As a result, its main merit seems to be as a convenient device to get those ever-present, pesky yods out of the way.
Since the initials of class C words were generally kept separate from the initials of words in other classes, I will indicate this by adding a yod (*j) after the initial.
Downer in 1957 argued persuasively that the retroflex sibilants were not actually followed by a yod even when placed in a third-division rime; he maintained that it would be better to write *su rather than *sju, *dzung rather than *dzjung, and *sek rather than *sjek, even at the stage of the Chiehyunn.
It appears that, initially at least, Pulleyblank's chief ground for suspicion about the division-three yod was the lack of any such element in numerous early transcriptions of foreign words and names in Chinese; frequently the yod of the Chiehyunn reconstruction is totally superfluous (Pulleyblank 1962: 99).
Even a cursory inspection of Coblin's 1986 Sinologist's Handlist reveals a significantly large number of cases of Chinese forms with medial yod corresponding to Tibeto-Burman forms lacking a comparable element.
In view of other evidence that suggests that the yod of the third division has a secondary origin, it makes better sense to see these forms as coming from a stage of Chinese independent of the Chiehyunn, in which a palatal medial had not developed in the forms cited.
In the Wang Renshiuh redaction of the Chiehyunn studied by Lii Rong (1957), there is a total of 3,633 distinct syllables; of these 52% belong to division three; that is, they are syllables containing a palatal medial or yod.
Division-three syllables must, from a phonological point of view, be considered marked in comparison to those of other divisions, in that they contain an added feature of palatality manifested in the presence of a reconstructed palatal medial or yod.
Downer has shown that the yod after the retroflex initials *ts, *tsh, *dz, and *s is non-distinctive.