20 ROYAL SOCIETY OF CANADA 



obtainino- by this means a zero reading, or reading on the bridge wire when 

 both ends were at O^ C. Leaving the shorter end in the mixture, the longer 

 end was passed out of the shanty into the river, and any difference in tem- 

 perature, from the mixture of snow and water was indicated by a change 

 in reading on the bridge wire. The thermometer had a scale of 20 cm. 

 to the degree centigrade, with vernier reading by means of lens to 1/100 

 of a milHmetre. One millimetre then corresponded to -005 ° C. It was pos- 

 sible, under suitable laboratory conditions, to measure to the ten-thou- 

 sandth part of a degree. In the present case, however, it was exceedingly 

 difficult to be accurate to more than one-thousandth. The j^eculiar want 

 of proper laboratory conditions in the observation shanty, together with 

 the care that had to be exercised in making the mixtures of snow and 

 water for the steady end of the thermometer, rendered the work much 

 longer and more tedious than would have been the case under a more 

 favourable environment. The mixtures had to be prepared of sufficient, 

 magnitude to avoid errors of conduction, and of sufficient uniformity to 

 avoid errors of temperature throughout their mass. It was possible, 

 however, by taking proper precautions, to prepare mixtures of snow and 

 water that would be, within the limits of these measurements, sufficiently 

 uniform. The same procedure was carefully followed in every case, and was- 

 essentially as follows : To a large cask, provided for the purpose, was 

 added two or three buckets of snow. This was pounded down by means 

 of a flat ended club, and enough river water added to moisten the whole 

 mass. The " slush " thus produced was then further worked down, ex- 

 cess of water poured off and continually beaten until the mass became 

 firm and solid throughout. This was repeated with every successive 

 batch of snow until the cask was nearly full of a firm mass of moist snow. 

 Eiver water was then added in sufficient quantity to permeate the whole 

 mass. This mixture would remain good for several hours, and a much 

 longer time, of course, if the temperature of the shanty was so regulated 

 as to be only a little above 0° C. Where sets of readings were separated 

 by several hours, a fresh mixture had to be prepared in order to secure 

 absolute uniformity of temperature throughout the mass. The snow re- 

 maining in a mixture that had already been prepared, served as a begin- 

 ning for a fresh supply, by pouring off all excess of water and Avorking 

 down until thoroughl}^ firm and compact. The time required to make 

 one of these mixtures of sufficient uniformity to be relied upon to 1/1000 

 of a degree, was from one-half to three-quarters of an hour. 



The following table will show the probable error to be found in care- 

 fully prepared freezing-point mixtures, as determined by a number of 

 zero readings at different times : 



Same Mixture. Different Mixtures. 



+ -00045 ° C. ± -00090 ° C. 



+ -00070 ° C. ± '00020 ° C. 



+ -00020 ° C. + -00005 ° C. 



+ -00025 ° C. -f -00085 ° C. 



