of Edinburgh, Session 1881 - 82 . 
493 
plication, and promises well so far. If it continues to answer our 
expectations I shall take the liberty of laying a detailed description 
of it before the Society. 
There is one point of great interest on which all these experi- 
menters are agreed, viz., that the curve whose abscissae is the spark 
length, and whose ordinate is the spark potential, is not a straight 
line as it ought to be by the ordinary electrostatic theory, combined 
with the notion of dielectric strength above explained, the air 
between the plates being assumed to be a uniform dialectric. There 
can be no question that at very short distances the falling off from 
this law is very great. I think that there has been a slight tendency 
to overstate the discrepancy for distances over '1 cm. Taking 
Bailie’s numbers, excluding all distances under *2 cm., I calculated 
by the method of least squares the formula of the form V' = as + b , 
which would best reproduce his observations. V' denotes calculated 
spark potential for s centimeters ; V the observed spark potential. 
The formula found was 
V' = 99*593s + 4*997 . 
The following Table gives the comparison : — 
s 
y 
V ' 
V - V ' 
•1 
14-67 
14*96 
- *29 
•2 
25*51 
24*92 
+ *59 
•3 
35*35 
34*88 
+ *47 
*4 
44*77 
44*83 
- *06 
*5 
54*47 
54*79 
- *32 
*6 
63*82 
64*75 
- *93 
73*78 
74*71 
- *93 
•8 
84*86 
84*67 
+ *19 
•9 
94*72 
94*63 
+ *09 
1-0 
105*50 
104*59 
+ *91 
The error nowhere exceeds 3 per cent., and if we consider the 
increased probability of error from leakage at the high potentials, 
