In recent years, thanks to Poser’s seminal 1990 paper, the two-mora foot has come to be recognized as one of the fundamental units in Japanese prosody, weakening the role of the syllable in the analysis of several accentual and other phenomena. In the wake of Poser’s paper, some linguists, such as Laurence Labrune, have returned to the traditional Japanese model of prosody, which did not recognize syllables as a category at all, but only recognized regular and “special” morae. However, while much has been made of the evidence for the existence of the two-mora foot and how special morae can be used as an alternative to a syllable-based analysis, I have yet to find any source that simply explains how feet are parsed in simple terms. Thus in order to glean as much information as possible on how morae are parsed, I elected to analyze a Japanese rap song, specifically “Ren’ai Circulation,” a theme song from Bakemonogatari.
The rhythmic system of Japanese is fundamentally different from that of English. Whereas the basic rhythmic unit of English is an acoustic unit, the syllable, the basic unit of rhythm in Japanese is a more abstract unit, called a mora, that may or may not correspond to a syllable in English. To be sure, most of all the possible morae are syllables, specifically syllables that do not end with a consonant. However, there are several types of sounds that occupy a mora rhythmically, even though they would not be analyzed as syllables in English. They are:
-the moraic nasal, written as n or n’; thus the honorific san would be analyzed as containing two morae, “sa” and “n,” whereas a native English speaker would analyze it as a single syllable.
-the second half of a long vowel; thus ookii, which begins with a long o and ends with a long i, consists of four morae, “o,” “o,” “ki,” and “i,” with the second half of the long o and the second half of the long i being separate morae.
-word-medial vowels without preceding consonants, particularly i and e; thus nai has two morae, “na” and “i,” whereas an English speaker would analyze the word as a single diphthong.
-the first part of a geminate (doubled) consonant; thus “yatta” (“[I/you/he/she/it/we/you/they] did [it]!”) contains 3 morae, “ya,” “t,” and “ta.” An English speaker would probably have trouble detecting the doubled consonant at all, let alone analyze it as a separate rhythmic unit.
These morae are called “special” or “deficient” morae by various authors; everything in Japanese that looks like a syllable ending with a consonant or vowel glide is actually a sequence of a normal mora followed by a special mora. Being less sonorant than morae that contain full vowels, they have some special restrictions on them. Most notably, in the Tokyo dialect they are not allowed to carry word accents, so that if a regular accentual rule would result in one of these morae being accented, the accent is pushed back to the preceding mora instead. (The only word as far as the author knows that is lexically accented on a deficient mora (not due to pre-accenting particles like tte or shika) is obaasankko, “gramma’s boy,” which is accented on the n.)
When several morae follow each other in sequence, they are divided into feet, which are units of two morae each, with the first mora in each foot being stressed. The first mora is said to be the “head” of the foot or to be in the strong position of the foot. (Which raises the question of whether the mora in the weak part of the foot is in the “foot” of the foot.) Here is where the situation becomes unclear. Some sources seem to suggest that feet are parsed from left to right, with the first mora in a word becoming the head of the foot, the next mora becoming the corresponding weak part of the foot, the third becoming the head of a new foot, and so on to the end of the word. Thus there might be a stray mora at the end if the word has an odd number of morae. Other sources, including Poser’s paper, seem to suggest that parsing goes from right to left, so that the last two morae in a word form a foot, then the next two preceding morae become the next foot preceding, and so on to the beginning of the word, so that if there is a stray mora it will be at the beginning of the word. There are also sources that claim that feet must always stay within a morpheme, so that e.g. “wakarimashita” would be parsed “(waka)ri(mashi)ta” rather than “(waka)(rima)(shita)” because the verb root “wakar(i)-” and the suffix “mashita” are separate morphemes. Laurence Labrune asserts that a word’s accent must fall on the strong part of a foot, so that e.g. “tokeru” would be parsed as “to(keru)” rather than “(toke)ru” because the accent falls on “ke” and so “ke” must start a new foot. Other sources suggest that foot parsing and accentuation are completely independent of each other. (Incidentally, based on my own subjective impressions from listening to Japanese, I would be more inclined to the latter opinion.) Labrune also asserts that special/deficient morae are not allowed to head a foot; thus e.g. “wakatta” would be parsed “wa(kat)ta,” because the “t” in the middle is not allowed to head a foot. However, while many sources make assertions about many individual rules of foot parsing, nowhere is there a simple list of rules of how feet are parsed. In order to take steps to remedy this situation, the author of the present paper decided to undertake a study of a Japanese rap song to analyze how feet are parsed in it.
Being curious about the process of foot parsing, I decided to study one of the several Japanese rap songs I had heard in order to analyze its foot structure. “Ren’ai Circulation,” the fourth opening theme of the anime Bakemonogatari, was chosen for this purpose, for several reasons.
First, it was catchy and easy to memorize.
Second, it happened to be the first Japanese rap song I had heard.
Third, “Ren’ai Circulation” is the theme song used in the “Nadeko Snake” arc, and Sengoku Nadeko is one of the author’s favorite characters in the Monogatari series. He particularly likes the parts where (POSSIBLE SPOILER ALERT!) Nadeko is revealed to have gone insane, with half the story of the “Nadeko Medusa” arc as she had narrated it consisting of outright lies; and where Kaiki reveals that he had seen the manga that Nadeko had secretly been writing (“And the plots were surprisingly perverted, too”), causing Nadeko to revert from a raging insane goddess to an embarrassed schoolgirl. (END OF SPOILERS)
Data collection and analysis proceeded in four steps. First, the author of the present paper listened to the song “Ren’ai Circulation” in the following video:
Actually, it was not this video, but the author does not think it worth the time and effort to find the first video.
He then looked up the lyrics at the following URL:
Then he listened to the song again until he had memorized its rhythm and lyrics.
Finally, he went ahead and analyzed it.
The lyrics of the rapping portions of the song are provided below in romaji for easy reference.
Kotoba ni sureba kiechau kankei nara
kotoba o keseba ii ya tte
omotte ta, osorete ta,
da kedo are? Nanka chigau ka mo…
Senri no michi mo ippo kara!
Ishi no you ni katai, sonna ishi de
Chiri mo tsumoreba yamato nadeshiko?
“Shi” nuki de―iya, shinu ki de!
Watashi no naka no anata hodo
anata no naka no watashi no sonzai wa
mada mada ookiku nai koto mo
Ima kono onaji SHUNKAN
kyouyuu shite ru JIKKAN
Chiri mo tsumoreba yamato nadeshiko!
Ryaku shite? Chiri-tsumo-yamato Nadeko!
Figures 1.1 and 1.2 shows the scansion of the lyrics into two-mora feet.
Figure 1.1: Verse 1
Scansion of the lines “omotte ta, osorete ta, da kedo are? Nanka chigau ka mo” uncertain
Figure 1.2 Verse 2
Scansion of the lines “mada mada ookiku nai koto mo” uncertain
Several points stand out upon a first look at the data. For one thing, foot parsing seems to go from left to right starting with the very first mora of each line, with the first mora being articulated as the head of a foot, the next being articulated as the weak mora of the foot, the next being the head of the next foot, and so on. This process does not respect either word boundaries or morpheme boundaries, so that the weak part of a foot can be in a different word or morpheme from the strong part of the foot; for example, “Kotoba ni sureba,” “Chiri mo tsumoreba yamato nadeshiko,” etc. If the last mora before a pause falls in the strong position, then that mora lengthens to cover the weak part of the foot as well, and the next mora begins a new foot, e.g. “Senri no michi mo ippo kara,” “Watashi no naka no anata hodo anata no naka no watashi no sonzai wa”
There are several lines where deficient morae are consistently placed in the weak position of the foot. For example, “Kotoba ni sureba kiechau kankei nara,” “Senri no michi mo ippo kara!” etc.
However, what is much more interesting is that there are also many cases where a deficient mora falls into the strong part of the foot. This creates a syncopated rhythm, depicted in musical notes in figure 2. The most obvious case is the line “Ishi no you ni katai sonna ishi de,” where almost all the feet in the string “… no you ni katai sonna …” are syncopated. Other examples include: “… anata no naka no watashi no sonzai wa,” “… wakatteru keredo.” This syncopation flatly contradicts Labrune’s assertion that deficient morae are not allowed to head feet. More research would be required to determine whether this occurs only in metrical speech or also in normal speech.
A similar effect occurs when a devoiced vowel falls into the strong position, as in “yamato nadeshiko” or “ryaku shite.”
Figure 2: Syncopated rhythm
Syncopated feet would also explain several loanword truncations that do not quite add up in either a syllabic analysis or a purely moraic analysis. In Japanese, words are often truncated into a two- or three-mora form; thus “animeeshon” is shortened to “anime,” “conbiniensu sutoaa” is truncated to “conbini” (consisting of two 2-mora halves) etc. But there are some words whose truncated forms are cut off in the middle of (what an English speaker would call) a syllable, for example “demonsutoreeshon” to “demo,” with the n of the second “syllable” left out of the truncation. This has been used as evidence against the existence of syllables as a prosodic category in Japanese. However, if, as Labrune holds, deficient morae are not allowed to head feet, then this truncation poses a problem under a purely moraic analysis as well. Going by Labrune’s analysis, the footing of “demonsutoreeshon” would have to be “de(mon)suto(ree)(shon)” (I’m not sure how she would treat the “-suto-” in the middle). The sequence “-mon-” forms a foot to itself, and so the truncation splits a foot down the middle―which, going by the logic used in arguing against syllables, would then constitute evidence against the existence of feet. On the other hand, if we allow for syncopated feet, the problem is solved: the footing becomes “(demo)(nsu)…,” which makes the truncation “demo” quite natural.
Another point (admittedly anecdotal evidence) that seems to support this conclusion is that the author has often heard voice actors lengthen the first mora in a word. As it would not make sense to lengthen a mora in an unstressed position at the beginning of the word (lengthening at the end of a word can be explained by an exaggerated or trailing-off intonation), this lengthening would imply that the first mora in a word naturally tends to be stressed, regardless of the morae, deficient or otherwise, that follow.
Labrune uses the idea of deficient (i.e. single-mora) feet in conjunction with her claim that accents must fall on the strong portion of a foot to explain the accentuation of loanwords. However, in view of the tendency of single morae to lengthen when placed in the strong portion of the foot with no mora following, I find the idea of deficient feet difficult to believe, as one would expect that in such cases, the single mora would lengthen to cover the rest of the foot. This phenomenon is not restricted to metrical speech; for example, it is found regularly when counting (e.g. in “ichi, ni, san, shi,” the monomoraic numerals “ni” and “shi” are lengthened) or when a monomoraic word appears in a sentence without a case-marking particle (e.g. hi tsukeru “set fire” is pronounced “hii tsukeru,” with lengthened i).
The claim that accented morae must fall on the strong part of the foot also seems dubious in light of the metrical structure of the song. In particular, in the line, “Watashi no naka no anata hodo anata no naka no watashi no sonzai wa …,” the lyricist seems to have gone out of her way to make sure that the metrical accents fall on the first and last morae of “anata” each time the word appears; the extra pauses inserted after the second “no” in “Watashi no naka no…” and after the first “anata” in “… anata hodo anata no naka no…” certainly have this effect. However, the accent of “anata” falls on “na.” Thus the lyricist has deftly avoided placing the metrical accent on the same mora as the pitch accent. Rather, it seems that she has prioritized placing the metrical stress at the beginning of the word rather than on the accented mora. Elsewhere too throughout the song, the placement of word accents with relation to metrical accents is regularly ignored; e.g. “sureba” and “keseba,” both accented on their second mora, are metrically stressed on their first and last mora. This situation is strikingly similar to that of Ancient Greek, another pitch accent language, which also disregarded word accent when assigning metrical stress.
Lastly, the singer’s articulation of the word “unmei” in the chorus of the song is also interesting. As I understand her, Labrune analyzes the mora as containing two “slots,” an onset position where a consonant would appear, and a nucleus position where a vowel would appear. Deficient morae are analyzed as having an empty slot in one position. Thus the moraic nasal would have an unspecified nasal consonant (articulation being determined by assimilation to the following consonant) and an empty vowel position, as shown in figure 3.
Figure 3: Internal structure of a mora
Now, in the song, the moraic nasal in “unmei” falls on its own note. In order to articulate this, the singer pronounces the word as “u-um-mei,” in IPA, [ɯ ɯm mej]. This does not seem to be consistent with the internal structure of the mora in Labrune’s account. If the moraic nasal consisted of an unspecified nasal consonant in onset position followed by an empty vowel slot, then one would think the singer would pronounce it more like [ɯ mm̩ mej], with a nasal in the onset rather than an epenthetic vowel before the consonant.
VI. CONCLUDING REMARKS
In conclusion, Nadeko is the author’s favorite girl in the Monogatari series. However, he must admit that although Nadeko is his favorite character, he actually likes Kanbaru’s theme songs best.
Moreover, the concept of syncopated feet (feet with a deficient mora in the strong position, observed to occur in rap) seems to explain several loan truncations that at first do not seem to work under either a syllabic or a moraic analysis.
Metrical treatment of accented morae seems to suggest that accent and stress are completely independent, contrary to Labrune’s assertion that accented morae must start a foot.
Based on how the meter of Japanese rap works, the author is inclined to think that foot parsing begins with the first mora of the word and proceeds to the end. This process is not affected by deficient morae or accent, but the author conjectures that it is cut off by morpheme boundaries.
The author also thinks that the idea of dividing all morae, including deficient ones, into an onset position and a vowel position is taking the whole dividing-stuff-into-parts thing a bit too far.