68 
Proceedings of Royal Society of Edinburgh, [jan. 17 , 
ing tlie thermometers might he kept constant, and the errors due 
to imperfect mixing he as small as possible. It may he also 
mentioned that the same thermometers were not kept in the same 
screens, but they were changed from time to time, to check any error 
that might arise from any unknown difference in the thermometers, 
and every precaution that could be thought of was taken to 
check the results. 
The maximum thermometers used were all graduated on the stems 
and placed vertically in the screens with their bulbs in the usual 
position. They were held in their places by spring clips, to prevent 
the position of the index being altered by shaking, a source of error 
to which the metal-framed instruments loosely hung are very liable. 
The instruments were carefully tested with their stems vertical. 
The common maximum thermometers with metal frames and 
placed horizontally were discarded, as they were found to give most 
uncertain results, and never agreed with the others in the same screen, 
owing to the index in these instruments moving too easily, and 
more or less easily at different parts of the bore, thus forming a 
longer or shorter air space at different parts of the scale. Further, 
under certain conditions they gave different readings from the ordi- 
nary thermometers with freely exposed bulbs. They read too low 
if the high temperature remained only a short time, owing to their 
greater inertia; and in the open Stevenson screen they read too high 
when radiation remained strong for any length of time, owing to 
their larger surface causing them to he much more heated than the 
thermometers with freely exposed bulbs, as the frame and thermo- 
meter really act as one surface. The maximum thermometers were 
changed from screen to screen, generally every day or second day, 
as an extra check on possible errors. 
A great number of observations were made, extending from the 
15th of September to the end of November. On many days 
when there was much radiation, a great number of readings were 
taken at short intervals, and on as many other days as possible the 
temperature was taken by maximum thermometers. The result has 
been entirely to confirm my first conclusions. The closed screen 
always gave lower readings than the open one. 
The fan apparatus was not put on trial, as there were too many 
readings to he taken without it. In a trial like this the readings of 
