Translation as Treason

First published in the “Japan in Translation” series on the Global Literature in Libraries Initiative blog on 12 May 2018, at the invitation of David Jacobson. Other contributors to the series included Emily Balistrieri, Phyllis Birnbaum, Ash Brown, Juliet Winters Carpenter, Zack Davisson, Peter Goodman, Kathryn Hemmann, Cathy Hirano, Terry Hong, Sako Ikegami, Sally Ito, Deborah Iwabuchi, Roland Kelts, Eve Kushner, Tony Malone, Paul McCarthy, Melek Ortabasi, Roger Pulvers, Lynne Riggs, Jay Rubin, Bruce Rutledge, Fred Schodt, Laura Simeon, Ginny Takemori, and Avery Fischer Udagawa. Originally written in April of 2018. See also Lucy Ferriss’s “How to Teach the Rhythm of Language: Stop Counting Syllables!”

In The Book of Tea, Kakuzo Okakura wrote that “Translation is always a treason, and as a Ming author observes, can at its best be only the reverse side of a brocade.” And so it seems to be—the art of translation is always a compromise—something will always be lost. This reality seems especially true for poetry, which relies heavily on nuance, allusion, cultural reference, wordplay, and other matters of art and craft that cannot always find a direct correlation in the destination language. Translation is therefore always a matter of tradeoffs, yet we would be impoverished as a globally connected society if we did not continue to try.

        Translation is also a victim, at times, of misunderstanding, and a case in point is haiku poetry—and such misunderstandings have led to even more being lost in translations than is necessary. The fundamental misunderstanding in this case is the belief that haiku is a poem of 5-7-5 syllables, which is really an urban myth, despite how widely haiku is taught this way in the West. The word “haiku” itself illustrates the issue—it’s two syllables in English but three sounds in Japanese. In fact, as noted Japanese scholar and translator Haruo Shirane emphasized in introducing Kōji Kawamoto’s The Poetics of Japanese Verse, “the term syllable is an inaccurate way of describing the actual metrical units of Japanese poetry.” As such, because “syllables” are not what poets count in Japanese haiku (they count sounds, or “on”), the pattern of 5-7-5 syllables in English is really a violation of the Japanese form, not a preservation of it. Thus, a poem of 17 syllables nearly always produces a poem that is significantly longer (with more content) than what is said in the seventeen sounds in Japanese. For this reason, nearly all respected haiku translators generally do not follow a 5-7-5 syllable pattern, more commonly producing versions of about 10 to 14 syllables. This choice is more faithful to the length and rhythm of the original, and helps the translator focus on natural speech rhythms and line breaks in English rather than padding or chopping lines to force an arbitrary and awkward adherence to an incorrect 5-7-5 pattern. Quite simply, readers need to stop counting the syllables in haiku to see if the poet (or translator) “did it right.” Moreover, this change of emphasis also enables the translator (and thus the reader) to pay attention to more important aspects of the haiku genre that are frequently obscured by an obsessive and misguided assumption that 5-7-5 syllables is accurate for English.

        Two of those more important aspects are the kigo, or season word, which is often straightforward to translate (although the allusive and cultural overtones of some Japanese season words may be lost on English readers), and the kireji, or cutting word, which is a sort of spoken punctuation in traditional Japanese haiku that divides the poem into two parts. English does not have cutting words, but their space-creating effect can be reproduced by giving the poem two juxtaposed and grammatically independent parts. To illustrate, here is a haiku by Bashō, which I translated with Emiko Miyashita, from Kunio Kobayashi’s book Bonsai (PIE Books, Tokyo, 2011):


                色付や豆腐に落て薄紅葉

                irozuku ya tōfu ni ochite usumomiji


                turning color—

                a lightly tinted leaf

                fallen to the tofu


In this case the cutting word, “ya” (や), divides the poem into two parts, with the cut marked in the translation by the em dash, but also by the grammatical independence of the first line from the rest of the poem. The season word, “usumomiji” (薄紅葉), or “lightly tinted leaf,” spans from early October in northern Japan to early November in the south—a late-autumn season word by how the Japanese count their seasons. The translation also avoids compromises to fit a misunderstood sound pattern. Other treasons may remain, but one hopes that the essence of the poem and its moment of celebration still comes through to the thoughtful reader.


Michael Dylan Welch served two terms as poet laureate of Redmond, Washington. His haiku, tanka, longer poetry, and translations have appeared in hundreds of journals and anthologies. His latest books are Jumble Box, Earthsigns, and Seven Suns / Seven Moons. He has been a longtime officer of the Haiku Society of America, cofounded the Haiku North America conference (started in 1991), cofounded the American Haiku Archives (1996), founded the Tanka Society of America (2000), and founded National Haiku Writing Month (NaHaiWriMo, held every February, starting in 2011; see www.nahaiwrimo.com). In 2013, he was keynote speaker for the Haiku International Association annual conference in Tokyo. Michael’s website, devoted mostly to haiku, is www.graceguts.com.