Report of the Gravity Stirvcy Committee. 219 



(d) However carefully we construct our pendulums we shall 

 always have to correct the observed vibration number for 

 temperature and air pressure, and the corrections are by no 

 means small. Temperature changes affect both the dimensions 

 of the pendulum and the general properties of the medium in 

 which it swings, while pressure affects only the second of these. 

 We therefore require to determine both these quantities with 

 considerable accuracy ; for pressure this is easy enough, a good 

 syphon barometer gauge being all that is wanted ; but the deter- 

 mination of the temperature is not so simple, and observers are 

 by no means agreed as to the best method of attaining it. The 

 general assumption appears to be that a thermonieter will follow 

 the changes of temperature of the air moi^e quickly than will the 

 pendulum. Accordingly many observers, including Sabine and 

 ]\Iendenhall, sink the thermometer bulbs in a metal bar of the 

 same thickness as the pendulum rod, while von Sterneck encloses 

 the whole thermometer in a wide glass tube. The latter plan is 

 almost certainly bad; for the heat has first to make its way through 

 a glass tube, then across a layer of air, and then to heat up the 

 thermometer. But is Sabine's plan very much better? If we 

 consider the structure of a thermometer bulb, viz.: a thin layer of 

 glass, which is notoriously a bad conductor of heat, and then a 

 cylindrical mass of mercury about as thick as, and a worse 

 conductor than, the pendulum rod, and if we bear in mind that 

 convection currents probably play only a secondary part in 

 equalising the temperature of different parts of the thermometer 

 bulb, and further that all delicate thermometers are sluggish in 

 their indications, I think we shall see that an unprotected 

 thern:!ometer with a tolerably large bulb, set as near as possible 

 to the pendulum, will probably lag in temperature behind the air 

 of the containing vessel by about the same amount as the pendu- 

 lum itself, and in consequence the thermometer is more likely to 

 give the actual temperature of the pendulum if arranged in this 

 way than if sunk in a metal bar or otherwise modified. In any 

 case delicate thermometex^s ai'e required ; they should register at 

 least to one-twentieth of a degree Fahrenheit, preferably to one- 

 fiftieth of a degree Centigrade. 



(e) But little need be said as to the containing apparatus. It 

 should certainly be of juetal, in order as far as possible to secure 



