for uneducated speakers will more readily adapt a familiar sound to a
new meaning (as when my gardener called his Pomeranian dog a Panorama)
than take the trouble to observe and preserve the differentiation of a
new sound. There is no rule except that any loss of distinction may be
a first step towards total loss.[13]

[Footnote 13: To give an example of this. In old Greek _we_ and _you_
were [Greek: aemeis] and [Greek: umeis]: and those words became
absolutely homophonous, so that one of them had to go. The first
person naturally held on to its private property, and it invented
_sets_ for outsiders. Now the first step towards this absurdest of
all homophonies, the identity of _meum_ and _tuum_, was no doubt the
modification of the true full _u_ to _ii_. The ultimate convenience
of the result may in itself be applauded; but it is inconceivable
that modern Greek should ever compensate itself for its inevitable
estrangement from its ancient glories.]

It is probable that the working machinery of an average man's brain
sets a practical limit to his convenient workable vocabulary; that is
to say, a man who can easily command the spontaneous use of a certain
number of words cannot much increase it without effort. If that is
so, then, as he learns new words, there will be a tendency, if not a
necessity, for him to lose hold of a corresponding number of his old
words; and the words that will first drop out will be those with which
he had hitherto been uncomfortable; and among those words will be the
words of ambiguous meaning.

[Sidenote: No direct proof]

It is plain that only general considerations can be of value, unless