440 PROCEEDINGS OF THE AMERICAN ACADEMY. 



and secondly the necessity of knowing the pressure with the greater pre- 

 cision, since the tenth of a millimeter changes the boiling point by 

 0°.004. Hence, without a barometer of the best construction, and many 

 precautions, one can hardly hope to attain a result accurate to within 

 0.01°. 



The causes of error in the three fixed points just mentioned are in- 

 herent in the nature of the equilibria, and to these must be added those 

 dependent upon the method of thermometry. The behavior of the 

 liquid mercury, usually used to indicate temperature, leaves little to be 

 desired ; but as is well known, the glass receptacle is far less satisfactory. 

 Its sluggish expansion and contraction may easily introduce errors of a 

 degree or more at high temperatures, and the effect of this cause of error 

 becomes the more important as the temperature-interval increases. 

 Even at 100° serious mistakes may thus result. Hence it is important 

 to choose as a standard some temperature as near as possible to the tem- 

 perature read. In this respect 32°. 383 is far superior to either of the 

 usually accepted standards, because most temperature measurements lie 

 in the neighborhood of 20°. 



The careful study of the transition temperatures of other salts has been 

 continued, and while it has led to results which may be of subsidiary 

 importance, no salt has been found which seems to be as generally useful 

 as sodic sulphate. Hence we unhesitatingly recommend this substance 

 as providing one of the most convenient means known for attaining a pre- 

 cisely, definite temperature. It has even been used in this Laboratory 

 with success on a large scale in a bath containing over ten kilograms ; 

 the description of this undertaking will be found in the preceding paper. 



The writers are much indebted to the Rumford Fund of the American 

 Academy for pecuniary assistance in this investigation. 



Summary. 



The present paper contains the record of experiments which verify in 

 a satisfactory manner the earlier estimate of the transition temperature 

 of sodic sulphate. Its exact place on the international hydrogen scale is 

 found to be 32°. 383 ± 0.001, by means of four thermometers standard- 

 ized by the Bureau International des Poids et Mesures. It is recom- 

 mended as being at least as trustworthy as the other two fixed points 

 used for defining temperature, and is possibly the most generally useful 

 of the three. 



Cambridge, Mass., 1900-1902. 



