> Perhaps even "fee" isn't quite the correct ancient pronunciation for > phi.
It is more than likely that the ancient pronunciation (even if known) is inaccessible to the average English-speaking tongue (if not ear)--just as both of our "th" sounds are inaccessible to the average French-speaking tongue, while their "u" sound is inaccessible to the average English-speaking tongue. In these circumstances, I'm very much in favor of using the nearest approximation that English provides, and letting it go at that. I'm even willing to extend this convention to moving accented syllables around: Thus, to an English speaker, "Cauchy" becomes "KO-she" in place of the French "ko-SHEE" (and even that's an approximation--the French sounds are subtly different from the English sounds).
As to hearing, I, for one, have been immune since I was a young adult to the differences between the sounds of the hard and soft versions of certain Russian consonants. I simply could not hear any differences, no matter how many times the sounds were repeated for me. I imagine my spoken Russian was always a horror to a native speaker--even more than its utter lack of correct grammar would have supported. (Not any more--it no longer exists...)
--Lou Talman Department of Mathematical and Computer Sciences Metropolitan State College of Denver