( 355 ) 
expresses perfectly in what manner the effect changes with the 
changing of the stimulus. When examining some of the curves 
added by TicErsreptT to his publication we will observe directly the 
same deviation. In his figures 2, 3, 4, 6, 9, 13 we see that the 
curves, instead of being invariably concave to the z-axis, show in 
the beginning a marked tendency to become slightly convex. I believe 
a good reason to exist for this fact, which, though not expressed 
in my formula, may easily find expression in it by a small alteration. 
I trust to be able to explain this question in a later communication. 
We have now still to examine more closely WALLER’s series. 
WALLER gives with his experiments not only the value of the 
lifting height for each magnitude of stimulation, but adds also the 
magnitude of the negative variation responding to each nerve-sti- 
mulation. In fig. 9 of his publication he presents a continuation of 
the same experiment, in which only the negative variation is recorded, 
and showing that the negative variation increases regularly with 
increasing strength of stimulation. In his figure 8 (l.c.) the graphic 
representation of this fact is given, reproduced by me in fig. 9; 
when stimulation is increased, the negative variation continues as a 
perfectly straight line, at any rate within the limits of the intensities 
used for excitation. 
We may represent this relation by the simple mathematical expres- 
sion y= P+ Qr, y representing the negative variation, x the mag- 
nitude of stimulation and P and QY two constants. Inverting this, 
we may therefore also represent the magnitude of stimulation as a 
function of the negative variation: 
pte Tot, Sy it Balti pity OVSE ten 
wherein L and S are new constants. At any rate we may assume 
here the negative variation to be a reliable indicator of the magnitude 
of stimulation used, and we may therefore, with the aid of the 
formula (10) correct the numbers stated for the value of stimulation 
by means of the numbers found for the negative variation, by 
putting the values 
L= 1,41989 and S= 0,04811 
in that formula. We then obtain for A new values, communicated 
in the next Table X: (p. 356) 
When from these new values for Rand a few new constants 
E is again calculated, it will be seen that the agreement is as fine 
as we can desire, and that the mean error has fallen from 0,331 to 0,109. 
Something more still is proved by these numbers of WALLER. 
In the first place that the magnitude of stimulus possesses another 
thresholdvalue for the negative variation than for the muscle- 
