e receiver) will pass a spark that cannot go
across 0.6 of air at one time, it will immediately after, and apparently
under exactly similar circumstances, not pass a spark that can go across
0.8 of air?

1391. It is probable that part of this variation will be traced to
particles of dust in the air drawn into and about the circuit (1568.). I
believe also that part depends upon a variable charged condition of the
surface of the glass vessel _a_. That the whole of the effect is not
traceable to the influence of circumstances in the vessel _a_, may be
deduced from the fact, that when sparks occur between balls in free air
they frequently are not straight, and often pass otherwise than by the
shortest distance. These variations in air itself, and at different parts
of the very same balls, show the presence and influence of circumstances
which are calculated to produce effects of the kind now under
consideration.

1392. When a spark had passed at either interval, then, generally, more
tended to appear at the _same_ interval, as if a preparation had been made
for the passing of the latter sparks. So also on continuing to work the
machine quickly the sparks generally followed at the same place. This
effect is probably due in part to the warmth of the air heated by the
preceding spark, in part to dust, and I suspect in part, to something
unperceived as yet in the circumstances of discharge.

1393. A very remarkable difference, which is _constant_ in its direction,
occurs when the electricity communicated to the balls _s_ and S is changed
from positive to negative, or in the contrary direction. It is that the
range of variation is always greater when the small bulls are positive than
when they are negative. This is exhibited in the following Table, drawn
from the former experiments.

Pos. Neg.
In Air the range was 0.19 0.09
Oxygen 0.19 0.02
Nitrogen 0.18 0.11
Hydrogen 0.14 0.05
Carbonic acid 0.16 0.02
Olefiant gas 0.22 0.08
Coal gas 0.24 0.12
Muriatic acid 0.43 0.08

I have no doubt these numbers require considerable correction, but the
general result is striking, and the differences in several cases very
great.

* * * * *

1394. Though, in consequence of the variation of the striking distance
(138