434 PROCEEDINGS OF THE AMERICAN ACADEMY. 



whole length.* To prevent the introduction of an error, on the other 

 hand, from refraction in the glass of the jacket as well as of the ther- 

 mometer, readings were made from various sides, and the average taken ; 

 this error of refraction was of barely perceptible magnitude, and was 

 surely corrected by the diversity of our readings. Great care was taken 

 to place the thermometer in a vertical position. 



In all, five thermometers were used. The oldest Baudin thermometer 

 No. 9389 and the Tonnelot No. 11143 have been described in the 

 former paper. The latter was standardized in June, 1894. The new 

 Baudin thermometers were especially constructed for this transition- 

 temperature work, each one including the temperature 32°. 38. They 

 are numbered 15200, 15275, and 15276, and were standardized in 

 June, 1900. Their bulbs are no larger in diameter than the stem. 



No. 15200 has no enlargement of the capillary throughout its entire 

 length, and the scale runs from —5° to 104°. Each degree is about 

 six millimeters in length. 



No. 15275 has an enlargement between +2° and 30°, being intended 

 to give readings from —11° to +2° and from 30° to 112°. Its degrees 

 are about as long as those of No. 15200. After its use in the present 

 research, it will serve as a means of verifying thermometers to be used 

 for determining the freezing and boiling points of solutions. 



No. 15276, lastly, has one enlargement between 38° and 65°, another 

 between 67° and 98° ; it is intended to give very accurate readings 

 between— 2 and 38°, its degrees being each eight millimeters long. 



On account of the diversity in the forms of these thermometers, each 

 had to be calibrated by the Bureau in a different fashion, and especial 

 confidence may therefore be placed in the average of their readings. 



In the first place it seemed desirable to determine if sodic sulphate 

 made in various ways would yield the same value for the transition 

 point. 



The first sample was made from commercially "pure " Glauber's salt. 

 A filtered solution was allowed to crystallize ; this recrystallization was 

 found to raise the transition temperature only a few hundredths of a 

 degree. After five recrystallizations, readings of its transition tempera- 

 ture were taken on the old sensitive but only approximately standardized 

 Baudin thermometer, No. 9389, which alone was used in the preliminary 

 comparisons of the various preparations. 



* After tlie experiments were finished, it was found that in spite of the rapid 

 circulation of water tlirough the jacket, its temperature was about 0°.4 lower than 

 that in the thermostat. Hence a correction of 0°.002 must be added to the final 

 result. 



