76 Chemical and Physical Notes 



such as to introduce much error into observations of this kind. 

 The temperature of the room, and no doubt that of its walls, 

 was very constant, but of course there was no vacuum. 



The instrumental deformity introduced by the necessity 

 of a stem for the thermometer must always introduce some 

 deviation from the normal rate of cooling, but it is, as thermo- 

 meters are made, not practically of much importance. The 

 disturbing element which takes precedence of all others is 

 the air. 



The conditions in which the experiments quoted in the 

 table were made were as favourable as they could be, and it 

 would not be possible to get air more motionless than it was. 

 But however motionless the mass of the air may be, a thermo- 

 meter, or any other object, suspended in it, and having a higher 

 temperature, must produce convection currents in its im- 

 mediate neighbourhood, which will be the more energetic the 

 greater the difference of temperature. Hence the conditions 

 under which a thermometer cools in air are complex. In the 

 first place it cools by radiation to its surroundings, and, setting 

 aside instrumental imperfections, this takes place indepen- 

 dently, as it would in a vacuum, according to the logarithmic 

 law, losing equal fractions of heat in equal times. In the 

 second place it loses heat by contact with the air, and the 

 rate at which this loss takes place depends on the rate of 

 renewal of successive envelopes of fresh air, and this diminishes 

 as the temperature of the thermometer approaches to that of 

 the air in which it is cooling. This explains why the term of 

 cooling when the thermometer is only one or two degrees 

 warmer than the air is greater than when it is five to fifteen 

 degrees warmer. If differences of temperature amounting to 

 10 or I5C. are used, the terms of cooling found are very 

 concordant. 



In the case detailed in Table IX, p. 74, the observa- 

 tions were made at equal intervals of ten seconds, and the 

 mean logarithmic difference (dl) was found to be 0-0669, an d 

 the logarithm of the fraction remaining after the lapse of the 

 first interval (logX) was 1-9331, whence y^ = 0-8572. Now the 

 fraction f is expressed by the circulating decimal 0*857142, 



