678 
Proceedings of the Poyal Society 
and whenever the temperature was at all steady, readings were taken 
of the thermometers in the different screens. 
The result of these tests has been satisfactory. The thermometer 
in none of the screens rose much above the standard, even during 
the most trying conditions of the experiments. The thermometer 
in screen fig. 3, under many conditions read almost the same as the 
standard; but on one or two trying occasions, when the wind died 
away and the sun shone brightly, the error was quite decided, and 
amounted to an average of a little over half a degree. From my 
notes on these occasions, I see that it read sometimes exactly the 
same as the standard, but at other times it was one degree too high; 
generally, however, the difference was not so great, and gave an 
average of 0°‘6. It may be objected that we cannot average errors 
of this kind, because the maximum error may be the true error of 
the instrument, as the maximum temperature to be registered might 
happen just when the error was greatest. From the manner in 
which the experiments were made, it is, however, I think the 
average, and not the maximmm that is the true error of the screen. 
The occasions on which the error was greatest was when the 
temperature was changing. The readings were taken while the tem- 
perature was falling ; and as the fan thermometer follows the changes 
more quickly than the screen, it had fallen more than half a degree 
before almost any change was effected in the screen thermometer ; 
so that when the readings were taken the screen thermometer was 
still recording the higher temperature to which it was previously 
exposed. These tests do not show that the maximum of screen 
was ever one degree higher than the maximum of the standard. A 
more correct method of testing would have been to employ maximum 
registering thermometers, and to take the maximum temperature 
indicated during a certain time. This method was intended to have 
been followed, and special registering thermometers were being pre- 
pared, but they were not ready in time for the experiments, which 
had to be made before the hot weather was over for the season. 
The screen shown in fig. 4 has almost exactly the same error as 
screen fig. 3. Though its indications are nearly as true as those of 
the other, yet it does not seem so susceptible of improvement. It 
may be possible to reduce its error by changing the form of the 
screen at the bottom of the draught tube. It can, however, scarcely 
