I believe for most of the Normanized loans, the antepenult stress is due to a shift from older final stress (that I think can be seen in Chaucer etc.) by means of alternating strong-weak syllables.Ephraim wrote: The normanized loans are basically treated as if they were borrowed via French (and they often are). The base may be very latinate (in spelling at least) but the endings are greatly reduced, so that -itās becomes -ity, -tia and -cia becomes -cy or -ce, -us becomes -0 etc. These don't follow Latin stress rules, and like other Anglo-Norman words, they don't follow French stress rules either. I'm having trouble formulating rules for how these words are stressed, but somehow English speakers seem to perceive the stress as more natural. There is certainly a tendency towards initial stress (I think it might be especially strong in nouns) which would make them closer to the native vocabulary. But there are many exceptions and the exceptions often don't agree with Latin or French stress either. I think analogy plays a role.
Examples include: nominative (< nōminātī́vus), accusative (accūsātī́vus), Anthony (< Antṓnius), vocabulary (< vocābulā́rium), dictionary (< dictiōnā́rium), theory (< theṓria < θεωρία), history (< história < ἱστορία)
Heresy is a normanized loan, while diæresis and synæresis are direct loans. For heresy (< hǽresis < αἵρεσις), English stress actually agrees with Latin which also happens to agree with Greek. But this is probably just a coincidence. More importantly, heresy is stressed on the initial syllable.
Really, it's the normanized loans that are truly weird. Why is nominative stressed on the initial syllable but accusative on the second? Neither agrees with Latin.
So "charitee," etc., stressed on the last syllable, gets a secondary stress on the third-to-last syllable. Then that secondary stress becomes the primary stress. This kind of stress-shift back two syllables from the French position also explains "analogue," "analogy," "analogical" (and also "monologue" and the rest of the "-ologies" and "onomies.")
I think the same shift also occurred in "nominate," "nominative," "vocabulary," even though these aren't directly from French. (Compare "relate" and the like, where there was no antepenult syllable to receive secondary stress so the stress remained on the ult.)
What's unclear to me is when the original stress becomes secondary stress, or when it becomes completely unstressed in modern English.
It's completely unstressed in "charity" and "nominative," but secondarily stressed in "vocabulary." It's secondarily stressed in all antepenult-stressed verbs ending in "ate," but completely unstressed in (almost?) all antepenult-stressed nouns ending in "ate." For some classes of words, like "inventory," it's completely unstressed (and deleted) in standard British English but gets secondary stress in American English.