A colleague asked why it was that dictionaries seemed to prefer fetuses rather than, say, fœtii, to follow the same rule (she said) as octopus, rather than the “stupid sounding” octopuses.
Well, first of all, the plural of an -us ending is -i, not -ii; the Latinate plural of octopus is octopi, not octopii. Only words that end in -ius pluralize to -ii.
Second, octopi is not really any more correct than octopuses. Octopus was a loan word in Latin and is a loan word in English, and in each case the language has applied its own inflection ending for the plural. The original is Greek octopous (“eight” + “foot”) and the plural of that is octopodes, though those who insist on saying octopodes in English conveniently forget that we don’t say octopous for the singular.
In regard to œ versus e, in many words we have gotten from Latin, the digraph has been simplified in North American English, but that’s hardly the first spelling change ever enacted on Latin loans, and efforts to retain Latin etymology (or resurrect it) have had a lot to do with the poor match between spelling and pronunciation in English. In this case, however, fetus is the more etymologically correct spelling; fœtus is an error – a misconjecture. The original Latin is fetus with a long e.
Anyway, feti is used, but rarely. Fetuses is used commonly because, after all, we’re speaking English, and we more often than not conform loan words to English morphological patterns rather than keeping them in the morphology of the source language. (Quick, what’s the plural of sauna? And why do you say that? Also, why does nobody object that the alcohol and the albatross are redundant, since the al in the source means “the”? Answer: they’re ours now [evil laugh].) I suspect that the fact that feti would sound like “feet eye” has some little something to do with the preference in this case – we don’t always like to confuse ourselves. At any rate, dictionaries document usage. They can have some prescriptive effect, but their main function is to tell people what educated people use a word to mean and how they spell and inflect it. So the usage comes first. Even Noah Webster, when he made a number of spelling reforms in his dictionary, used only spellings that had already been used in real life. (And not all of his changes stuck, either.)
Latinate plurals serve nicely as a sign of desire to sound erudite, and they keep the language nice and difficult the way we like it, but they do have practical limits, beyond which they become rather funny. I seem to recall some humorous prose or verse referring to travelling on omnibi and so forth. (-ibus, by the way, is an inflectional ending of its own and not -ib plus -us, so -ibi is no kind of Latin).