The purpose of this dissertation study is to investigate how Korean-English bilinguals process compound words in both English and Korean. The major research question is: when Korean-English bilinguals process Korean or English compound words, what information is used to segment compound words into their constituents and, in particular, does morphological information play an independent role irrelevant to the form and semantic information?

Four masked priming experiments were conducted with adult Korean-English bilinguals. Compound words (e.g., bedroom, deadline) and monomorphemic words with a compound-like structure (e.g., hammock) served as targets and were preceded by brief masked primes corresponding to the constituent of the target stimulus (e.g., bed, room, dead, and mock). In Experiments 1 and 2, within-language prime-target pairs (Korean-Korean for Experiment 1 and English-English for Experiment 2), co-varying morphological decomposability, semantic and form relatedness were presented. In Experiments 3 and 4, cross-language prime-target pairs (Korean-English for Experiment 3 and English-Korean for Experiment 4), varying morphological decomposability, semantic and phonological form relatedness were presented.

In Experiment 1, results showed that morphological information plays a role independent of the form information when Korean-English bilinguals decompose compound words into their individual constituent morphemes in their L1 (Korean). In Experiment 2, however, there was no significant priming effect in all conditions, indicating that morphological decomposition is not relied upon in their L2 (English) processing. In Experiment 3, morphological information plays an independent role in the early stage of cross-language activation irrelevant to the semantic factor at the prime duration of 36 ms. However, morphological decomposition is constrained by semantic transparency in the later stage of cross-language activation at the prime duration of 48 ms and 100 ms. There was no significant priming effect at the two short prime durations (both 36 ms and 48 ms). However, there was a marginally significant priming effect in the +M+S-P condition at the longest prime duration (100 ms) in Experiment 3. Based on the pattern of these results, it seems that at the earlier stage of processing, phonological relatedness was important for morphological processing. In Experiment 4, there were no significant priming effects in all conditions across all of the prime durations. These findings together point to a clear asymmetry in the masked cross-language priming between L1-L2 and L2-L1 directions.