OF ARTS AND SCIENCES. 167 



the open mouth of the overflow tube dipping about six or eight inches 

 deep under concentrated sulphuric acid. This overflow indicated the 

 least chauge of tension of the hydrogen in the apparatus, and would 

 have shown the least leak if it had existed ; but the apparatus as thus 

 constructed remained absolutely tight so long as it was in use. 



With the hydrogen drawn from this apparatus, the first determina- 

 tions were not wholly satisfactory, and the cause of error was traced 

 to the air dissolved in the dilute hydrochloric acid with which the 

 generator was charged. In all the succeeding determinations the 

 greatest pains was taken to remove the last traces of air by boiling 

 the dilute acid, and allowing it to cool in a stream of hydrogen ; and 

 as additional precaution, while the solution was still warm, the gas 

 was exhausted from the containing vessel and pure hydrogen run in, 

 several times in succession, the pure acid being finally conveyed into 

 the generator entirely out of contact with the air. The need of all 

 these precautions will be seen when it is considered how small an ad- 

 mixture of air or of nitrogen will materially influence the weight of the 

 hydrogen. If only one ten-thousandth of the volume of the hydrogen 

 were replaced by air during the process of filling the globe, this would 

 cause an apparent increase of weight in the hydrogen of five tenths of 

 a milligram, and that, other things being equal, would reduce the 

 atomic weight of oxygen two hundredths of a unit. 



The precautions used in filling the globe have already been described 

 in detail, and with hydrogen from the apparatus, constructed and 

 charged as just described, were made the five consecutive determina- 

 tions whose results are given as of the first series in the table on page 

 173. These, and all the determinations given in the table, were made 

 by the writer's pupil and assistant, Mr. Theodore William Richards, 

 to whose experimental skill the success of the investigation is largely 

 due, and without whose assistance the work could not have been com- 

 pleted in the present condition of the writer's sight. The mean of 

 these first five results, as will be noticed, is but little less than that 

 obtained by Dumas, and the probable error, ±0.0048, is considerably 

 less than that of Dumas. In order to understand how this result ap- 

 peared to the writer, it must be remembered that he started with a 

 certain prepossession in favor of the hypothesis of Prout, based on his 

 previous work on antimony ; and, furthermore, that* the effect of the 

 causes of error which had been encouutered and overcome all tended 

 to lower the atomic weight ; and the result obtained was a maximum 

 which had been reached after every known precaution had been taken. 

 But although this maximum was essentially the same as that obtained 



