SUMMARY
This research aimed to describe the proficiency, selection and placement of krama lexicon. A sociolinguistics approach with the utilization of a descriptive qualitative collaborative method was used in the data analysis. The qualitative analysis employed bagi unsur langsung (BUL), permutation, and substitution methods. The results showed that: (1) the lexical proficiency of krama from Semarang coastal community reached 38.1%, (2) Javanese language contained overlapped speech, (3) there was self-kramanisasi in Javanese language, and (4) there was Indonesian language interference on Javanese. The deviations were caused by an error in language inheritance, low proficiency in krama lexicon, and the lack of comprehension of the Javanese speech levels usage.