266 I'HILOSOPHICAL TRANSACTIONS. [aNNO 177". 



pansion of the glass tube and the scale. But this error is in almost all cases so 

 small as to be not worth regarding ; we have however, in the note below, given 

 a rule for computing the value of it.* 



In making experiments with thermometers, it evidently is equally necessary 

 that the quicksilver in the tube should be of the same heat as that in the ball, as 

 it is in adjusting the boiling point : for this reason, in trying the heat of liquors 

 much hotter or colder than the air, the thermometer ought, if possible, to be 

 immersed as far as to the top of the column of quicksilver in the tube. As this, 

 however would often be very difficult to execute, the observer will frequently be 

 obliged to content himself with immersing it to a much less depth. But then 

 as the quicksilver, in a great part of the tube, will be of a different heat from 

 that in the ball, it will be necessary, where any degree of accuracy is required, to 

 make a correction, on that account, to the heat shown by the thermometer. If 

 the heat of the quicksilver in the tube be known, the correction may readily be 

 made by help of the annexed table ; the only difficulty lies in estimating what 

 that heat may be. In all probability the heat of the quicksilver in the tube will 

 not be very different from that of the air which surrounds it •,'f- but as that air 



* The usual way of adjusting thermometers is, to mark the boiling and freezing points on the 

 glass tube, and not to set off those points on the scale till some time after, when the tube and scale 

 may both be supposed to be nearly of the temper of the air in the room ; consequently, when the 

 thermometer is exposed to a greater heat than that, the scale, if of brass, will expand more than the 

 glass tube, and the divisions on it will be longer than they ought to be ; but, if tlie scale be of wood, 

 it will expand less than the glass tube, and the divisions will be too short. Let now the heat of the 

 air, when the divisions were set off on the scale, be called a ; let the degree of heat which the tlier- 

 niometer stands at in the experiment be called d ; and let the degree answering to that point of the 

 scale in which the thermometer is fastened to the scale be called f. Then, if all parts of the ther- 

 mometer and scale are heated equally, and the scale is of brass, the thermometer will appear to stand 



lower than it ought to do by the part of a degree, observing, that ifo— f x d — a 



is negative, it wUl stand higher than it ought to do; but if the scale is of wood, it will stand higher 



than it ought to do by the z.—- part of a degree. If tlie tliermometer be fastened to the 



scale by the ball, or any part of the tube lower tlian tlie observed heat, the error w ill be the same, 

 whether that part of the tube and scale, which is above the obsen-ed degree, be of the same heat as 

 the ball or not : but if the thermometer is fastened to die scale by the top of the tube, as is fre- 

 quently done, then the error will vanish whenever tliat part of the tube and scale, which is above 

 the observed degree, is not much heated. This nile is founded on Mr. Smeaton's experiments, who 

 found, that white glass expands -j^ of an inch in a foot by 1 >S0° of heat ; that brass wire expands 

 j%^ff^ ; and that wood expands scarcely sensibly. — Orig. 



f This must evidently be the case, unless tlie quicksilver in tlie tube is considerably heated by its 

 contact with that in the ball. To see whether this was the case, some sand was iieated in a small 

 copper dish, over a lamp, to the heat of about '..'12°, and die abovementioned tube, widiout a ball, 

 laid horizontal with the end extending about half an inch over the sand ; but to prevent its being 

 heated by it, a piece of wood, about a quarter of an inch thick, was laid between the sand and it. 

 After it had remauied a sufficient time in diis situation, die division which die quicksilver stood at was 



