seems, however,
to me to summarise _all_ that is of the slightest value in Romanes'
wordy paper. I have asked Newton (to whom I had lent it) to forward to
you at Birmingham a proof of my paper in the _Fortnightly_, and I shall
be much obliged if you will read it carefully, and, if you can, "hold a
brief" for me at the British Association in this matter. You will see
that a considerable part of my paper is devoted to a demonstration of
the fallacy of that part of "Romanes" which declares species to be
distinguished generally by useless characters, and also that
"simultaneous variations" do not usually occur.

On the question of sterility, which, as you well observe, is the core of
the question, I think I show that it could not work in the way Romanes
puts it. The objection to Belt's and your view is, also, that it would
not work unless the "sterility variation" was correlated with the
"useful variation." You assume, I think, this correlation, when you
speak of two of your varieties, B. and K., being _less fertile with the
parent form_. Without correlation they could not be so, only some few of
them. Romanes always speaks of his physiological variations as being
independent, "primary," in which case, as I show, they could hardly ever
survive. At the end of my paper I show a correlation which is probably
general and sufficient.

In criticising Romanes, however, at the British Association, I want to
call your special attention to a point I have hardly made clear enough
in my paper. Romanes always speaks of the "physiological variety" as if
it were like any other _simple_ variety, and could as easily (he says
more easily) be increased. Whereas it is really complex, requiring a
remarkable correlation between different sets of individuals which he
never recognises. To illustrate what I mean, let me suppose a case. Let
there occur in a species three individual physiological varieties--A, B
and C--each being infertile with the bulk of the species, but quite
fertile with some small part of it. Let A, for example, be fertile with
X, Y and Z. Now I maintain it to be in the highest degree improbable
that B, a quite distinct individual, with distinct parents originating
in a distinct locality, and perhaps with a very different constitution,
merely because it also is sterile with the bulk of the species, should
be fertile with the very same individuals, X, Y, Z, that A is fertile
with. It seems to me to be at least 100 to 1 that it