return to research page    

From 5-7-5 to 8-8-8
Haiku Metrics and Issues of Emulation --
New Paradigms for Japanese and English Haiku Form


Publication: Language Issues: Journal of the Foreign Language Education Center (vol. 1)
Prefectural University of Kumamoto, Kumamoto (March 2000) Japan.



The question of how English-language haiku form may best emulate Japanese 5-7-5 haiku (or whether it even should at all) has been hotly debated for decades. A recent trend in Japanese poetic analysis, however, interprets haiku in terms of 3 lines of 8 beats each onto which the 5-7-5 -on are mapped. This paper presents an overview of this trend, supported both by theory from metrical phonology and by observed experimental data of subjects reading haiku in Japanese. It was found that the 8-8-8 metrical pattern is indeed verifiably present in haiku reading, and that this pattern serves to map both haiku with 5-7-5 -on and other  -on counts. Based on these findings, implications for English haiku form, especially with respect to emulation, lineation, and metricality are discussed within the context of the North American haiku movement. It is proposed that haiku in both Japanese verse and English free verse may naturally fit into a similar metrical form. It is hoped that a metrical analysis, operating across both languages, may help clear up some misconceptions regarding the Japanese haiku in the West, while providing an impetus to bridge the gap between the Japanese and world haiku movements. 


In 1952, R. H. Blyth eloquently described some of the qualities of Japanese haiku that make this poetic form one of the most unique in world literature--qualities which have sparked a worldwide study and practice of haiku, in numerous languages. His description has lost none of its relevance:

It is not merely the brevity by which [the haiku] isolates a particular group of phenomena from all the rest; nor its suggestiveness, through which it reveals a whole world of experience. It is not only in its remarkable use of the season word, by which it gives us a feeling of a quarter of the year; nor its faint all-pervading humour. Its peculiar quality is its self-effacing, self-annihilative nature, by which it enables us, more than any other form of literature, to grasp the thing-in-itself (vol. 4, p. 980).

To a large extent, the evolution of form in English haiku has been wedded to the qualities Blyth (along with others) outlines above, whether considered from an aesthetic, experiential, or literary perspective. That is, the Japanese haiku, and the literary culture which bore it, has provided a model example for a new form of English, and indeed global, poetry. As this form has evolved, the question of how the English haiku might best emulate the Japanese has given way to a growing body of poetic works that maintain themselves autonomously, outside of Japanese-haiku referents. Nowadays, some North American poets may be inspired by a growing body of English haiku composed by poets outside Japan, and have little interest in Japanese culture or literature. "An important change that is occurring in American haiku is the decrease in those being introduced to haiku through Japanese culture and an increase in those discovering haiku from the poetry-writing arena . . . I think that the world's haiku poets recognize our common heritage in Japanese haiku, and at the same time acknowledge that Japanese and American haiku will likely grow apart" (Gurga, 1997).

English haiku has firmly established itself as a distinct free-verse poetic form, and in general, the passion for some sort of "mirror-like" emulation of the Japanese haiku has in many quarters either devolved into or achieved (depending on your point of view) mere inspiration, complete autonomy, or divergence. Speaking of devolution, Ross (1993) wrote that,

The fourth generation [of the mid-1980s on] of American haiku poets has through experimentation all but obliterated the requisite form and substance of classic Japanese haiku: there is a consistent lack of seasonal references, surrealist techniques and figurative expression are introduced, regular prosody is eliminated, and human, rather than nature, subjects are increasingly emphasized. Contemporary American haiku has been made a poetic vehicle for eroticism, psychological expression, political and social commentary (p. xxiii).

One could argue that statements along these lines have likewise been made in recent decades in Japan, regarding the Japanese haiku, so perhaps it is modernity itself that is in question. Nonetheless, recent polemics beg the question of what English haiku form is, and how to delimit and define this form from other short poetic forms, as well as in relationship to the Japanese haiku. While these questions of haiku content and aesthetic are of pressing relevance, and surely depend upon critical contributions for the English haiku to flourish, in this paper we would like to address the "framework" of haiku by taking a new look at the Japanese haiku, and offer some suggestions for how a metrical approach might be considered as a bridge between haiku forms and languages, English and Japanese.

We wish to demonstrate the validity of a novel metrical approach to haiku form, whose theory and practical application has only recently been demonstrated in phonological studies,[1] providing a definitive validation for both 'tighter' on-counts in Japanese and flexible syllable-counts in English. We will show that the 5-7-5 moraic[2] structure of Japanese haiku "floats," as it were, upon a deeper template, a timed, metrical substructure, composed of three segments[3] of four bi-moraic feet, making for a total of 24 mora, into which the 17 (+/-) mora of 5+7+5 mora fit. This structure can be viewed graphically in Figure 3, "A Foot-based Template," below.)

We will then apply an analogic form of this same metrical template to the English haiku. This metrical approach is not intended to provide a prescriptive "rule" for haiku composition in English in terms of a bilingual mirroring; nevertheless, the template is serendipitously inclusive of the bulk of acclaimed contemporary free-verse haiku. Since English haiku has reasonably abandoned strict syllable-counting, a metrical approach may serve to redefine, delimit, or refine English haiku length, meter, and closure, beyond the concept of 'as long as a single breath,' or 'short,' or 'about X (choose your number) syllables.' We might instead say that the English haiku approximates the tripartite, 24-mora metrical structure of the Japanese haiku, and that analogical forms of metrical emulation are possible. A metrical approach allows for a great deal of flexibility in the English haiku syllable-count, as an intrinsic constituent of its definition, while maintaining an identity to Japanese haiku metrical form. Some may find this an elegant solution. It seems that as long as we continue to use the term "haiku" in English to define the poetic form, linkage to the Japanese haiku is desirable.

In traditional Blues, a 12-bar musical structure, there are nearly an infinite variety of styles, tempos, and contexts. Blues-like musical structures can be heard in the indigenous musics of many countries. Thus, radically divergent contexts can share similar musical structures. As musicians know, feeling the sense of rhythm and sharing rhythmic structures from beyond one's own shores bridges languages and cultures. At the dawn of the modern free-verse movement, the Imagist Manifesto (item #3) stated: "As regarding rhythm: to compose in the sequence of the musical phrase, not in [the] sequence of a metronome."[4] Current linguistic theory, and public readings too, indicate that the English haiku has music and, as in all languages, metricality, though the English haiku does not possess as regular a meter as the Japanese haiku.[5] It may be that many English haiku abide analogically within the 24-mora metrical structure of Japanese haiku, not in prosodic particulars (involving a one-to-one identification of Japanese mora with English syllables), but in linguistically valid metrical terms. It is not our intention to provide a proof at this point in time, but rather to explore some implications. We hope that this paper serves to encourage further consideration and study of Japanese and English rhythmic haiku expression, and promote cross-cultural dialogue. 

Contemporary English Haiku Examples and Issues

Just as a person may best be known not through analysis in absentia but through actual meeting, the English haiku is perhaps best met through example rather than definition or analysis. Although many authors discuss the "English haiku tradition," this tradition, traceable to the first "hokku-like" success in English by Ezra Pound (1913),[6] for the most part begins in the post-WWII era--so is a tradition barely fifty years old.[7] Variability, variation, and experiment remain rife and vital in all aspects of the poetic form.

Foment continues within English haiku circles regarding the definition and further evolution of English haiku. There are questions concerning emulation, including whether any formal emulation of the Japanese haiku is advisable or necessary; unresolved issues related to season words (kigo) and settings (kidai); syllable counting; lineation; content; aesthetics; inclusiveness of contemporary Japanese haiku which substantially depart from traditional form--the list goes on. In fact, Sato (1999a) has suggested that, "Today it may be possible to describe haiku but not to define it. . . . All you can say is that a haiku, be it composed in English, Japanese, or any other language, is what the person who has written it presents as haiku."[8] The large majority of English haiku do have consistent defining features, both aesthetic and formal, as you may observe in the seventeen selections below, chosen to illustrate a range of presentational styles. Many are recently published in major anthologies, and/or have won prizes in national or international contests:

From Snow on the Water: The Red Moon Anthology of English-Language Haiku, 1998:
my head in the clouds in the lake

- Ruby Spriggs (1)

frost predicted
     the sizzle
of fried green tomatoes

                - R.A. Stefanac (2)


Just enough of rain
To bring the smell of silk
From umbrellas

     - Richard Wright (3)

sitting here
without the mountains

     - Gary Hotham (4)
From Haiku in English, 1967. First Prize, Japan Airlines and American Haiku Contests, 1964:
      Lily :
      out of the water . . .
      out of itself.

             - Nicholas Virgilio (5)
From Yomiuri Shimbun,1998. First Prize, International Haiku:
     spring    wind --
     I      too
     am     dust

           -Patricia Donegan (6)



From Haiku Moment, 1993:

in morning sun two white horses      the autumn aspen

        - Elizabeth Searle Lamb (7)


the old irrigation ditch    gathering in     autumn dusthaze

                            - Elizabeth Searle Lamb (8)




     as the wave retreats
                  into itself

      - Geraldine Little (9)

waterfall at night--
          her long

- Charles Nethaway, Jr. (10)


   a barking dog
   little bits of night
                                  breaking off
      - Jane Reichhold (11)

     arms folded
     i watch the crane
     standing on one leg

       -Cor van den Heuvel (12)


       the cat in
       the fog in
                - Vincent Tripi (13)
From Red Moon Anthology,1997:
     the river
              the river makes
      of the moon

            - Jim Kacian (14)

From The Haiku Anthology, 1999:

the fingers of the prostitute cold          - Bob Boldman (15)

subway woman asleep
picked daisies
in her hand

     - Raffael De Gruttola (16)

   20,000 feet
   traces of masking tape
   on the jet engine

            - Dee Evetts (17)

In contrast to several of the above selections, many haiku typically take a "short-long-short" three-line form. Our acclaimed haiku (above) range from 6-16 syllables, and the actual genre-range goes both higher and lower. The overall locus of 11-12 syllables emulates the semantic reality (amount of information contained) within translated Japanese haiku, as divined by Higginson and others.[9] Nevertheless, as can be seen in our example haiku that vary from the average, information is also a qualitative affair. In English, a haiku with 6 syllables may "read" (be spoken typically) at a time-length similar to a haiku of 16 syllables. In terms of emulation of the Japanese haiku, translators give us a guideline, and a syllable-count locus, but the relation of total time-length and "amount of information" is a fuzzy-logic equation in English.

Swede (1997) reports that in two surveys, the first conducted in 1980, and the second in 1997, roughly 93% of the haiku published in the two major North American periodicals, Modern Haiku and Frogpond, were presented in three lines. "Despite the efforts of some to promote one-, two-, and four-line haiku as well as visual haiku, the combined use of these forms has actually gone down . . . to an overall average of 6.6%." [The 21st century has seen a substantial rise in the number of one-line English haiku.] In terms of 5-7-5 syllable-counting in English, roughly 75%--80% of the haiku published in the above-mentioned periodicals are "other-than," while some 22% on average are in a 5-7-5 form--actually, 9% in Frogpond and 35% in Modern Haiku. The editors of these periodicals report neither encouraging nor discouraging the 5-7-5 English-syllable form. In general, the evidence suggests that editorial criteria are not prejudicial. Clearly, a large majority of haiku are no longer being composed according to the 5-7-5 syllable count. Many familiar with English haiku, the present authors included, consider strict 5-7-5 syllable-counting to be a poor method of emulating the Japanese form, if that is one's compositional intention.

 Some of the main issues in contemporary English haiku are that: 1) the syllable counts and 2) rhythms in English haiku are more variable than prescriptive guidelines for emulation of the Japanese haiku allow. Also, 3) rhythmical divisions are more varied. In Donegan (Ex. 5) we have a haiku of 6 total syllables. De Gruttola's haiku (Ex. 15) contains 12 syllables, twice as many as Donegan. Which takes longer to read in a typical reading by the same individual? We can see that Donegan, through word-spacing and selection, choice of line breaks, and punctuation, has created qualities which suggest rhythmic lengthening. Thus, total syllable counts cannot be considered apart from their intimate relation with rhythm and phrasal cadence. Donegan's haiku is a good example of the creative possibilities of free-verse English poetry, and it is this tradition that most adequately defines the basis of English haiku, in terms of rhythmic and verse-line variation. One of the foundational statements of the Imagist Manifesto (item #2), emphasizes the need for variability and creative novelty in free-verse poetry:

To create new rhythms -- as the expression of new moods -- and not to copy old rhythms, which merely echo old moods. . . . We believe that the individuality of a poet may often be better expressed in free-verse than in conventional forms. In poetry, new cadence means a new idea (Poetry, 1913).

Many poets these days no doubt find substance in the above statement. Without such a basis, one might wonder whether English haiku form would have so rapidly evolved. It appears that in terms of form, the English haiku movement is unabashedly a free-verse movement, and it resists or ignores attempts to restrict the main formally creative elements or tools of the free-verse poet: 1) the freedom to select a variety of syllable-counts (and accents) per line, 2) to select phrasing, 3) line-breaks, and 4) adapt the visual presentation to serve, on the whole, 5) the sense of cadence in its unique relationship to images and "idea." Of course, free verse "freedoms" must be seen as operating within the overall haiku aesthetic in the poem, which delimits its total length and influences its style and diction, thus creating a unique poetic form. 

Historical Evolution, In Brief

In the 1950s, works from R.H. Blyth, Harold Henderson, D.T. Suzuki, and Allen Watts became available. This was at a time when there was a resurgent interest in Asiatic arts, culture, and spirituality (especially Zen Buddhism), in the years following the American occupation of Japan, and also the following the Korean War.[10] It was this intellectual and spiritual ground that provided a means for haiku to firmly take root as a Western poetic form in English.[11]

Primarily, the two scholars, Blyth and Henderson, founded the tradition of English haiku, and their works (along with Yasuda, 1957), continue to remain important sources for haiku study. Both men served in Japan during the years of the U.S. occupation, sharing their interest in Japanese literature and haiku. (Blyth accepted a job offer from Henderson, then Special Advisor under General McArthur, but was unable to find housing in Tokyo, resulting in his accepting a professorship at Peers' School, and becoming tutor to the crown prince.[12]) In 1949, the first volume, Eastern Culture, of Blyth's four volume work, Haiku, appeared, followed in 1952 by the remaining volumes, Spring, Summer-Autumn, and Autumn-Winter. Henderson's An Introduction to Haiku: An Anthology of Poems and Poets from Basho to Shiki, appeared in 1958. This was an expanded and revised edition of his earlier work, Bamboo Broom, published in 1934.[13]

Even considering the increased interest in haiku form and its development, Japanese haiku and possibilities in English might have remained minor cultural footnotes if it hadn't been for the publication and popular success of Jack Kerouac's The Dharma Bums, in 1958 (following the publication of On The Road, in 1957, and his resulting rise to fame). Kerouac did something for the haiku movement no amount of scholarship alone could, in creating the character Japhy Ryder, a scarcely-veiled portrait of the poet Gary Snyder. Japhy transplants something of the Japanese haiku ethos, or an imagination of it, into the heart of American vernacular. Japhy seems like a modern-day gloss on Basho--a kind of Basho cum Li Po cum Oregonian lumberjack: "From the beginning a woods boy, an axeman, a farmer . . . . his face was a mask of woeful bone, but his eyes twinkled like the eyes of old giggling sages of China, over that little goatee . . . . he'll make the top of your head fly off, boy, with a choice chance word." In his pilgrimages into natural settings and intuitive feeling for nature, acquaintance with Zen practice and philosophy, simple lifestyle and dwelling-place, Japhy tantalized and inspired readers with novel possibilities for perception, spirituality, lifestyle, and poetic process.

Early in the novel, Kerouac (as Ray Smith), living in Allen Ginsberg's (Alvah Goldbook's) "rose-covered cottage" in Berkeley, California, notes that, "On the walls are hundreds of books everything from Catullus to Pound to Blyth" (p. 17). These signal authors are the only ones mentioned. A few days later, Ray, hiking with Japhy, exults:

     "Oh this is like an early morning in China and I'm five years old in beginningless time!" I sang out and felt like sitting by the trail and whipping out my little notebook and writing sketches about it.
    "Look over there,"sang Japhy, yellow aspens. Just put me in the mind of a haiku . . . . A real haiku's gotta be simple as porridge and yet make you see the real thing, like the greatest haiku of them all probably is the one that goes 'The sparrow hops along the veranda, with wet feet.' by Shiki. You see the wet footprints like a vision in your mind and yet in those few words you also see all the rain that's been falling that day and almost smell the wet pine needles" (p. 59).
The haiku by Shiki is a quotation from Blyth (1952, vol. 2, p. 517), who provides commentary:

Nureashi de suzume no ariku rouka kana

The sparrow hops
Along the verandah,
With wet feet.

This might be taken as a model for all haiku. It is poetical and yet extremely matter-of-fact. It is like one of those perfect jokes, so simple, so inexplicable. The delicate three-pronged little marks on the floor of the verandah, so soon to dry up and vanish for ever, as transitory as the pyramids or the solar system, --what an infinity of meaning in them!
It is an interesting contrast, the almost domestic, earthy vision of Japhy and the sense of Buddhistic universals in Blyth's speculations on impermanence. These two contemplative polarities, the sensual and the philosophic (especially Buddhist), emanated as prime Western responses to the presentational immediacy of haiku, and continue to inform North American haiku culture.

Books devoted to haiku study, along with some excellent anthologies and chapbooks, might grace the Poetry bookshelves of the local bookstores--and are sought by those in pursuit of a deeper understanding. Kerouac's novel on the other hand, remains popular[14] and available in numerous languages, thus continues to provide an entree to haiku for younger generations of readers, for whom the book has a strong appeal.[15]

The 1960s saw the beginning of several important haiku periodicals, the first being American Haiku, in 1964. The same year, Japan Air Lines (which sponsored a number of international haiku contests), received over 41,000 submissions--these contests stimulating a further popular interest in haiku. In 1967, the first issue of the Canadian periodical, Haiku appeared, and the same year Haiku in English by Henderson was published. This book remains one of the few guides for those who wish to both study and teach haiku form and practice. In the following year, the Haiku Society of America (HSA) was founded by Henderson and Leroy Kanterman, with the encouragement of the Japan Society of New York. From this point on, haiku in North America began to be widely written; many local writing groups and societies sprang up, and numerous periodicals came into existence.

In the 1970s, the Western haiku flourished and became an international phenomenon. By this time haiku was being taught in many American grade schools,[16] often as a means of syllable-study. Thus, English haiku as 'ever and only' 5-7-5 syllables got caught in the cultural craw. In 1974, the first edition of The Haiku Anthology, edited by Cor van den Heuval, was published, attracting international attention, and the HSA became similarly well known. Along with haiku festivals, a raft of small press publications (in many ways the lifeblood of the haiku movement) came into existence. At this time, dissent increased concerning what exactly was and was not a haiku, in terms of form, style, and subject matter. To date, a careful chronicling of these debates has not been published. The publisher and haijin (haiku poet), Jane Reichhold, referring to the formal issue of syllable-counting, remarks:

Did you know that haiku wars were waged in the 70s over this issue of onji "syllable" counting? Friendships were permanently destroyed. Haiku groups split up. New ones formed. Persons were reviled. There was much sneering, jeering, and rejection. It was terrible. The problem remains and is just now entering the tanka scene. . . How can we count our syllables and equate them with this unknown factor which the Japanese count and hold in such high esteem? (J. Reichhold, July 11, 1998. Personal communication)

The bizarre and mistaken term onji, and suitable replacement terms (on, moji, ji) are discussed in Sato (1999b) and Gilbert (1999). Arguments over terms, haiku form and syllable counting, were emblematic of misunderstandings and reductive half-truths (continuing to be promulgated), caused by language differences between Japanese and English, and a pervasive lack of information, scholarship, and cross-cultural dialogue. For the most part, useful new information continues to be published only in a small number of journals in limited circulation.

The majority of haijin in the West, while interested in the Japanese haiku aesthetic, are not engaged in studies of the Japanese language, and so rely upon translated materials. Japanese scholars addressing North American audiences are likewise rarely well versed in the contemporary North American literary tradition, especially in terms of its vernacular rhythms, free-verse poetics, and other contextual intricacies. As you may notice from the contemporary haiku examples above, free-verse English haiku rely on a tradition intimately wedded to vernacular speech patterns--a most difficult area to attune to, except for a highly skilled second-language learner of English. A third element in continuing misunderstandings has to do with the lack of accurate information. One finds little academic research devoted to explanations of Japanese prosody in the available literature; what does exist has been written for linguistics specialists, not practicing poets or students of the English haiku, who as a group find little benefit in these specialized discussions. Scholarly cross-cultural bilingual multi-disciplinary generalists with an ear for poetry and a desire to communicate to an audience of poets don't crop up often, though haiku literary culture is blessed when they do, as can be seen in the cases of Blyth, Henderson, Yasuda, Higginson, Ueda, Sato, and others.

The 1980s brought further refinement and questioning[17] of the English haiku tradition, as well as increasing recognition for individual poets. The year 1985 saw the publication of another important work The Haiku Handbook: How to Write, Share, and Teach Haiku by William Higginson and Penny Harter. It remains the most comprehensive reference work in one volume for the aspiring haijin,[18] though it is in need of revision in some areas. Particularly, this essay will address Higginson's discriptions of traditional Japanese and English haiku form, in terms of recent developments.

As the 1990s roll into the new millennium, North American haiku is rapidly becoming a subset of international haiku, with the common medium of exchange being the English language. [19] We are in the midst of another sea-change in the haiku movement; haiku appears to be the world's first global poetic genre. Some fundamental questions regarding haiku definition and form are very much with us. The haijin Lee Gurga, quoted by Takiguchi (1999) states, "Some say that they are now tired of this 'What is haiku?' question repeated countless times. Others still insist that this and other fundamental questions on haiku 'have not become old questions.'" Takiguchi, organizer of the "World Haiku Festival 2000" in Britain (among the first international conferences of its kind), writes:

While not only Japan but also the rest of the world seems to be enjoying an unprecedented popularity and proliferation of haiku, there are some worrying signs as well. The history of haiku is a succession of prosperity followed by decline. So, the ups and downs of the haiku movement are nothing new. What is different in today's haiku scene is that prosperity and deterioration are there simultaneously. It has been pointed out symptoms of the include stagnation existing movements, lowering standards quality haiku, commercialization factional rivalries, self-aggrandizement corruption generally. They have seen to be sapping health vitality yet precious little seems done about it. How this state affairs come can we possibly free ourselves from situation?[20]

It is hoped that an investigation of Japanese haiku metrics will help vitalize the haiku movement by providing a new perspective on haiku rhythm, and aid in a redefinition of Japanese haiku, encouraging alternative modes of conceptualization and definition relating to English (and other-language) haiku. 

Modes of Emulation

Those only tangentially acquainted with English haiku may recall hearing that haiku are composed of five, plus seven, plus five syllables, equaling seventeen. Henderson, in a letter written in 1971, mentions that "the rigid 17-syllable requirement . . . does not exist in Japan . . . and who started it for English syllables I do not know." Probably, the 5-7-5 "rule" began after any number of people read that "haiku are composed of 5-7-5 syllables." Some few authors who described the Japanese haiku were aware they were not referring to English syllables, and were also aware of the moraic nature of Japanese. However, to this day these distinctions are rarely spelled out succinctly anywhere in the literature.

Henderson himself, in his Haiku In English (1967), seems a bit mixed concerning 5-7-5 syllable-counting in English. He states quite clearly that: "The 5-7-5 norm of duration-count has proved appropriate for classical Japanese haiku. What form is best for haiku in English is still an open question. It has, we think, been shown that an invariable 5-7-5 syllable count is not an adequate answer (p. 32)." Nonetheless, his advice to teachers (especially teaching beginners) is: "The most general area of agreement is on form. It is generally taught that the form should be 17 English syllables divided into three lines of 5, 7, and 5. A few modify this by adding "about" or "approximately." Almost all specify that a haiku should be unrhymed. [There are] advantages, for beginners, of using a strict form . . . (p. 47)" As well, the book contains twenty "Suggestions for Beginners and Others" by J. W. Hackett. Rule #7: ""Don"t write everything in 5,7,5 form, since in English this often causes padding and contrivance." Rule #8: "Try to write in three lines of approximately 17 syllables." We also find that nearly all of the haiku quoted in the book are in 5-7-5/17 syllable form. So, we discover that an "invariable 5-7-5 syllable count" is not the answer, but the implication is that something quite close (nearly invariable?) is desirable. Hackett's "don"t write everything in 5,7,5 form . . ." has a similar implication.

This influential book became counsel to many "who would like to write serious haiku," as the blurb mentions on the back cover of our edition. Indeed, this book remains valuable, with a wealth of well-presented information. Nevertheless, in the matter of form it has been misleading, due to the linkage of 17 "Japanese syllables" to 17 in English, and the implied approximate goal of how these syllables are divided.

As for Blyth, he did not choose to address the subject of how to write English haiku, concerning himself mainly with issues of translation and comparative literature. However in his sixth and last major work on Japanese haiku, A History of Haiku, Volume Two (1963), within the final chapter, "World Haiku," he addressed some formal concerns:

But wherever haiku are composed, the problem of the form must arise. Europeans and Americans have to decide whether their haiku are to be in rhyming couplets or triplets, alliterative verse, free verse, what some rude people call "a dribble of prose," or in five, seven, five syllables as in Japanese. As far as the last is concerned, a strict adherence to 5, 7, 5 syllables in English has produced some odd translations of Japanese haiku. . . . The fact is that "syllable" does not have the same meaning for the Japanese, the Romans and Greeks, and the English.
In a potato,
Those groans whose forced prayers change nought,
Can never occur
This is 5, 7, 5, but to eye and ear, and to the sense of counting, the 5, 7, 5 has no meaning whatsoever . . . . The haiku form is thus a simple and yet deeply "natural" form, compared to the sonnet, blank verse, and other borrowed forms in English. The ideal, that is, the occasionally attainable haiku form in English would perhaps be three short lines, the second a little longer than the other two; a two-three-two rhythm, but not regularly iambic or anapestic; rhyme avoided, even if felicitous and accidental. A season word is not necessary, nor even a season, but is greatly advantageous, as suggesting one quarter of the year in time (pp. 350-352).

Some of Blyth's suggestions continue to be followed. The English haiku has settled on an unrhymed "dribble of prose" certainly, usually in three short lines, many with the second line longer than the first--and a number of haiku do fall into a 2-3-2 rhythm (though more do not; it seems doubtful this is a purposefully selected rhythm in the composition of a majority of contemporary haiku).

Other than discussions of ellipsis and how to break verse lines, often related to the kireji (the "cutting word" in the Japanese haiku), Blyth's and Henderson's suggestions seem to well-express the limits of formal description, in major published works until Higginson's Haiku Handbook (1985). By the 1980s, Higginson was able to state that, in terms of translation, "approximately twelve English syllables best duplicates the length of Japanese haiku in the traditional form of seventeen [-on][21] (p. 102)." Higginson also mentions independent corroboration by Sato and Maeda Cana. In the same chapter, "A Traditional Form for English Haiku," Higginson mentions these main points:

1. For haiku in English an overall form consisting of seven accented syllables, plus unaccented syllables up to a total of about twelve, would yield a rhythmical structure native to English and at the same time approximate the duration of traditional Japanese haiku. A major grammatical pause between the second and third or fifth and sixth accented syllables would provide the sense of division created by the Japanese kireji.
2. While Japanese are used to reading texts in which rhythms are not visually identified, the Western notion of a printed poem-text incorporates the idea of a line of type equaling a rhythmical unit, or verse-line. Therefore, a three-line structure of two, three, and two accented syllables, respectively, would establish rhythmical proportions similar to those of traditional Japanese haiku.
3. Since the most commonly encountered short structure in traditional English poetry is the "heroic couplet" with two five-beat lines, the two-three-two beat structure with a strong grammatical break after the second or fifth beat, as proposed, would yield a sense of rhythmic incompleteness similar to that in Japanese haiku.
4. Grammar should be stripped to the minimum that seems reasonably natural. Complete sentences may or may not occur; articles ("a, an, and the") and prepositions should be used sparingly, but not naturally omitted (pp. 105-106).

Higginson added this caveat:

. . . those who wish to write what they think of as "traditional haiku" in English may use them as guidelines. But since haiku have only been written in English for a few score years, it seems a bit early to call any kind of haiku in English "traditional" (p. 106).

Higginson summarized the traditional form of the Japanese haiku as:

1. A Japanese haiku  . . . usually has two rhythmical units, one of about twelve [-on] and one of about five, the break between them often marked with  . . . a kireji. 
2. Since the break between the two rhythmical units is equally likely to occur after the first five [-on] or after the first twelve, the normal rhythm  . . . in Japanese is five, seven, and five [-on]. 
3. The form of traditional haiku originated in the incomplete opening stanza of a longer poem; the haiku form is therefore rhythmically incomplete. 
4. Haiku often omit features of normal grammar . . . (p. 104).

Higginson's guidelines, being the most refined, extensively detailed, and widely known, will be addressed further on in the section "Implications for the English Haiku." 

The Metrics of Japanese Haiku

Of Moras and Syllables

Every Japanese schoolchild knows traffic-crossing aphorisms such as te o agete oudan houdou watarimashou (raise your hand/to cross at the crosswalk) as well as the rather tongue-in-cheek anti-traffic aka shingo minna de watareba kowakunai. (at the red light/all-together, crossing/is fearless!) Such aphorisms, called hyogo in Japanese, are almost invariably based on the traditional haiku pattern of 5-7-5 -on. Japanese children are thus exposed to this pattern from quite a young age. When they get a bit older, they study hyakunin-isshu, a set of 100 classical Japanese waka[22] patterned in 5-7-5-7-7, which form the basis of a card game often played around the New Year at home and in school. A popular TV program starts off each morning by introducing haiku sent in by viewers--but calls these haiku "5-7-5" (rather than haiku), presumably to avoid pedantism and encourage participation. A recent episode of the popular "Chibi maruko-chan" shows a reticent old man who ends up winning a haiku contest by inadvertently producing "masterpieces" such as dou shiyou nani mo haiku ga ukabanai (What to do? Not one haiku comes to mind.) These few examples show that rhythms based on 5s and 7s come quite naturally to people in Japan.

As discussed in the previous section, English speakers trying to discover the essence of haiku hit upon the magical mystical 5-7-5 -on triplet, and first decided that the closest equivalent to this would be 5-7-5 English syllables (this also turned out to be a boon for elementary schools trying to teach younger children about syllable structure). However, the difference between a syllable and a mora, the linguistic term for units which behave like Japanese -on, is great. The structure of an English syllable is usually linguistically characterized as a 3-part affair: onset (0-3 consonants), body (single vowel or diphthong) and coda (0-4 consonants). The latter two are often grouped together as the rime, a unit with which every preliterate child is at least somewhat familiar, thanks to the Cat in the Hat and Sesame Street.

The mora, on the other hand, comes in two kinds: the strong mora is linguistically structured as (C)V (consonant/vowel), but for many Japanese is barely conceivable as a set of one or two phonemes. A mora is a mora, not a set. It is written with a single graph: [ka] does not break down into /k/ and /a/ any more than English "I" /ai/ breaks into /a/ and /i/.

There are also three weak mora in Japanese: /N/ (a syllabic nasal which cannot appear word-initially), /Q/ the first half of a geminate consonant, and /V/ or the second half of a long vowel. Naturally none of these comes anywhere near to syllabic standing, and yet they are all (usually) fully counted in Japanese haiku.

In terms of English syllable structure, it may be helpful to compare the strong mora to a unit equivalent to the onset plus the nucleus (with the caveat that the onset may be only 0-1 consonant and the nucleus a monopthong). The weak mora then approximates the final part of the syllable--the coda. In the illustration below, the Japanese "pan" contains two mora as: onset-nucleus /pa/ plus coda /n/.[23] It is clear from this conceptualization that the mora is at best only about 1/2 the length of an English syllable. 

Figure 1. Three syllable structures: flat, onset-rime (English), head-coda (Japanese, cf. Katada 1990) (Yoneoka, unpublished manuscript).

This inherent difference in length between the mora and syllable (and consequent difference in terms of the volume of information that can be conveyed) was perhaps the biggest reason for reformulations of emulation in terms of "11 or 12 syllables," as Higginson concluded, or "usually . . . totaling fewer than seventeen syllables" as defined in the HSA haiku definition.[24] In the next section, we will see that this instinctive number of 11 or 12 is indeed a linguistically valid and appropriate figure, based on the underlying rhythmic template found in Japanese haiku.

It should be stressed at this point that although we have been contrasting mora with syllables, the two are actually quite different units. Specifically, the mora functions in Japanese both as a temporal and as a structural unit, whereas the syllable is solely a unit of structure in English. Thus, the mora itself is considered to have a (not necessarily isochronous) temporal value, while the length of a syllable unit may vary greatly with the number and quality of phonemes, as well as with its suprasegmental qualities such as stress, intonation, etc. This is yet another reason why the equation 17 -on = 17 syllables just does not work for haiku in English. 

A Foot-based Template

    Music is not the universal language; rhythm is.
      Plenty of people are tone-deaf, but everyone has a heartbeat.
              - Chico Hamilton, Jazz drummer[25]

Metrical phonology (cf. Hayes 1981, Goldsmith 1990, Kenstowitz 1979), a subfield of current linguistic theory, proposes a rhythmic structure in language made of metrical tiers. As put forth in Prince (1989, p. 50 ff), the lowest tier of this structure is the metrical position (MP) (cf. Halle and Keyser, 1966). The second tier is that of metrical foot (F), made up of two metrical positions, one of which is stronger than the other. Above the foot is the metron (D), made up of two feet with a similar strong/weak (S/W) pattern:

Figure 2. Hierarchical organization of metrical structure in language (Prince 1989, p. 51, Figs. 5a and b, cf. Also Goldsmith 1990, p. 171).

The assumption here is that all metric structure is essentially binary, although it is admissible for one MP per foot to further break into two sub-positions to account for ternary meters such as anapests. The foot itself, however, forms a temporal unit not unlike a measure in music, and is thought to be a universal phenomenon. The strong/weak alternation stipulated for the foot and metron tiers explains, for example, the phenomenon of stress clash in English, (where thirt'een shifts to th'irteen 'men, and Mississ'ippi shifts to M'ississippi R'iver).

With its strict moraic phonological structure and regularized timing (Port et al., 1985), Japanese would assumedly be a prime language in which to find evidence of foot structure, and the epithetic haiku form a prime context in which to look for it. And indeed, Japanese verse theorists of recent years have come to agree that the 17-on haiku is actually based on a 24-beat template which divides into 3 lines of 8 beats each, including 3, 1, and 3 silent beats, respectively. Figure 3 shows an example of (1) Basho's frog pond haiku and (2) a famed traffic aphorism (hyogo) superimposed on this template. Comparing the two, we can see how this template can handle not only 5-7-5 haiku, but haiku with a greater number of -on as well (specifically, 6-8-5 in the case of (2):

            Figure 3. Japanese Haiku Metrical Template: 24 mora in 3 segments of 4 bi-moraic feet.[26]

We see also from Fig. 3 that the 8 beats in the segment are arranged into 4 sets of two beats each. Paralleling the discussion above in metrical phonology, we can consider these beats as similiar to metrical positions (or mora), thus we end up with 3 segments of 4 bimoraic feet, or 12 feet altogether for a full haiku. However, the analogy may be extended further, as these 12 feet may further be hierarchically arranged into binary sets of 6 metrons and 3 segments, as follows:

                              Figure 4. Hierarchical organization proposed for Japanese haiku. (Sakano 1996, p. 132)

The hierarchy shown in Figure 4 looks suspiciously like the one in Figure 2. In other words, we have a clear binary pattern which, except for the stress patterns, would seem to perfectly match the proposals of the metrical phonologists. As for stress, the concept of strong/weak itself is lacking in Japanese prosody; however, we do find evidence in haiku for the prominence of the first position of each foot, in that (1) morophological markers (such as kireji) tend to occur or end on these positions and (2) pauses do not occur in such positions (except in cases of syncopation, see below). 

Variations in the base template rhythm

Henderson (1971) has gone so far as to say the rigid 17-syllable requirement for haiku does not exist in Japanese. Indeed, it is true that even traditional Japanese haiku form allows for a certain "aboutness" in number of -on and that variations such as 6-8-5 (=19), 5-7-4 (=16), etc., are often found. It is telling that such variations are called hachou or broken rhythm (rather than broken form) in Japanese, implying an underlying conscious or unconscious conceptualization of haiku as a meter rather than a form. Within the 8-8-8 template, however, such variations are easily accounted for.

However, variation in haiku metrics does not come about solely because of a different number of -on. The internal structure of the haiku itself may lead to other variations of 5-7-5 within the 8-8-8 base template, especially with respect to the placement of silence beats. Such variation usually comes about because the internal structure of the poem may differ from the most common 5-(fragment)/7-5 (phrase) structure, and therefore pauses must be rearranged in order to keep the semantic units intact. Especially, it is common to find haiku that have natural breaks at about the middle of the haiku, on the 8th or 9th beat. Even if they differ from the base, however, silence beats generally occur only in a weak MP position (i.e. in the second half of the foot) or at the beginning of a segment, creating syncopation. In the following sections, both hachou alternative-phrasing and silence beats will be discussed in some detail. 

Hachou (Broken Rhythm)

Broken rhythm or hachou in Japanese is the product of a haiku with either more (jiamari or amariji) or fewer (=jitarazu) than 17 -on. The 24-mora template provides a structural basis for jiamari, the little "extra" mora often found in most of the classical masters' works. One example Is (Sakano 1996, p. 140):

yuki wa shizukani yutakani hayashi kabaneshitsu   - Hakyo

The first segment of this haiku contains 7 beats rather than 5, which are grammatically and semantically connected with the second segment. Thus Sakano recommends the following as the preferred reading:

We will see later that this recommended reading, with its initial syncopation, did not prove to be the most popular reading in our pilot tests; nevertheless the initial extra beats do throw a monkey wrench into the works, which is best taken care of by utilizing the available silence beats in some manner.

In cases of jitarazu, or haiku having fewer than seventeen syllables, Sakano suggests two methods of compensation.  First, a mora can "double" once or more in order to fill out the missing beat(s), as in the following (Sakano 1996, p. 144):

taka no su ya taikyo ni kiyomeru hi (--) hitotsu       - [Hashimoto] Keiji

In this haiku, the final hi of hi hitotsu naturally lengthens to fill out the second beat of the final segment. It is not shown in writing, but rather felt intuitively through the structure of the overall haiku. From our experimental results, we will see that such -on lengthening over two or even three beats seems to be common, not only for jitarazu haiku, but for the final fifth beats of the first and third segments of 17-syllable haiku as well. Alternatively, the silence between the -on may be lengthened, as in the first segment of this tanka (Sakano 1996, p. 144):

Fu-i ni kokoro wa kanashi yuki furi hairu
furusato no ike ni mogoi wa ikimu
                             fu-i ni
X X X X X / kokoro wa kanashi               -Miya Shuji

Here, the added length between the fu-i ni (suddenly) and kokoro wa kanashi (heart is sad) may serve to stress the shock of the suddenness expressed in the first segment. This added space is not necessarily actually preserved in terms of temporality when the haiku is read aloud, but is probably more felt through the absence of the final two spoken beats in the first segment. 

Pauses and Syncopation

The superimposition of a 17-mora verse on a 24-mora template means that 7 mora are left "floating" to provide ma, or meaningful pauses (to simplify this complex psycho-aesthetic term). In principle, six of these are to be found in the first and last segments, and the 7th, in the middle segment. For the sake of simplicity and ease of analysis, we will limit our discussion of pauses to this midline beat.

Conceivably, the midline beat could occur in eight different positions within the segment. In classical haiku, however, it is said to occur only at the eighth, fourth, or sixth position in descending order of frequency (Sakano 1996, p. 172). This is to be expected, as each of these positions falls on a weak beat in the foot structure of the template.  Similarly, classical five-beat segments include two silent beats at the 7th and 8th beats of the first segment, and silence either at the 6th beat (i.e. the last three beats are silent), or the fourth beat. Both of these patterns, too, respect the foot structure, falling only on the weak beat of each foot. Note that we do not find pauses on the second beat of the first foot, even though such a placement would not violate the foot structure. This reflects the tendency to avoid a stranded single mora (Sakano 1996, p. 172), which in itself provides further evidence for an underlying foot structure in general in Japanese.

Where there are feet, and pauses, there is also the possibility of syncopation. We have already seen examples of syncopation in haiku with jiamari in the previous section. The question of syncopation use in haiku, however, often seems to be one of fashion and personal taste. For example, Basho's "frogpond" poem is found superimposed on the 8-8-8 template in two different ways in the literature[27]

Although both conform nicely to the 24-mora template, the two versions represent different interpretations of how these haiku might be spoken. The pauses in the former version serve to support the internal foot structure of the template. On the other hand, the latter example includes syncopation, which serves to keep the 12-on phrase together. These two versions may be interpreted as reflecting a choice on the part of the reader to preserve the foot structure of the haiku over its internal semantic structure, or vice versa. They also may reflect the desire to preserve word pitch accents, which are known to differ in different regions of Japan.

It is well known that most haiku are composed of a fragment, or shorter portion giving an idea or concept, and a phrase, or longer portion concerned with some aspect of the fragment. According to Imaoka (1999), "classic Japanese haiku are composed of two major parts of varying lengths, such as 5-12, 12-5, 8-9, 9-8, 7-10, and 10-7, in the generally decreasing order of prevalence...".  The commonest break by far between fragment and phrase occurs at 5-12, coinciding with the end of the first segment. Breaks after the 12th mora, too, coincide with the end of the second segment.

Note that the next two sets of numbers (8-9, 9-8) place the break between fragment and phrase exactly at the middle of the second segment. Thus, we may conceive of a kind of caesura occurring at exactly half the length of the poem. Haiku that break into a fragment/phrase of 8 and 9 -on respectively do so at this caesura, for example:

ware-to kite asobe-ya (9) oya-no nai suzume (8)
    (come play with me -- you motherless sparrow)   - Issa

uguisu-no naku-ya(8) chiisaki kuchi akete (9) 
         (uguisu singing -- with the small mouth open)   - Buson[28]

This midpoint caesura was noted by Sakano (1999, p. 132) as a hidden element providing symmetry within the haiku, as follows:


The symmetry here provides a mirror image of 5+4||3+5 or 5+3||4+5 respectively. It can be argued that this mirror-image symmetry, along with the existence of the 8-8-8 template, gives the overall poem a sense of closure within its "incompleteness." 

In Search of the Japanese Template: Some Experimental Observations

To understand more about the Japanese haiku metrical template and how it is internalized in the language, we plan to conduct a large-scale experiment in the future. Preliminary pilot studies, however, have in themselves shown some useful results, which will be summarized in this section. 

Pilot Test 1: Metronome-timed Readings of Classical Japanese Haiku with a 5-12 Structure

The six haiku, shown below, with an internal structure of 5-12 morae (in other words, a kireji was present at the end of each first segment) were given to 7 Japanese (5 female, 2 male) between the ages of 14 and 23. After studying the meaning of these haiku and practicing to their satisfaction, the subjects were asked to read the haiku into a computer microphone. They were then instructed to use a metronome placed at a comfortable speed and read the same haiku again using the metronome as they felt natural. This procedure resulted in a set of 84 readings (12 for each subject). The readings were then measured and analyzed by computer using a Goldwave digital audio editor.[29]  

Pilot Test 1: The Six Haiku (from Blyth, Haiku, 1952)

1. furu-ike ya kawazu tobikomu mizu no oto   - Basho (vol. 2, p. 246)
         The old pond; / A frog jumps in,-- / The sound of the water.

2. Yoku mireba nazuna hana saku kakine kana   - Basho (vol. 2, p. ix)
        Looking carefully, / A shepherd's-purse is blooming / Under the hedge.

3. Asagao ni tsurube torarete morai-mizu   - Chiyo-ni (vol. 1, p. 219)
        The well-bucket / Having been taken by the morning-glory, / I borrow water.

4. [w]Ochi kochi ni taki no oto kiku wakaba kana    - Buson (vol. 1, p. 245)
        Listening to waterfalls / From here and there: / The young leaves!

5. Sabishisa ni meshi wo kuu nari aki no kaze   - Issa (vol. 1, p. 411)
        Eating a meal / In loneliness, / The autumn wind blowing.

6. Yuku ware ni todomaru nare ni aki futatsu   - Shiki (vol.2, p. 97)
        I going, / You remaining,-- / Two autumns.


Overall Lengths

Table 1 shows the relative lengths of the metronome and non-metronome files for each subject. It can be seen that each haiku tended to become longer with each reading. With the exception of the first haiku (3.49), however, we find the average lengths to be remarkably similar, between 3.62 and 3.73. In other words, although individual variation in speed was great between the subjects (with F.U. being especially fast), there was no large variation between the speeds of the individual haiku themselves. This indicates that the overall haiku tempo may be felt to be more or less invariant. (The relatively faster tempo used in the first haiku may be attributed to the fact that this particular haiku is well known to most Japanese, and this familiarity could have led to a high level of comfort and hence higher speed. It may also be due to initial nervousness regarding the experiment itself).

Table 1. Haiku lengths for each subject with and without metronome (measured in seconds).

Table 1 also shows that the average lengths of haiku readings without metronome (3.31) were shorter than those with the metronome (3.96), indicating that the task of "timing" the haiku slowed the subjects down considerably. Just how considerably can be seen from the percentages of non-metronome reading lengths with respect to the timed ones for each reader, which varied from 74% to 96% (shown in the final column of Table 1).

The metronome speeds chosen seemed to have a direct effect on these percentages: three of the subjects chose speeds around 130 and spoke their haiku with a 2/2 foot-based rhythm, whereas another three chose speeds around 90 and spoke their haiku with a 4/4 metron-based rhythm. Finally, F.U. alone read haiku with a foot-based rhythm and metronome speeds averaging 179, exactly twice as fast as the metron-based rhythm group. It is interesting to note that both F.U. and the subjects who chose a 4/4 metron-based beat rhythm (=lower metronome speed, ca. 90) over a 2/2 foot-based rhythm (=higher metronome speed, ca. 130) came closer to imitating their original speeds (86% to 96%) than did the latter (74% to 81%).

In theory, the 24-beat 8-8-8 template should yield a tempo which equally divides into both 8 and 24 -- either a foot-based tempo for 12 metronome clicks (or 11, with the last beat on silence after the haiku ends), or a metron-based tempo with 6 metronome clicks altogether.[30] This is exactly the number of clicks we find consistently from our pilot tests: = 11 for the foot-based readers, and 6 for the metron-based readers.

Interestingly there were several cases of -on lengthening to fill out the 8-8-8 pattern. The former group often lengthened their final -on to meet the final 12th beat (such cases are marked with a final (x) in Table 4). In addition, we observed three occasions where metron-based readers got ahead of the beat, but in each of these cases they used a similar compensatory lengthening technique to fill in the space up to the final 6th beat.

The only exception to this regularity in number of metronome clicks was F.U., who started out with 11 beats, but consistently sped up until she got to 8. It is suspected that F.U. tended to ignore the metronome rather than work with it, as we find metronome beats falling between -on as well.  For this reason,  F.U.s results were omitted from further analysis. 

Relative Lengths of Portions of the Haiku to the Whole

In order to judge the relative similarity of the rhythmic patterns with and without the metronome, measurements were taken of two sections of each metronome-timed and untimed reading: (1) the length of the second segment and (2) the length of the initial pause between the first and second segments. These measures are shown in Tables 2 and 3, respectively, as proportions of the overall haiku lengths. Proportions of the untimed readings to the timed readings are shown as well.

                      Table 2.  Average proportions of middle segments to overall haiku length.

If the haiku readings with and without metronome are similarly patterned, we would expect to find that the proportions of the parts (i.e. the second segment and initial pause lengths) are similar to the proportions of the whole. Specifically, our 8-8-8 template indicates that the second segment should take 1/3 or approximately 33.3% of the whole poem.[31]  Table 3 shows that this 33% is exactly the average percentage we find for both the timed (which should be expected) and the untimed haiku. Moreover, the averages for each haiku deviate little from this norm, ranging from 31% to 37%. For individual readings, we do find values ranging from 24% to 42%, but of course the more extreme percentages could be to "mistimings" on the part of the readers.  This fact that the untimed haiku show the same patterning as the timed haiku is especially revealing, and indicates the regularity and consistency of the overall rhythm.

      Table 3. Average proprtions of initial segment pauses to overall haiku length.

Likewise, the initial pause in the first segment should most typically extend over three beats, and thus the proportion we would expect would be approximately 3/21, or about 13.6%. As Table 3 shows, we do find a percentage that approximates this number for the timed haiku: 12%. For the untimed haiku, however, the percentage falls to only 10%. There are two main reasons for these lower figures: (1) many readers extended their fifth -on in the first segment at least to the sixth beat (as noted previously, this seems to be common), and (2) there was a slight tendency to jump the gun and to begin the second segment before the beat. The first tendency could be seen especially in the case of M.O., who had recently experienced several hyaku-nin isshu (100 waka) contests in which such elongated readings tend to be the norm. The second tendency was common among many of the subjects, especially for the untimed readings. 

This "pause-shortening phenomenon" indicates that the metricality of Japanese haiku may be more subjectively relative than temporally isochronous, which is exactly what we should expect from a poetic form. It also indicates that variation in the tempo of haiku probably occurs with more frequency at the initial 3-beat pause space than anywhere else. 

Positioning of the beats

The next question we asked was whether the beats fell consistently at similar positions within the haiku as read by different readers. Table 4 compares the placement of beats used by each subject (except F.U., who results were discarded due to the irregularities in tempo mentioned above):

Table 4. Location of metronome beats in the haiku (pilot study 1).
( X = beats used by both the metron- and foot-timed readers; x = beats used by foot-timed readers alone.)

Depending on the haiku, we see quite a bit of variation between readers. This variation, however, tends to occur mainly in the second segment, with the metrical phrasing of the first and third segments being remarkably uniform. Specifically, we find three main patternings of the second segment: (A), with the pause falling on the final beat (=classical);  (B), in which the pause falls on the first beat (=syncopation); and (C), with the pause falling on the fourth beat (=pre-caesura lengthening). These three patterns occur in 6, 5 and 4 of the haiku respectively.

In addition, two aberrant rhythms were found; two readings are marked pattern (D), indicating that the reader got ahead of the beat. This phenomenon occurred once for M.O. (haiku 4), in which the first mora of the second segment was pronounced before the beat, and once with Y.U., whose second segment did not contain a pause beat. The end result was that both of these haiku ended before the 21st beat of the template, but in both cases the final syllable was extended to cover this beat.

Finally, we see a case of miscalculation and adjustment of timing in pattern (E). Here, the actual rhythm used by M.I. was as follows:


In other words, two -on ended up sharing the same metrical position both with /tsuru/ and /(#)to/, where # indicates a slight pause in the rhythm, almost as if the following /t/ of to had been geminated (i.e. a little tsu was placed before the /t/).  This "beat-splitting", which parallels the MP splitting phenomenon in metrical phonology discussed earlier, seems to occur as a repair strategy. M.I. had used the syncopated pattern A for the previous two readings, but this pattern places the /ru/ of tsurube in a strong metrical position, which may have sounded odd to her ears. To repair this situation, the /ru/ shifted back to share the preceding weak beat with /tsu/ placing the word-final -on /be/ in an appropriate foot-initial position. This |da da DA| pattern was then repeated in the next foot by inserting a weak mora tsu, perhaps to achieve a rhythmical symmetry.

Before we leave this section, it should be noted that haiku 6 was the only one which showed a single patterning (B). This is due to the fact that the internal structure of the second segment of this haiku is 4:3 (to do ma ru   na re ni) rather than 3:4 as with the other 5 haiku. This internal structure requires the fourth -on /ru/ to fall on a foot-final position, resulting in only two possible meters:


Thus neither pattern (A) nor (C) could have been employed for this haiku. 

Pilot Test 2: Metronome-timed Readings of Classical Japanese Haiku with a Non-5-12 Structure

To find out whether variation is naturally felt and expressed within the 8-8-8 template, the following four irregular (in the sense that they do not follow the most typical 5-12 pattern) haiku were selected to be read: 

(A)  ware to kite asobeya (9) oya no nai suzume (8)  - Issa

(B)  neko no meshi shouban suru ya (12) suzume no ko(5)  - Issa

(C)  nagorinagori (6) chirumade wa mizu ume no hana (12)  - Chiyo-ni

(D)  yuki wa shizuka ni (7) yutaka ni hayashi kabaneshitsu (12)  - Hakyo

(A) and (B), taken from Imaoka (1999), have breaks in places other than the 5th mora; specifically after the 9th and 12th mora respectively, as noted. (C) and (D), on the other hand, were chosen for their extra amariji morae in the first segment. Also, as discussed above, (D) prefers a reading of the first and second segments without a pause; thus, the first segment should either begin after the beat or contain an extra internal beat. 

Four subjects (2 male, 2 female) were recorded using a method similar to the first test, and results can be seen in Table 5. 

Table 5. Location of metronome beats in the haiku (pilot study 2).
( X = beats used by both the metron- and foot-timed readers;  x = beats used by foot-timed readers alone.)

Interestingly, none of the subjects were able to break away from the original 5-12 pattern for the first two haiku (A) and (B), even with the obviousness of the kireji /ya/. This is a somewhat puzzling result, as the template should allow more metric variation. However, upon explanation of the difference, subjects readily agreed with the interpretations and imitated the rephrasing easily. One subject remarked that such phrasing would not occur to someone without a formal study of haiku, whereas another commented that the 5-12 pattern was so deeply ingrained in Japanese prosody that it was difficult to imagine breaking out of the mold.

On the other hand, results for haiku (C) and (D) were varied. In haiku (C), there is an extra beat in the first segment complicated by the fact that the internal structure is 3/3 (nagori is repeated twice). Here, we find Y.U. using a syncopated first beat and T.A. pausing after the first nagori. Both of these strategies have the effect of placing the second na at the beginning of the foot, which feels intuitively "right." The other two readers chose rather to add the sixth beat at the end of the five beat segment.

 Haiku (C) shows variation in the first segment only, again because the internal structure of the second segment is 4:3 rather than 3:4, as with Haiku 6 in Pilot Test 1. Haiku (D), however, shows variation both in the first segment and the second segment. With respect to the first segment, it should be noted that none of the subjects chose an initial syncopation, as predicted by Sakano. Rather, 3 of the 4 paused after the 3rd -on /wa/, whereas M.O. ran the second segment into the first, delaying the problem pause until the very end of the haiku. Notably, however, because this reading strategy had the effect of making the haiku shorter than 6 beats, he lengthened the final -on /tsu/ to reach the sixth beat. 

Implications for English Haiku


Before proposing a metrical means of emulating the Japanese haiku in English based upon the previous experimental results, we would like to discuss some of the current received wisdom concerning how best to emulate the Japanese haiku. Because questions of emulation rest upon descriptive analysis (if not definitions) of the Japanese haiku, these descriptions will also be critiqued.

The most comprehensive information along these lines in a widely published volume is to be found in Higginson's discussions of traditional haiku form in The Haiku Handbook. These descriptions will provide the basis for our analysis. 

Critique of Higginson's Description of a Traditional Form for Japanese Haiku

It should be mentioned at the outset that this critique of Haiku Handbook is not intended as a disparagement. This book remains one of the best broad introductions to haiku. In fact, it is due to the book's popularity and resulting impact that a critique seems especially necessary. Higginson summarized the traditional form of the Japanese haiku as:

1. A Japanese haiku . . . usually has two rhythmical units, one of about twelve [-on] and one of about five, the break between them often marked with . . . a kireji.

This description reduces the reader's perception of rhythm in Japanese haiku to two separated rhythmical units, if we focus on what Higginson alludes to as the "usual" haiku. Actually, as we have seen, the Japanese haiku has two phrasal elements, contained in a tripartite metrical structure of three segments, each segment containing two metrons, four bi-moraic feet, and eight metrical (moraic) placements. In the sense that rhythm is "movement or fluctuation marked by the regular recurrence or natural flow of related elements,"[32] the haiku is all of a piece, a dynamical rhythmic production, with a sense of beginning (first metrical line), middle (second metrical line), and end (last metrical line). Minimally, three metrical phrases are evident.

As for where the two lexical phrases are parsed, the most common break is after the first five morae, that is, a 5-12 phrasing. But importantly, the rhythm of this phrasing can change. We need to remember that the 5 spoken sounds (-on) occur in the first metrical line of eight morae. The second phrase may begin at the first mora of the second segment, or it may not. It could begin, for instance, in the second moraic position of the second segment. In other words, there can be syncopation occurring metrically in the second phrase. Syntactical and grammatical elements all affect the overall rhythm. Thus, having a 5-12 pattern does not lead to the implication that there is any single unitary rhythm divined by that phrasal arrangement. There are also additional phrasings in the haiku. The second most common is the 12-5 phrasing, followed by the 8-9 or 9-8 phrasings, with the kireji placed in the middle of the second segment. Many traditional haiku utilize these and other, less common, phrasings and internal rhythmic structures. Why should we restrict our appreciation of the haiku to only the most common phrasing? We might rather recognize a multitude of both subtle and dramatic formal variations.[33] In any case, all rhythmical elements, from the bi-moraic foot to the phrase, are at all times relating to the fundamental metrical template. For instance, we have found caesura between the second and third metrical lines, with haiku that contain a final 12 -on phrase spanning these two lines. There are a number of additional examples. The main point of contention here is that the metrical template affects the spoken rhythm; there are many more rhythmic "units" than two phrasal units; phrasal units come in several shapes and sizes, not merely 5/12; and, the Japanese haiku has many levels of rhythmic symmetry, asymmetry, and interaction.

2. Since the break between the two rhythmical units is equally likely to occur after the first five [-on] or after the first twelve, the normal rhythm  . . . in Japanese is five, seven, and five [-on].

As mentioned above, statistically, there is a hierarchy of likelihood of where phrases break. The most likely break is after 5-on, and next most likely after 12-on, next at 8-9/9-8, and etcetera, down the range of statistical likelihood, which ends with the exceptional or dramatic departure. It remains unclear what "the normal rhythm" is; our data suggests at least three common rhythmic patterns even within the 5-7-5 -on haiku on the part of different speakers--divergent rhythmic readings also occur within the same haiku (see Tables 4 and 5). Secondly, the basis of the rhythmic pattern is not syllabically determined. The rhythmic patterns are first determined by the metrical template, which is more fundamental than either the total number of -on, or where the phrase breaks occur. For instance, haiku having 7-7-5 -on or 6-7-5 -on also map themselves onto the metrical template. Third, haiku rhythm has much to do with the way syntactical and semantic elements of the haiku interact with the metrical pulse. Our conclusion is that haiku rhythm is not first determined by 5-7-5 at all; rather, the metrical template is the more fundamental feature. Indeed, 5-7-5 is not a necessity for haiku, though it is a highly selected choice, and accords well with the metrical template. Nevertheless, this choice is related with 1) typical language conventions, and 2) an individual poet's or 3) haiku school's aesthetic--and is not a formal rule.

3. The form of traditional haiku originated in the incomplete opening stanza of a longer poem; the haiku form is therefore rhythmically incomplete.

The form of the traditional haiku originated as the opening stanza of a longer poem. That is, the typical 5-7-5 -on occurring in a moraic structure of 8-8-8 was usually linked to a further two lines of 7-7 -on contained in two metrical lines of 8-8, in waka and renga.[34] The basic pattern is ancient, probably pre-historic, and can be found in the "call and response" pattern of the katauta, as Yasuda (1957, p. 109) points out. At issue here is the idea of rhythmic incompleteness, functionally related to a felt-need for rhythmic completion. The question is, does this rhythmic incompleteness really exist, and if so, is it a function related to an absence of metrical closure? It may be true, in terms of literary history, that the haiku is a rhythmically incomplete form. Nevertheless, the ubiquity of the haiku as an autonomous form in Japanese culture, and its use as a mnemonic device over the past century, points to the necessity for either clarification or some alteration of the above view. The tripartite metrical structure of the haiku template has rhythmic closure. Due to the long-term popularity of the haiku (or 5-7-5 pattern), the sense of rhythmic incompleteness may have become highly mediated. Higginson's lack of clarification may lead one to mistakenly assume that "rhythmic incompleteness" equates with a lack of metrical closure.

Metrical closure is also indicated by formal symmetries. In the (most common) 5-12 phrasing, the 3 silent beats at the end of the first (introductory) metrical line are echoed in the last three silent beats of the third line (resolution). As well, there is caesura at the end of the second bi-moraic foot of the second metrical line, that is, exactly bisecting the metrical template; the haiku is thus metrically symmetrical from its midpoint: a 12-12 moraic pattern.

In any case, there seems not to be a pregnant rhythmic incompleteness emanating from the 5-7-5-on form as-such, these days. Arguably, aphorisms such as: "at the red light/crossing all-together/is fearless" (aka shingou/mina de watareba/kowakunai) have rhythmic closure. A case for incompleteness in haiku might better be made by examining the semantic relationship of the phrases (especially the final phrase) in relation to the metrical template. Especially striking is that haiku (and all 5-7-5-approximate poetic structures, as aka shingou, above), end with silent beats, usually three. What is striking is that these silent beats are part of the formal structure of the poem, defined by the metrical verity. It is a unique outgrowth of the moraic nature of Japanese, and difficult, if not impossible, to replicate in English. Many traditional haiku place a final kireji-type expression at the end of the haiku, (kana, etc.). So, one could say that it is not so much rhythmic incompleteness, as a dissolving of the poem as an exhalation into space that lends haiku its unique resolution--evoked through the complex of the poem's entire sensibility and the formal resolution in silence that is uniquely present within the metrical template. Our data would indicate that the 8-8-8 metrical structure has metrical closure, and 'missing' metrical lines are not evoked as a matter of course.

4. Haiku often omit features of normal grammar . . . (p. 104).

The Japanese haiku admits a number of further possibilities in terms of language. Imaoka (1999) states that "syntactic relationships, [and] word units [can] become independent and can be moved about more freely within a sentence or a clause without affecting its meaning" and "there are [many] more places [than in English] where a Japanese phrase can be divided without disrupting its meaning." A further analysis is off our topic, but point #4, above, begs expansion. 

Critique of Higginson's Description  of "A Traditional Form" for English Haiku

Continuing on, Higginson develops a set of four prescriptive dicta for emulating the Japanese haiku in English, based upon his analysis of the Japanese haiku and an application of stylistics drawn from the English poetic tradition. The first is:

1. For haiku in English an overall form consisting of seven accented syllables, plus unaccented syllables up to a total of about twelve, would yield a rhythmical structure native to English and at the same time approximate the duration of traditional Japanese haiku. A major grammatical pause between the second and third or fifth and sixth accented syllables would provide the sense of division created by the Japanese kireji.

Any discussion of how to formally describe the English haiku, especially in terms of how it emulates, or does not emulate, the Japanese haiku, is bound to arouse disagreements. We know there are still those for whom English haiku will only ever be 5-7-5 English syllables. Higginson's suggestion, above, is certainly an improvement over the 5-7-5 prescription. Nevertheless, many of the acclaimed English haiku presented earlier in this paper do not fit into Higginson's pattern (depending upon how one scans the accented syllables). In our presented examples, haiku have accented syllables from a minimum of 1-1-1 for a total of 3 in Tripi (Ex. 13), to Wright's haiku, which can be read as 3-3-2, with a total of eight accents. So, there is a range of 3--8 accented syllables in our example haiku, with a total of 6 accents (2-2-2 or 2-1-3 or 3-2-1 in our examples), predominating, slightly, over totals of 7 accents. In either case, we suspect that poets are not composing haiku by counting strong and weak syllable accents as a formal principle.

Hopefully, the reader recognizes the success of English haiku through the excellence of our presented examples (see Section 3). Most haiku have one, two, or three accented syllables per line. It is hard to have less than one or more than four accents per line; more than nine or less than three accented beats, total. Metrical divisions are usually limited to one main division, to accord with the Japanese kireji, and this division is most often at the end of the first or second poetic line (though a mid-second line pause crops up here and there in some haiku). Total syllable-count is limited to emulate the Japanese form, and most haiku (93+%) have a three-line lineation, thus intuitively (or serendipitously) emulating the three-line metrical template of the Japanese. Which brings us to:

2. While Japanese are used to reading texts in which rhythms are not visually identified, the Western notion of a printed poem-text incorporates the idea of a line of type equaling a rhythmical unit, or verse-line. Therefore, a three-line structure of two, three, and two accented syllables, respectively, would establish rhythmical proportions similar to those of traditional Japanese haiku.

We have already addressed the problems of restriction to 7 stressed syllables, above. A 2-3-2 approach only compounds these problems. For example, Wright (Ex. 3) might have a 2-3-2 pattern of accented syllables, if we leave off the accent (demote the second syllable) in "enough" in the first line; otherwise we have a 3-3-2 pattern. Evetts (Ex. 17) could be read in a 2-3-2 pattern, but also in a 2-2-2 pattern. Perhaps it is enough to say at the outset that Higginson's mode of emulation for English haiku has not been found especially useful by the English haiku movement. The 2-3-2 pattern as a singular form of emulation impoverishes the Japanese haiku by implication, and is creatively restrictive in English, excluding innumerable free-verse techniques for working with rhythm/cadence.

The metrical template we have described for Japanese haiku provides an alternative for creating an emulative metric structure in English, without restricting it in terms of numbers or stress patterns. It will been seen in the next section that rhythmicities of 6 and 11 or 12 are naturally present in haiku form in English, even in our examples. Such rhymicities however do not necessarily have to coincide with numbers of syllables, stressed or unstressed; they should rather be interpreted as pulsations though the poem; beats of timing, as it were. The fact that many non 5-7-5 haiku do show syllable counts nearing one of these numbers demonstrates that this metricality has been felt intuitively, if not recognized formally, in English haiku.

Higginson makes a useful point here that in English poetry, a line of type equals a rhythmic unit (perhaps with the exception of run-on lines). The Japanese haiku metrical template is composed of three metrical lines, and this makes a clear-cut case for a three-line lineation in English haiku as a normative form of emulation. In terms of lineation, we have to disagree with Sato (1999b, p. 55), who says "I regard the Japanese poem as a one line poem and translate it accordingly." Sato is ignoring the metrical template which is intuitively and unconsciously grasped by the Japanese-native speaker, as our research shows. A singular metrical template does not exist in the English haiku. A line of Japanese haiku poetry and an unpunctuated, normally-spaced sequential-line translation of English haiku are perceptually divergent in terms of how line parsing, on the part of the reader, occurs. To best emulate the Japanese haiku, the foundational metrical structure needs some formal emulation, we would argue. Since this is already being done in virtually all acclaimed haiku composed in English (with successful departures as in Spriggs, Ex. 1), and also in translations, the point seems moot. Nonetheless, to think of the Japanese haiku as similar to a one-line unpunctuated or otherwise metrically indeterminate English poetic form ignores the regular tripartite internal metrics of virtually all Japanese haiku ever written, traditional or modern,[35] and often forces the English native-speaker to perform conceptual gymnastics in discerning the rhythmic parsing of the metrically indeterminate one-liner.

3. Since the most commonly encountered short structure in traditional English poetry is the "heroic couplet" with two five-beat lines, the two-three-two beat structure with a strong grammatical break after the second or fifth beat, as proposed, would yield a sense of rhythmic incompleteness similar to that in Japanese haiku.

As previously mentioned, rhythmic incompleteness in the Japanese haiku is a complex and debatable topic. Likewise, the idea of basing the English haiku on the heroic couplet is debatable. When you break up a single five-beat line into two, it takes on a completely different rhythmic character. Why not put in a vote for the four-beat rhythm? "This is the most common of all the possible rhythmic patterns, if every kind of verse is taken into account. It is the basis of most popular music, including rock and rap, of most folk, broadside, and industrial ballads from the Middle Ages to the twentieth century, of most hymns, of most nursery rhymes, and a great deal of printed poetry. It is popularity is not limited to the English language, either . . ." (Attridge, 1995). In fact, it comes as no surprise that the prevalent 2-2-2 accented-syllable pattern in English haiku successfully aligns with the four-beat rhythm. Pointedly, several rhythmic models and modes are utilized to advantage in English haiku, and there is a problem in prescribing any one exclusive emulatory model or method of composition. An investigation into the scansion of existing English haiku leads to several formal metrical possibilities, each having innumerable possible variations.

4. Grammar should be stripped to the minimum that seems reasonably natural. Complete sentences may or may not occur; articles ("a, an, and the") and prepositions should be used sparingly, but not naturally omitted.

This point, on brevity and concision, finds some resonance in two Imagist maxims: "To produce poetry that is hard and clear, never blurred nor indefinite." And, "Finally, most of us believe that concentration is of the very essence of poetry (Poetry, 1913)." 

A Metrical Approach to English Haiku Based on the Japanese Template: A Musical Analogy

Based on our discussion of the Japanese metrical template and observed experimental results, we propose that the best conceptual means of approaching the problem of emulation of the Japanese haiku is metrical, musical, and analogical. One of the reasons that most modes of emulation of Japanese haiku have been found to be unacceptable to the English haiku world is that fact that they are bound to be based on the lowest level of the metrical grid--i.e. the mora or MP level. Countless discussions of whether or not a syllable equates to a Japanese mora, or more, or less, are fruitless as they do not take into account what is happening at a higher level in the metrical grid.

Instead of looking at the metrics from the bottom level only, then, we propose trying to find a pattern of emulation which will spring from a higher level of the metrical hierarchy. Remember that Japanese haiku are metrically composed of three 8-beat lines which are in turn composed of 4 feet and 2 metrons each. A top-down approach, then, would begin by dividing the whole time-space of the haiku (which has been described as "about a breath") first into 3 metrical lines (here we do not necessarily mean 3 lines on paper), next into 6 metrons, and finally into 12 feet.

Using a musical analogy, each metrical line would represent a measure, so a complete haiku would be composed of three measures (=lines).  A 2/2 or 4/4 time signature at the beginning of these measures would then dictate a reading at either a metron level, for a total of 6 beats, or at a foot level, for a total of 12 beats.  Consider the following :

We can see here that haiku with a lower number of syllables overall "feel natural" using a 2/2 time signature, i.e. a slower beat.  On the other hand, English haiku with a higher number of syllables lend themselves better to a foot level metricality (i.e. 4/4 time).  However, it is possible for a single haiku to work well at either level, e.g.:

It is also possible for a haiku to work best at a line level, i.e. with a single beat per line (1/1 time signature), as follows[36]:

In a haiku which incorporates temporal extension, a fermata(*)[37]can be used, as in:

The meters suggested here should not be confused in any way with traditional verse meters such as iambic pentameter, which are mapped onto each syllable of a poem and leave little room for pauses or variation. The readings given here are only suggested ones, and make use of devices we have already observed in Japanese haiku, such as syncopation and beat-splitting. Naturally, use of such techniques in different parts of the haiku will yield an infinite number of possible readings. In addition, variation in readings may also be achieved through use of fermata-like beat lengthening, caesura accenting and other techniques.

The experimental results found here have another implication for the metricality of haiku in English. It has often been assumed in English haiku that if the second line has 2 more syllables than the other two lines, that it should also have a corresponding increase in length. However, in Japanese haiku the 8-8-8 template dictates that the length of the overall middle line should be exactly the same (i.e. 8 beats) as the first and third lines, and this is just what we have demonstrated experimentally. This means that if emulation of the proportional lengths in Japanese haiku is desired, the second line of an English haiku should be felt as being metrically equivalent to the first and third lines (which include final pauses) even though the actual number of words/syllables of the line may be longer.

In conclusion, our top-down approach allows us to hypothesize that haiku readability--at (at least) the first two levels of our metrical hierarchy --segment and metron--may form a natural and universal basis for overall haiku form. Furthermore, traditional Japanese haiku must also be readable at the third, or foot level--in other words, the foot level is language specific. Applying this requirement is then unnecessary for the English-language haiku. 


In this paper we have mainly been concerned with issues of Japanese haiku form, and the manner in which English haiku emulate this form. The apparent dissimilitude of the two languages, combined with the dissimilitude of the highly regular syllabic quotient in the Japanese haiku vs. the free-verse syllabic, accentual, and verse-line variability in the English haiku, has presented a scenario in which commentators tend to view the two traditions as formally divergent or diverging. While form is not the only issue, it is, along with the use of kigo, a signal issue both obvious and troublesome. Some would say irreconcilable. We hope that the above views are worthy of reconsideration in the light of a metrical approach to haiku--an approach that may have a further application in the many languages in which haiku are now being written.

To our knowledge, a demonstration and application of Japanese haiku metrics has not heretofore been presented in English. It may perhaps be only through a metrical analysis that formal and structural similarities between the English and Japanese forms become evident and compatible. Because Japanese morae are so unlike English syllables, any formal rule for versification that remains at the lowest metrically-hierarchical level is bound to be misleading, reductive, and restrictive in its treatment of haiku form, whichever cultural form you are looking at. A metrical approach seems an obvious boon for Japanese haiku study on the part of Westerners; it may seem a less palatable approach as a means of English free-verse poetry analysis. In fact, within the science of English versification, "free verse" is defined as "non-metrical verse." But what the versification studies are really saying is that free verse is not regularly metrical when compared to the regulated metrical verse forms we find historically, that is, the historic cannon of English poetics.[38] There is a world of difference between "non-metrical" and "not regularly metrical." "Non-metrical" in its bald sense, is a linguistic oxymoron. All language has meter. Free-verse poetry is not regularly metrical only at the level of the syllable--the basis of traditional verse analysis. We have found that English haiku, perhaps due to its brevity, combined with aesthetic and semantic verities, is amenable to a straightforward metrical treatment, but only at a metrical level hierarchically higher than where any single syllable might be placed.[39]

It is at this higher-order metrical level that formal emulation becomes a reasonable prospect. For those who must have some sort of singular equation between Japanese morae and English syllables, a metrical approach as we suggest will never work, just as for those who demand an English haiku of 5-7-5 syllables, a prescription of "about 11-12 syllables" will never work. One must keep in mind that the higher-order metrical approach in English does not create identities between syllables and metrical pulses. It is a musically analogical approach that finds the typically varied readings of English haiku to have one of several similar temporal structures and metrical patterns (and these patterns are yet to be fully explicated). One can make the same statement about the Japanese haiku, though readings in Japanese are, taken as a whole, less varied than English readings. The metrical pattern is flexible, and is cognate with the musical measure and musical time signatures.

Our approach to metrical emulation first began with an intuition. Listening to modern English haiku and listening to Japanese haiku, there are, of course many obvious differences. And yet there appears to exist a metrical character with a sense of congruency--not perhaps in every case and for every reading (this is yet to be determined), but for many. It has not been our intention to prove a metrical verity for English haiku, nor to prescribe rules for formal emulation. It is unclear at present how definitive such analyses and rules can hope to be. Nonetheless, by demonstrating a paradigm in which many acclaimed English haiku already metrically emulate Japanese haiku, and by providing a means of appreciating the formal complexity that is the music of the Japanese haiku, it is our hope that both English and Japanese haiku literary cultures will benefit. 


We would like to gratefully acknowledge Professor Ryoji Matsuno (Kumamoto Prefectural University), who devoted many hours to authoring the metronome/voice recording software utilized in our data collection. In addition, we would like to thank reviewers Professor Masahiro Hori (Kumamoto Gakuen University) and Professor Sadahiro Kumamoto (Kumamoto University) for their helpful comments and suggestions. Any errors remain our own. 


American Heritage Dictionary.(1992). New York: Houghton Mifflin.

Anakiev, D. and Kacian, J., Editors. (1999). Knots: The Anthology of Southeastern European Haiku Poetry. Winchester, VA: Red Moon Press, dist.

Attridge, D. (1995). Poetic Rhythm. Cambridge: Cambridge University Press.

Blyth, R. H. (1952). Haiku. vols. 1-4. Tokyo: Tuttle.

Collier, G. L. and Wright, C.E. (1995). Temporal Rescaling of Simple and Complex Ratios in Rhythmic Tapping. Journal of Experimental Psychology: Human Perception and Performance, (Vol 21, No. 3), 602-627.

Cummins, F. and Port, R. (1998). Rhythmic Constraints on Stress Timing in English. Journal of Phonetics 26, 145-171.

Gilbert, R. (1999). Stalking the Wild Onji: The Search for Current Linguistic Terms Used in Japanese Poetry Circles. Language Issues: Journal of the Foreign Language Education Center, (Vol. 5, No. 1). Prefectural University of Kumamoto, Kumamoto, Japan. Also Frogpond: Journal of the Haiku Society of America, XXII: Supplement, 7-32.

Gilbert, R. (2000). Universalism vs. Particularism in International Haiku. Haiku Novine, March Issue. Nis: Yugoslavia.

Goldsmith, J.A. (1990). Autosegmental and Metrical Phonology. Cornwall: Basil Blackwell.

Gurga, L. (1997). American Haiku Today. Frogpond: Journal of the Haiku Society of America, XX: Supplement, 33-34.

A Haiku Path. (1994). New York: Haiku Society of America, Inc.

Halle, M. and Keyser, S.J . (1966). Chaucer and the study of Prosody. College English 26, 187-219.

Han, M. S. (1994). Acoustic manifestations of mora timing in Japanese, J. Acoust. Soc. Am. 96 (1), July, 73-82.

Henderson, H. G. (1971). [Letter to Anita Virgil, dated May 5, 1971] In A Haiku Path, (1994), 46. New York: Haiku Society of America, Inc.

Henderson, H. G. (1967). Haiku in English. Tokyo: Tuttle.

Higginson, W. (1996). Haiku World: An International Poetry Almanac. Tokyo: Kodansha.

Higginson, W. & Harter, P. (1985). The Haiku Handbook: How to Write, Share, and Teach Haiku. Tokyo: Kodansha.

Imaoka, K. (1999). Forms in English Haiku. Manuscript, available at: Previously published in Woodnotes Haiku Journal (1996?).

Jyouo, H. (1989). nihonjin no nihongo shirazu [What Japanese People Don"t Know About Japanese]. Tokyo: Arc Books.

Kacian, J. (1999). Haiku and Music in the Western Tradition. Frogpond: Journal of the Haiku Society of America, XXII: Supplement, 33-41.

Kacian, J., Editor. (1999). Snow on the Water: The Red Moon Anthology of English-Language Haiku. Winchester, VA: Red Moon Press.

Kacian, J., Editor. (1998). Red Moon Anthology 1997. Winchester, VA: Red Moon Press.

Kacian, J., Editor. (1997). Red Moon Anthology 1996. Winchester, VA: Red Moon Press.

Katada, F. (1990).  On the Representation of Moras: Evidence from a Language Game. Linguistic Inquiry, 21:4, 641-646.

Kiparsky, P. and Youmans, G., Editors. (1989). Phonetics and Phonology: Rhythm and Meter. San Diego: Academic Press.

Lamb, E. S. (1994). Haiku in English to 1978. In A Haiku Path, pages 3-21. New York: Haiku Society of America, Inc.

Lehiste, I. (1990). Phonetic investigation of metrical structure in orally produced poetry, Journal of Phonetics 18, 123-133.

Lynch, T. (1989). An Original Relation to The Universe: Emersonian Poetics of Immanence and Contemporary American Haiku. Unpublished Ph.D. dissertation, University of Oregon.

Poetry: A Magazine of Verse. (1913). Harriet Monroe, ed. & copyright. March Issue. Chicago.

Port, R. F., Cummins, F. & Glasser. (1995). A Dynamic Approach to Rhythm in Language: Toward a Temporal Phonology. Technical Report 150, Indiana University Cognitive Science Program, Bloomington, IN.

Port, R. F., Cummins, F. & McAuley, J. D. (1995). Naive Time, Temporal Patterns and Human Audition. In Port, R. F. & van Gelder, T., Editors, Mind As Motion, pp. 339-437. MIT Press, Cambridge, MA.

Port, R. F., Dalby, J. and O'dell, M. (1985). Evidence for mora timing in Japanese, J. Acoust. Soc. Am. 81 (5), 1574-1585.

Pound, E. (1913). In A Station Of The Metro. In Poetry: A Magazine of Verse. Harriet Monroe, ed. & copyright. March Issue. Chicago.

Prince, A. (1989). Metrical Forms. In Kiparsky, P. and Youmans, G., eds. Phonetics and Phonology: Rhythm and Meter. San Diego: Academic Press, 45-79.

Sakano, N. (1996). shichi go chyou no nazo wo toku. [The Key to the Riddle of 7-5]. Tokyo: Taishuukan Shoten.

Sato, H. (1999a). The HSA Definitions Reconsidered. Frogpond: Journal of the Haiku Society of America, XXII:3, 71-73.

Sato, H. (1999b). Haiku & the Agonies of Translation. Frogpond: Journal of the Haiku Society of America, XXII: Supplement, 50-66.

Swede, G. (1997). A History of the English Haiku. Haiku Canada Newsletter, vol. 10, no. 2/3.

Takiguchi, S. (1999). Haiku As A World Phenomenon. Ginyu, Issue 5. Tokyo.

Van den Heuval, C. (1974). Haiku Anthology. New York: Anchor/Doubleday.

Van den Heuval, C. (1986). Haiku Anthology, 2nd ed. New York: Simon & Schuster.

Van den Heuval, C. (1999). Haiku Anthology, 3rd ed. New York: Norton.

Merriam-Webster's Collegiate Dictionary, 10th Edition. (1999). Springfield, MA: Merriam-Webster Inc.

Wilmot, R. (1981). The Structural Dynamics of Haiku. Frogpond: Journal of the Haiku Society of America, IV:1.

Yamamoto. K. (1978). Kenkichi Yamamoto's Talk. Frogpond: Journal of the Haiku Society of America, I:4.

Yasuda, K. (1957). The Japanese Haiku. Tokyo: Tuttle.

Yoneoka, J. Implications of Japanese Speech Strategies of Children for the Status of the Syllable in Japanese. Unpublished manuscript.


[1] Cf. Cummins & Port (1998); Port, Cummins & Glasser (1995); Port, Cummins & McAuley (1995).

[2] Mora-- A linguistic term used to identify the sense of "phonic (hyouon) time-units" or "time-lengths" in Japanese speech. Mora, and morae, its plural, are English linguistics terms and also Japanese loanwords. Mora is defined by the American Heritage Dictionary, 1992, as "(Latin) The minimal unit of metrical time in quantitative verse, equal to the short syllable," and is used in Japanese linguistic circles as the katakana: mo-ra.

[3] The term segment is used here to mean one of the three traditional poetic or verse-lines (5-7-5) recognized in a Japanese haiku, to avoid confusion with the terms "line" (used in reference to lineation to mean a single line on paper) and "metrical line," used to mean the set of 8 beats on which a segment or poetic line may rest.

[4] Poetry: A Magazine of Verse (March, 1913).

[5] Japanese, being a moraic language, is more metrically regulated than English.

[6] "While the Imagists thought of the haiku as an ideal, none of them quite managed to ever write a true one. Pound's famous In A Station Of The Metro is often described as a haiku by persons with only a tenuous knowledge of the form (Swede, 1997)." Swede's view of Pound's poem is contravened by others, present authors included, who consider "Metro" an adequate and important example of haiku in English.The haiku aesthetic began to percolate through the English-speaking world from Pound on, influencing the Imagist and Objectivist poets, as well as those working in other artistic genres. A detailed history is beyond the scope of this article.

[7] Lamb (1994): "It was roughly the decade of the 1950s that saw the real beginning of what may be called the haiku movement in the Western world."

[8] Sato's statement echoes one made by Henderson (1971): "It is hard enough to define poetry . . . a definitive definition of haiku is probably impossible. When I asked professor Yagi [Yagi Kanetaro, Professor of Linguistics, then President of Matsuyama University of Commerce] he answered, 'There is no definition of haiku in Japanese. Haiku are what the poets make them.'"

[9] Cf. Higginson (1985, pp. 101-102). Also see the section, "Critique of Higginson's Description of 'A Traditional Form' for English Haiku" in this paper.

[10] Cf. Kacian (1999).

[11] Swede (1997): "Thomas Lynch (1989) . . . posits that an influential group of nineteenth-century New England poets, writers, and philosophers. . . the Transcendentalists created an intellectual and emotional climate receptive to the haiku. Lynch argues that  . . . Walt Whitman, Henry David Thoreau and Ralph Waldo Emerson, developed a homegrown philosophy quite similar to Zen Buddhism and that this way of thinking permeated their writing which, in turn, strongly affected the work of important twentieth-century poets such as Ezra Pound, Wallace Stevens, William Carlos Williams, Richard Wright, Allen Ginsberg, and Gary Snyder. Not surprisingly, these names appear whenever haiku scholars, such as Higginson (1985), list the major poets who have written haiku or haiku-like poems."

[12] Cf. A Haiku Path (1994, p. 5).

[13] This book presented Henderson's rhymed translations of Japanese haiku, a discussion of characteristics, and literal word-for-word translations. "Henderson spoke of his belief that there could and would develop an English-language haiku, but the time for that development had not arrived. With a few exceptions, serious interest in the study and writing of haiku would not come for another fifteen to twenty years" (Lamb, 1994).

[14] The Dharma Bums is currently ranked at approximately 3,500 in a sales-related index, of the million-plus titles sold at; by contrast the Haiku Handbook is about  30,000th in ranking.

[15] Kerouac also went on to write and record haiku, though his published output was small.

[16] The influential haiku critic Yamamoto Kenichi (1978), relates in a talk to the HSA: "Dr. Keene told me that the population of haiku makers in the States is probably larger than that in Japan and he smiled sort of mischievously. I was so surprised--so I asked him how that could be. In America, he said, haiku is very popular now and even in grade schools the teachers are getting students to write haiku."

[17] Van den Heuval (1986, p. 19): "After about twenty-five years of English language haiku do we know what haiku is? . . . [the lack of general consensus on this question is] a sign of its health and vitality."

[18] Cf. Van den Heuval (1999, Preface): "...The most comprehensive compendium of the haiku - its nature, uses and history - ever to appear in English, or any other language except Japanese."

[19] Recently acclaimed anthologies now include haiku submissions from around the world, translated into English. Particularly noteworthy are Snow on the Water (Kacian, 1999), Knots (Anakiev & Kacian, 1999), The Red Moon Anthology (Kacian, 1997, 1998), Haiku Anthology, 3rd ed (Van den Heuval, 1999), and Haiku World (Higginson, 1996).

[20] Similar concerns are also addressed by Gilbert (2000).

[21] On--(lit. "sound") A term used to count kana, or individual phonic units (hyouon) in poetry. It is this counter (or -ji), that is most commonly used to count the individual sounds in Japanese poetry. A Bracketed [-on] in this paper indicates that -on is replacing onji, a term not used in contemporary Japanese linguistics, and virtually unknown in Japan, which was mistakenly accepted as de rigueur in the West, after its erroneous introduction in the 1970s.

[22] Waka--a traditional generic term for poem, often used to refer to a tanka of 5-7-5-7-7 -on. Cf.

[23] This by no means implies any sort of moraic timing in English which might be equivalent to the onset-rime structure.

[24] Cf. Frogpond (1999): 1) An unrhymed Japanese poem recording the essence of a moment keenly perceived, in which Nature is linked to human nature. It usually consists of seventeen [-on]. 2) A foreign adaptation of 1, usually written in three lines totaling fewer  than seventeen syllables.

[25] Kiparsky (1989, p. 1).

[26] Adapted from Jyouo Hakutarou (1989).

[27] Cf. First version: Doi, 1922, quoted in Sakano 1996, p. 180 Second version: Jyouo 1989, p. 172.

[28] Adapted from Imaoka (1999).

[29] Shareware, available at

[30] We could also conceivably have had three 8-beat measures, which would have yielded an extremely slow metronome tempo of around 35-45. 

[31] In practice, as the haiku were measured from the beginning to the end of utterance, we are actually measuring only 21-22 beats rather than 24, because of the final silence beats. However, the 2nd line was measured in the same way, and thus for the most part only contained 7 (occasionally 8) beats; thus the overall proportions remain the same as those of  the entire haiku. 

[32] Merriam-Webster (1999).

[33] The English haiku is rarely parsed in the middle of the second line, as historically this was not considered to be an appropriate emulation of the Japanese haiku. Nonetheless, a midpoint caesura (say, indicated by a comma or en-dash, space, etc.) would emulate a recognized means of phrase parsing in the Japanese haiku.

[34] Renga--A poem usually composed by two or more poets, sometimes considered as a linked version of tanka, in which "one author wrote the 5-7-5 [stanza] and another responded and finished the poem by adding [a separate] . . . 7-7 [stanza]. Rather than ending the poem, another poet next wrote a new 5-7-5 [stanza] to "answer" the previous 7-7 link" and the poem continued in this linked fashion. Cf. Reichhold, J. []. Renga do not possess a linear narrative srtructure; typical lengths are 36, 50, 100 (or more) stanzas.

[35] With the possible exceptions of free-meter haiku, and those haiku (a statistically small number) that find exception to a 24 morae metrical structure.

[36] Tentatively, we may conclude that the inherent difference in length between -on and syllables would preclude the use of a MP-level rhythm for English haiku (=24 beats), but it is perhaps not impossible to construct a haiku of extremely short syllables that would read well with such a meter.

[37] Fermata--"A prolongation at the discretion of the performer of a musical note, chord, or rest beyond its given time value" Merriam-Webster (1999).

[38] Whitman is generally considered the progenitor of free verse, circa 1855.

[39] The musical analogy is that syllables are like notes in a musical measure. They can take on different temporal values and accents, and be placed anywhere in relation to the beat. The situation is not entirely free (or aleatoric), because there is a time signature, and a defined number of measures in the performance. As well, numerous diverse performances are metrically similar--as in the example mentioned of 12-bar Blues musical structures.

return to research page