( 2146 3 
because it appears from the rates during short periods, that abrupt 
changes in the temperature are reflected almost immediately. Not- 
withstanding this, I deemed it possible, at first, that the true expla- 
nation might be found in such a cause, by assuming that part of the 
effect of the temperature on the rate — perhaps by the intervention 
of the elasticity of the suspension-spring, — is only felt after along 
time. In this case however, we ought to find another and smaller 
temperature-coefficient from swift changes in temperature than from 
the comparison of summer- and winterrates. In reality, however, it 
seems, that the coefficients obtained in the two ways agree in the 
main, at least as far as can be judged now, before the completion 
of a more elaborate investigation by Mr. WeepeRr. 
Besides a change in the elasticity of the suspension-spring, lagging 
behind the yearly change of temperature, has become improbable since 
we found no trace of it in the amplitudes of the oscillations. 
Another possible explanation might be found in the hypothesis that 
the temperature of the different parts of the pendulum is permanently 
unequal and that the distribution of temperature varies systematically 
with the season, in such a way that it is not identical in the spring 
and the autumn. The influence of a small inequality of the tempera- 
ture is considerable. For if the temperature of the pendulum-rod changes 
only by so much as 0°.1 R., whereas that of the mercury remains 
constant, the daily rate changes by 05.065. 
The differences between the readings of the upper and lower 
thermometer in the clock-case must throw light on this distribution 
of the temperature. The information however must be defective 
1st. on account of the small accuracy of the thermometers, 2nd because 
we do not know the relation existing between the temperature of the 
steel and the mercury of the pendulum and that of the surrounding air. 
If we consult the mean values of these differences of tem- 
perature for the three periods, we see that in the two former the 
difference: Upper temperature—Lower temperature has been really 
found + 0°41 R. greater in April and May than in October and 
November. This would produce a difference in the rate agreeing in 
sign with that which is really found. In the Srd period, however, 
spring and autumn agree nearly perfectly. 
It seems to me still very uncertain, therefore, whether the cause 
of the phenomenon in question may be found in this distribution of 
the temperature. The fact that, whereas the clock was in very diffe- 
rent circumstances, the inequality of the rate was very nearly constant 
and also the fact that it seems to have diminished in the second period, 
seem, even a priori, contrary to such a hypothesis. 
