The aim of this study is to investigate learners’ acquisition of Japanese lexical accent in different prosodic environments. Nineteen Iranian learners and one Japanese native speaker participated in the experiment and were asked to pronounce a dialogue using the same three-mora non-word in three different positions in the dialogue. The results show that native speaker pronounce the non-word with three possible accent types: Atamadaka-gata (i.e., the first mora has a high pitch (H) and all subsequent morae have low pitches (L)), nakadaka-gata (i.e., one or more than one mora that is neither the first nor the last within that word has a high pitch) and heiban-gata (i.e., the first mora has a low pitch and all subsequent morae have a higher pitch, and the pitch gradually lowers ). However, accent realization does not change in different prosodic environments. On the other hand, when analyzing the Iranian learners’ pronunciation, a total of seven accent types were recognized, and the results demonstrate that accent realization differs by prosodic environment: while LHL (an accent form like nakadaka-gata) appears frequently in a focal environment, LHH (an accent form like heiban-gata), which is similar to a Persian accent, is realized more in other environments such as neutral or post-focal environments. The above result suggests that since, at the beginning of conversation sentences and in focal environments, the learners’ degree of consciousness is higher than for the whole conversation, L1 interference is less likely to occur.