“Loss of palatalization”: from tʂi tʂʰi ʂi to tʂɨ tʂʰɨ ʂɨ
The term “[i~j] is lost after retroflex sibilants in Mandarin” may not be the most accurate description. While it is true for two cases:
- As a medial /j/: 張: TCC ṭiâ̄ng (tyiāang) > MTCC ciong1 > Man zhāng - As a medial and final /ɥ~y/: 書: TCC shyū > MTCC shü1 > Man shū
It doesn’t quite explain the empty rime: 知: TCC ṭī > MTCC ci1 > Man zhī [tʂɨ˥]
I think the term “palatalization is lost”, as used in some works, may be more accurate. We understand when the medial /j/ becomes silent. That is a loss of palatalization. Likewise for when /ɥ~y/ loses its palatalization and unrounds to /w~u/. But this term also covers the loss of the palatalization, or the “frontness” of [i], which depalatalizes and backs to [ɯ], and which becomes [ɨ] allophonically after retroflex sibilants (“shibilants”). As such, the better way to transcribe this phoneme, when it is actually considered one (which is useful as analyses tend to merge it with /i/, which isn’t friendly to learners since merging [i] with [ɯ~ɨ] would only result in improper Mandarin. It isn’t a geminate consonant either and there is a clear vowel-like sound coming after the initials so calling those “just a lengthened buzz of the consonants” may lead to learners thinking that they’re just geminates.), it is generally written as /ɨ/ instead of /ɯ/. The reason for this transcription is likely because of pinyin, which writes /ɯ~ɨ/ as <i>, which may in turn be based off of its etymological (Middle Chinese) value. Hence, the phoneme may be better represented as /ɯ/.
Related point with historical Cantonese: The shift of TCC dsi tsi dzi si zi to Canto tsz ts‘z sz probably prevented the alveolar sibilants from palatalizing, thus merging in that period (when the alveolar and [alveolo]palatal sibilants were distinguished) with the Cantonese shibilants. The other way to look at it is, /i/ assimilates to the non-palatal-ness of the alveolar sibilants, thus losing the palatal nature, becoming “z”, as transcribed by Ball and Eitel. This again, is a “loss of palatalization”. The reason I think Ball recorded proper Cantonese at the period is because of his criticism of Eitel’s transcription, commenting that if one went to Canton, it would be evident that characters like “女” and “去” are (or were) not pronounced “nü” and “hü” as Eitel and the earlier Williams transcribed. Note that, though, Ball has transcribed TCC ci chi ji shi zhi as “chi ch‘i shi”, i.e. not with a <z>, and that the /i/ became presumably [ɯ] after alveolar sibilants, not the palatal shibilants.