This study deals with the development and recoverability of word-final codas in Chinese-Swedish interlanguage. The relation between consonant deletion and vowel epenthesis is investigated from both a developmental perspective and a grammatical-functional one. Longitudinal, conversational data from three Chinese beginner learners of Swedish were analyzed. First, it is shown that for these learners the acquisition of Swedish codas was U-shaped rather than linear such that they exhibited relatively high accuracy rates at early stages, lower accuracy rates at later stages, and again high accuracy rates at more advanced stages. It is also demonstrated that the epenthesis-deletion differential is closely related to second language proficiency in that the proportion of epenthesis to deletion errors increases over time. Furthermore, the data show that word-final codas that are relatively important for the retention of semantically relevant information generate lower overall frequencies of simplification and greater epenthesis-deletion proportions than codas containing information that is relatively recoverable from other segments or features in the context.