Are there minimal pairs distinguished only by length of [y] in Latin? Was the short variant of /y/ pronounced like [ʏ]?
Great question! I was previously unaware of any such pair, but browsing Lewis and Short (available online in various forms) brought up these:
- lўsis: loosening, rupture, talon, ogee
- Lўsis: a Pythagorean of Tarentum, instructor of Epaminondas
- Lȳsis: a small river in Asia Minor
The letter is so rare in Latin that I doubt there are examples without Greek names.
The existence of /y/ and /y:/ in Classical Latin is mainly postulated as a part of non-nativized pronunciations for loanwords from Greek. Vowel length was phonemically distinctive for Y in Greek, and it was accordingly distinctive in Latin as well.
In the popular language, it is thought that merger with /i/ and /i:/ was a possibility already in classical times.
My impression is that it is rather difficult to determine at which points in time Latin did or did not have qualitative distinctions between corresponding long and short vowels. It does seem plausible that [ʏ] would have been used alongside [ɪ] and [ʊ], but I don’t think we really have any actual evidence for this.
With respect to latter portion of the question only, I agree with @sumelic's answer that it is unlikely we have actual evidence of a qualitative difference between /y:/ and /y/, but I will supplement that with the following considerations:
1) It is generally thought that the Latin high vowels were somewhat lowered in Classical times to [ɪ] and [ʊ]. If that was the case, I think there would be a natural tendency to lower short Y too, to [ʏ].
2) On the other hand, /y/ was not a native Latin sound, and I've never seen any evidence that Greek short ι and υ were qualitatively any different from their long equivalents. That being the case, an educated Latin speaker would be likely to pronounce short Y as in Greek, viz. [y].
So, I think there are arguments for both possibilities, and it's quite likely (in my opinion) that both could be used by different speakers.