60 Proceedings of Royal Society of Edinburgh, [jan. 17 , 
from observations taken at longer or shorter intervals of time, a 
very little consideration will show us that these curves, as generally 
constructed, are in no sense curves of temperature. 
An examination of the one-minute observations given in fig. 2 
shows us that the curve of daily temperature is much too complicated 
to be capable of being represented in the manner it generally is. A 
few detached observations at hourly intervals tell us really very 
little about the matter, and to attempt to draw a curve from these 
can lead to no good, as the curve appears to give definiteness where 
in fact almost all is unknown ; this is particularly the case when 
there is any radiation. Observations taken at intervals so wide 
apart as one hour really tell us nothing about the state of 
matters in the intervals, and yet by connecting these hourly read- 
ings by means of curves we not only indicate the temperatures in 
the interval, but we may even represent the temperature at the hour 
of observation to be rising or falling when in reality it may be 
doing quite the opposite. 
To illustrate the small value we are entitled to put on curves 
drawn from hourly observations of temperature, the curves in PI. II. 
fig. 3, are drawn from one of the few sets of observations taken 
recently at regular intervals. The curve A is drawn from obser- 
vations taken at five-minute intervals, and it will be seen that 
during the two hours while the observations were made, there 
were great fluctuations in the temperature. An examination of 
fig. 2 will, however, show us that even during five-minute intervals 
considerable changes may have taken place. The curve A, fig. 3, is 
therefore not so variable as the actual state of the air was when 
these five-minute readings were taken. Suppose now that, in- 
stead of taking the temperature every five minutes, we had done it 
at hourly intervals. If, for instance, we had selected, not the 
hour but five minutes to the hour, for the time for taking our 
readings, then we should have got the curve B. If, however, 
we had selected five minutes past the hour for our observations, 
we should have got the curve C, while if the readings had been 
taken five minutes later the curve would have been I). This latter 
curve would have shown a higher maximum for the day of 2° '75 
above that given by curve B. This difference might have been 
brought about, as stated, by the hour at which the observa- 
