摘要： Korean verbs can be marked with both referent and addressee honorific morphology. An analysis of a teledrama corpus and a phone call corpus shows that these two morphological classes co-occur in a biased distribution indicating an association between the two classes. An experiment was conducted to determine whether Korean heritage speakers' acquisition of Korean was affected by this association as would be predicted by the distributional bias hypothesis. Twenty heritage learners of Korean performed a teledrama oral translation task which elicited two addressee honorific styles with and without referent honorific marking. A repeated measures ANOVA on the four possible addressee-referent honorific combination showed differences in performance. A post hoc analysis of pairwise contrasts indicated that performance was superior on the referent honorific (RH) plus hayyo addressee honorific (AH) combination relative to the RH plus hay AH combination. This result is incompatible with an account that explains acquisition in terms of the cumulative frequencies of the forms in input. It is also incompatible with accounts claiming that learners do not associate the forms during the acquisitional process. It is argued that the distributional bias hypothesis best accounts for the pattern of results and the frequency-driven conflation of semantically related concatenated affixes may have special significance for agglutinative languages such as Korean.

关键词: referent and addressee honorific morphology, distributional bias hypothesis, addressee-referent honorific combination, concatenated affixes