This Old Korean reconstruction is based on a Middle Korean form which has an intervocalic lenited fricative (/β/, /z/, or /ɤ/), and which has a non-lenited cognate (/p/, /s/, or /k/) in the modern Gyeongsang dialect. The Old Korean orthographical evidence, such as 初叱音 (*CHEsem) for Middle Korean 처ᅀᅥᆷ (chezem), and the general trends of diachronic sound change strongly suggest that the Gyeongsang dialect preserves the Old Korean forms.
Middle Korean features both lenited and non-lenited phonemes intervocalically, and the reason why certain forms escaped lenition is still poorly understood. In Korean-language scholarship, it is generally assumed that the non-lenited forms originally had a feature (e.g. a consonant cluster) that somehow blocked lenition. This is defended by Alexander Vovin, who believes that the non-lenited consonants were originally a consonant cluster involving initial *n or *l. Samuel Martin alternatively proposed that lenition occurred only before the minimal vowels <o> and <u>, and that apparently incompatible Middle Korean lenited forms were created by deletion of the minimal vowel.
The issue is further complicated by the fact that the same Middle Korean morpheme was sometimes lenited and sometimes not lenited. For instance, the inanimate agentive suffix 개 (-kay) is lenited in 울에 (wulGey, “thunder”) but not lenited in 벼개 (pyekay, “pillow”). This may suggest that lenition ceased to be productive at some point, explaining some of the more problematic contrasts.