Let, however, a word, instead of really coming under class A, only appear to do so. Let it be dealt with accordingly. The analogy then is a false one. The principle of imitation is a wrong one. The process affecting is a process of confusion.
Examples of this (a few amongst many) are words like songstress, theirs, minded, where the words songstr-, their-, mind-, are dealt with as roots, which they are not.
Ambiguous processes, extraordinary processes, processes of confusion—each, or all of these, are legitimate reasons for calling words irregular. The practice of etymologists will determine what definition is most convenient.
With extraordinary processes we know nothing about the word. With ambiguous processes we are unable to make a choice. With processes of confusion we see the analogy, but, at the same time, see that it is a false one.
[§ 331]. Could.—With all persons who pronounce the l this word is truly irregular. The Anglo-Saxon form is cuðe. The l is inserted by a process of confusion.
Can, cunne, canst, cunnon, cunnan, cuðe, cuðon, cuð—such are the remaining forms in Anglo-Saxon. None of them account for the l. The presence of the l makes the word could irregular. No reference to the allied languages accounts for it.
Notwithstanding this, the presence of the l is accounted for. In would and should the l has a proper
place. It is part of the original words, will and shall. A false analogy looked upon could in the same light. Hence a true irregularity; provided that the L be pronounced.
The L, however, is pronounced by few, and that only in pursuance with the spelling. This reduces the word could to an irregularity, not of language, but only of orthography.
That the mere ejection of the -n in can, and that the mere lengthening of the vowel, are not irregularities, we learn from a knowledge of the processes that convert the Greek ὀδόντος (odontos) into ὀδούς (odows).