ㅿ is a voiced dental fricative, hence /z/. The sound value is apparent from multiple sources of evidence:
- Earliest evidence for the sound is found in the 12th century text 鶏林類事 in which Korean words are transcribed in Chinese characters. Phonetic reconstructions of these characters give clues to the Korean sound /z/. Further evidence can be seen in similar texts coming later as well.
- A number of doublets exist showing variation between /s/ and /z/.
Example: 두ᅀᅥ (tuzeo ‘a few, a couple of’) and 둘서 (tulseo ‘id’, literally ‘two three’). (Note: the loss of /l/ here is a separate matter.)
- Similar to above, the sound can be deduced from compounds which etymologically go back to /s/. Hence, /s/ lenited to /z/ in specific phonological environments.
Example: 한ᅀᅮᆷ (hanzum ‘sigh’) is composed of 한 (han ‘big’) and 숨 (sum ‘breath’).
This sound is gradually lost between c. 1480 and the middle of the 16th century. This loss results in two phonological changes:
- In most cases, z > ∅
- In a few cases, z > j / [+nasal] _
The earliest example of this change can be found in the 1481 text 杜詩諺解 (Tusi ŏnhae) where ᄉᆞᅀᅵ (sozi ‘space, interval’) became ᄉᆞᅌᅵ (soi). Following this, a there is a period of variation over the next few decades. By the 1580s, ㅿ is exceedingly rare or completely absent.
The z > j change can be seen in texts such as 飜譯朴通事 (c. 1517) and 小學諺解 (1587). In both, for example, the Middle Korean word 손ᅀᅩ (sonzo ‘with one’s own hands’) is spelled 손조 (sonjo).
Notes: Font and rendering support for Middle Korean is limited by the OS. For Windows, it is supported from Windows 8 / 10 via the Malgun Gothic font. The font alone is not enough to shape the syllables, though and requires support from the OS. I cannot comment on the level of support in other OSes.