80 Proceedings of Royal Society of Edinburgh, [jan. 17, 
shown. The readings of the open Stevenson screen are used as a 
base line, and the difference between its readings and those of the 
closed Stevenson and the screen C are marked off, each series being 
connected by straight lines. It will be seen that on only two 
occasions did the screen C give the highest readings, and on these 
exceptional occasions it was only 0°T higher. As the days on which 
these occurred were dull and windy, the differences were probably 
errors of observation, such as are quite to be expected. It will 
be observed that the close screen generally held an intermediate 
position between the open one and screen C, its readings being 
better than the former, but not so good as the latter. 
Accepting screen C as our standard, which we may do for the 
present, its readings being the same as the fine silvered bulb, an 
examination of the maximum temperature curves shows us that in 
autumn and winter the Stevenson screen is frequently more than 1° 
too high. In the first half of October it was more than 2 o, 0 too 
high on two occasions, and in November it was a degree or more 
wrong eight times — in addition to the seven times shown on plate 
there was an error of 1°T on the last day of the month — and on 
one occasion in that month it was as much as 1°*75 too high. The 
differences shown in PI. IV. are the differences in the maximum 
readings, but these are not necessarily the greatest differences for the 
day. Of course, the differences in the maximum readings are what 
are practically required, yet it may be interesting to note that this 
may not be the greatest difference for the day. For instance, on the 
21st November the maximum readings for the day were 39° ‘6 for 
screen C, and 41 0, 2 for the open Stevenson, thus giving a difference 
in the maximum for the day of l° - 6 as shown. The day being 
very trying, a reading was taken at 1 o’clock, when the index of the 
maximum thermometer in screen C was at 38° and in the Stevenson 
at 40 o, 6, or a difference of 2 0, 6. This great difference was due to 
there being a dead calm a short time before. 
The weather on the different days is not entered on the plate, as 
it is quite unnecessary ; the curves speak for themselves. Whenever 
the reading of the screen C was much below the others, the weather 
was fine, and it was only during cloudy and stormy weather 
that all three agreed. 
The observations for December show that the maximum given 
