used together, forming a
rigid cylindrical rod, a modification, as far as yet known, absolutely
peculiar to it.

[Illustration: DENTITION OF THE SABRE-TOOTHED TIGER (MACHAIRODUS).]

In a similar way the _extinct_ machairodus, or sabre-toothed tiger, is
characterized by a more highly differentiated and specially carnivorous
dentition than is shown by any predacious beast of the _present day_. {111}
The specialization is of this kind. The grinding teeth (or molars) of
beasts are divided into premolars and true molars. The premolars are molars
which have deciduous vertical predecessors (or milk teeth), and any which
are in front of such, _i.e._ between such and the canine tooth. The true
molars are those placed behind the molars having deciduous vertical
predecessors. Now, as a dentition becomes more distinctly carnivorous, so
the hindmost molars and the foremost premolars disappear. In the existing
cats this process is carried so far that in the upper jaw only one true
molar is left on each side. In the machairodus there is no upper true molar
at all, while the premolars are reduced to two, there being only these two
teeth above, on each side, behind the canine.

Now, with regard to these instances of early specialization, as also with
regard to the changed estimate of the degrees of affinity between forms, it
is not pretended for a moment that such facts are irreconcilable with
"Natural Selection." Nevertheless, they point in an opposite direction. Of
course not only is it conceivable that certain antique types arrived at a
high degree of specialization and then disappeared; but it is manifest they
did do so. Still the fact of this early degree of excessive specialization
tells to a certain, however small, extent against a progress through
excessively minute steps, whether fortuitous or not; as also does the
distinctness of forms formerly supposed to constitute connecting links.
For, it must not be forgotten, that if species have manifested themselves
generally by gradual and minute modifications, then the absence, not in one
but in _all cases_, of such connecting links, is a phenomenon which remains
to be accounted for.

It appears then that, apart from fortuitous changes, there are certain
difficulties in the way of accepting extremely minute modifications of any
kind, although these difficulties may not be insuperable. Something, at all
events, is to be said in favour of the opinion that sudden and {112}
app