has differed in each case, yet the result is
identical. Therefore the order of exposure has no effect on the
result.

|----------+------------------------------------|
|Composite.|Successive places of the Components.|
| 1 2 | A B | D A | C D | B C |
| 4 3 | D C | C B | B A | A D |
|===============================================|

In 1 it has been A, D, C, B,
" 2 " B, A, D, C,
" 3 " C, B, A, D,
" 4 " D, C, B, A,

I will next show a series consisting of two portraits considerably
unlike to one another, and yet not so very discordant as to refuse
to conform, and of two intermediate composites. In making one of the
composites I gave two-thirds of the total time of exposure to the
first portrait, and one-third to the second portrait. In making the
other composite, I did the converse. It will be seen how good is the
result in both cases, and how the likeness of the longest exposed
portrait always predominates.

The next is a series of four composites. The first consists of 57
hospital patients suffering under one or other of the many forms of
consumption. I may say that, with the aid of Dr. Mahomed, I am
endeavouring to utilise this process to elicit the physiognomy of
disease. The composite I now show is what I call a hotch-pot
composite; its use is to form a standard whence deviations towards
any particular sub-type may be conveniently gauged. It will be
observed that the face is strongly marked, and that it is quite
idealised. I claim for composite portraiture, that it affords a
method of obtaining _pictorial averages_, which effects
simultaneously for every point in a picture what a method of numerical
averages would do for each point in the picture separately. It
gives, in short, the average tint of every unit of area in the
picture, measured from the fiducial lines as co-ordinates. Now every
statistician knows, by experience, that numerical averages usually
begin to agree pretty fairly when we deal with even twenty or thirty
cases. Therefore we should expect to find that any groups of twenty
or thirty men of the same class would yield composites bearing a
considerable likeness to one another. In proof that this is the case,
I exhibit three other composites: the one is made from the first 28
portraits of the 57, the second from the last 27, and the third is
made from 36 portraits taken indiscriminately out of the 57. It will
be observed that