride in the precipitate, there may be 

 an additional volatilization of chloride of antimony at this time ; but the 

 main loss, as we have constantly observed, takes place before the point of 

 conversion is reached.^ We therefore concluded that more trustworthy 

 results could be deduced from the weight of the red sulphide dried, as we 

 have described, than from that of the gray ; and, as will be seen, this 

 judgment was fully confirmed by subsequent experiments on the haloid 

 compounds. AVe have, however, in all but two instances weighed the 

 sulphide in both conditions, and we give the results of both weighings ; 

 and on comparing these results in determinations 8 to 13 inclusive 

 of the table on pages 36-7, which were made under the nearly identical 

 conditions we have above indicated, it will be seen that the differences 



