English is troubled by what appears to be an unsystematic plethora of spelling rules, not to mention the rules for pronunciation. In general, there seems to be a consensus on how Greek and Latin is romanised in English, but now and then something pops up to remind us that it is not so simple. Greek κ, for instance, is often romanised as c, apparently being based on the Latin spellings. The letter υ, however, is more problematic: Mostly it is (problematically) romanised as u,¹ but in some combinations, the y is preferred, e.g.:
- words with the prefix syn-, such as syncopate
- words with μυ-, ψυ-, φυ-, such as myth (but music), psyche, physics
- words with initial hard breath followed by upsilon, such as ὑποθήκη → hypothek
But there are so many situations in which the expected romanisation of Greek → English υ → y is instead turned to υ → u, such as in αὐ to au, ευ/εῦ to eu, not to mention the standard romanisation of Greek words where English u is (to the best of my knowledge) always used, such as:
hypothec […] Origin Early 16th century from French hypothèque, via late Latin from Greek hupothēkē ‘deposit’ (from hupo ‘under’ + tithenai ‘to place’).
― Oxford Dictionary: ‘hypothec’
Thus my question: When should Greek υ be romanised as English u and y? I suspect the rule is that words that have been taken into the English language prefers y for monophthongs and u for diphthongs, whilst any romanisation of unassumed Greek words prefers u, but I am not sure this observation is correct, based on words such as music. My working hypothesis is that the following applies:
- If a monophthong and borrowed either directly from Greek or via French, the preferred spelling during the great spelling reform was y.
- If a diphthong, the preferred spelling was u.
- If borrowed directly from Latin, the word having been naturalised in Latin, the preferred spelling was u.
¹ By this, I also refer to the transliteration rules that usually are followed, as per the quote, in which υ → u is the standard method.