The addition of these phonemes to the language is occurring between the generations of speakers that are alive right now. Older people might have trouble with [ti di ɸi ʦi] etc. and pronounce them as [ʨi ʥi ɸui ʦui or ʨi], but for probably almost everyone around 60 and younger, the first three of those are pretty well established, e.g. pātī, supagetti, patishe, birudingu, kōdinēto, figyua sukēto, fāsuto, kafe, forudā (this one, "folder," actually has a minimal pair, horudā "holder"; they used to be identical, but "folder" shifted to match English more closely). /ʦ/ I think is more recent than the others, and still at least somewhat non-standard, which I guess is the original topic, but I think that if you knock a couple decades off of my original age limit you will find most people are fine with [ʦa ʦi ʦe ʦo]. [si] and [zi], on the other hand, as you mentioned, show no sign of entering the language.finlay wrote:As for the phonemicity of Japanese palatalized stops, it's definitely true that the thinking is vastly influenced by the writing system, and I think I'd agree at least for t vs ch, but even then many Japanese people mispronounce ti as chi in English. They seem to have more trouble correctly pronouncing di than ti, too (many people call the city Ejimbara rather than Edinburgh, for instance - but then, this is almost certainly coming from the katakana spelling of the city, which they take as the "correct" pronunciation). But che and she (ie, チェ and シェ) are pretty well-established, now, and they only occur in loanwords, and people don't have trouble pronouncing them at all. They have much greater trouble distinguishing si and shi, and I'm partly convinced that katakana has something to do with this (because there's no スィ in general use). For native words, you can basically argue that they're not phonemic, but it doesn't work for the modern language with all its English loanwords.
More accurately, you can argue that there's a neutralization or something between [tj] and [tɕ], because I'm pretty sure you can find somewhere where [tɕ] clearly is derived synchronically from [tj].
As for /ts/, yeah they have a few loanwords now like pizza as ピッツァ (instead of ピザ), but I think I've heard [pitswa] for this. And with f, again they have a way of writing it in katakana, but that doesn't necessarily mean they will then be able to distinguish it from [h] or [hw] or something.
The pronunciation of Edinburgh (or any older borrowing) is not really a good test for the existence of the phoneme, because in Japanese Ejimbara is simply the established pronunciation, and is unlikely to change regardless of the phonemic state of /di/. It's a lot like how in English we pronounce Paris with an /s/ at the end: we're capable of dropping it, but we aren't going to because that pronunciation is established. We need to look to more modern borrowings to see how they are realized.
/di/:
http://www.youtube.com/watch?v=pmYAFtcnDiY#t=3m35s (edī rōson)
http://www.youtube.com/watch?v=KYL2-ZRO_co#t=0m27s (diran)
http://www.youtube.com/watch?v=KYL2-ZRO_co (komedian)
http://www.youtube.com/watch?v=Aid1PHL6kOg (ōdishon)
http://www.youtube.com/watch?v=_7BdtRw4xoI#t=0m24s (kōdinēto)
http://www.youtube.com/watch?v=_7BdtRw4xoI#t=0m48s (uedingu)