grammarians in their warnings and the Romance languages bear evidence to
this effect.

We noticed above that the final consonant was exposed to danger by the
fact that the syllable containing it was never protected by the accent. It
is also true that there was a tendency to do away with any difficult
combination of consonants. We recall in English the current
pronunciations, "February," and "Calwell" for Caldwell. The average Roman
in the same way was inclined to follow the line of least resistance.
Sometimes, as in the two English examples just given, he avoided a
difficult combination of consonants by dropping one of them. This method
he followed in saying santus for sanctus, and scriserunt for scripserunt,
just as in vulgar English one now and then hears "slep" and "kep" for the
more difficult "slept" and "kept." Sometimes he lightened the
pronunciation by metathesis, as he did when he pronounced interpretor as
interpertor. A third device was to insert a vowel, as illiterate
English-speaking people do in the pronunciations "ellum" and "Henery." In
this way, for instance, the Roman avoided the difficult combinations -mn-
and -chn- by saying mina and techina for the historically correct mna and
techna. Another method of surmounting the difficulty was to assimilate one
of the two consonants to the other. This is a favorite practice of the
shop-girl, over which the newspapers make merry in their phonetical
reproductions of supposed conversations heard from behind the counter.
Adopting the same easy way of speaking, the uneducated Roman sometimes
said isse for ipse, and scritus for scriptus. To pass to another point of
difference, the laws determining the incidence of the accent were very
firmly established in literary Latin. The accent must fall on the penult,
if it was long, otherwise on the antepenult of the word. But in popular
Latin there were certain classes of words in whose case these principles
were not observed.