This dissertation addresses the tension between the celebration of multilingualism and the assumed loss of translation: multilingualism will eventually turn translation from a professional task to an ordinary activity. Then, how should we deal with the loss in translation? What is exactly lost in translation? Should it be taken for granted? What if the loss is more or less a historical construct? Identifying the feeling of loss in translation, this dissertation argues that there is melancholy in language. What is assumed to be lost in translation was never there in the original because due to the heterogeneity of language transparent communication is never guaranteed. In assuming the loss in translation, one rather reveals a desire and intention for linguistic homogeneity and homolingual community, which are essentially colonial and at the same time a melancholic reaction to the ungrieved heterogeneity in language. Translation repeats the loss in language and returns to the heterogeneity, i.e., difference in multiplicity. This logic of translation that repeats and return to difference provides insight for the postcolonial paradox of returning to the colonial and for the diasporic return to the past for a break and new departure. The singularity of postcolonial diasporas thus reaches the heterogeneity of language through the temporal dialectic of repetition. Investigating this logic of translation, this dissertation focuses on Yi Sang and his bilingual writings during the Japanese occupation of Korea, Theresa Hak Kyung Cha's experiments with the mediality of language in DICTEE during the postcolonial Cold War era, and Ook Chung's diasporic language and dewriting [désécrire] of the origin in Kimchi. The multiplicity of languages in their works recounts difference as Yi Sang breaks the twosomeness of the colonial and the colonized languages into infinity and reveals the postcoloniality of difference as opposed to colonial monolingualism; Theresa Hak Kyung Cha focuses on the "Tertium Quid" that connects anonymized disconnections through second tongues; and the fourth square of writing in Ook Chung's Kimchi leads to an empty center of dissemination. Recounting multiplicity thus allows for working-through when translation repeats the loss of the original heterogeneity in language over and over again.