128 Proceedings of Royal Society of Edinburgh. 
back to its first stage of positive twist, there is a hint at a change in 
the law of the acquired polarity. It is obvious, however, that this 
is simply the result of torsional after-effect. The wire, in fact, has 
become elastically untwisted, although it has not become so as 
regards the relative positions of its end-sections. On continuing 
the untwisting past the original zero, we find that it is some time 
before a pronounced effect is produced. Even for the twist — 37'*5 
the difference in the average polarities for the two directions of 
current attains to nothing like the first difference for the twist of 
+ 37'*5. This comparative vanishing away of the polarity differ- 
ence effect as the wire is partially untwisted is quite analogous to 
what is observed when, with no sustained longitudinal magnetising 
force, the current along a wire is reversed at different stages of 
untwisting. 
There is one particular, however, in which the results of the 
present experiments differ from what might be expected if the 
accumulation of polarity was simply due to the magnetic effect of 
the current along the twisted wire. 
First, it should be noted that the twist in a wire subjected only 
to the magnetising influence of a longitudinal field varied cyclically 
has no determinate effect in causing a change in the average 
polarity. Whatever be the polarity produced by a current passing 
along the twisted wire, we might naturally expect a cyclic field 
superposed thereon to give rise to a symmetrical variation of 
magnetic intensity about this acquired polarity as a mean. Con- 
sequently the difference of the average polarities for positive and 
negative currents should be equal to the range of polarity when the 
current is reversed in the twisted wire. As a fact, however, it is 
much greater. For instance, in the case given above for twist + 25' 
the range of polarity produced by reversing the current, the field 
being zero, was only 25 ; whereas the difference of polarities as 
given in the table is 42. Similarly, for twist + 37'*5, the range 
was 49, as compared with the difference 82 5; for the (second) 
twist 12'"5 the range was zero ; and for twist - 37'*5 the range was 
6, as compared with the difference 15*5. 
The same peculiarity is shown to a more pronounced extent in 
nickel wire, of which I give here only one experiment. In spite of 
the greatest care in setting up the nickel wire, a change of average 
