Waka and ryūka are short poems originating in Japan and the Ryukyu Islands, respectively. Waka has a very long history and flourished during the Heian period, which began around the year AD 800. The waka studied here belong to Ogura Hyakunin Isshu, a collection of 100 waka compiled around the first half of the thirteenth century, and the karuta game based on it. Waka – nowadays often referred to as tanka – is widely used in contemporary Japan, from the New Year recital at the Imperial court to popular tanka magazines and newspaper columns to which anybody can submit their poems. Though they have existed in writing for centuries, the oral aspect of waka/tanka is still strong, both with regard to how they are conceived and how they are orally presented.

Ryūka is known from the seventeenth century onwards. It originated in the higher strata of society as poems with separate melodies that were accompanied on the three-stringed long-necked lute, sanshin. The genre spread to other social contexts, and perhaps the best-known ryūka poetess was Onna Nabii, who lived in a village in the eighteenth century (Map 6). At that time, Japanese society was under feudal constraints; but her poems were free, sometimes sarcastic about politicians, and sometimes filled with amorous passion. A film has been made about her life.¹ Ryūka are performed orally with or without instrumental accompaniment, using a basic melodic formula.

The Ryukyu Kingdom was an independent country before it was integrated into Japan in 1879. Owing to the integration policy of the Japanese government, particularly after the Second World War, the Ryukyuan languages have been severely suppressed, and the vast majority of people on the Ryukyu Islands are monolingual – in Japanese. There has been debate about whether Ryukyuan should be considered as an independent language, forming a language family with Japanese, or as a dialect of Japanese. It is widely agreed that they are related, and that Ryukyuan resembles Old Japanese more than Modern Japanese. Here we will use the terms Japanese and Ryukyuan, as these terms are commonly used in Western literature. Because of differences with regard to phonology, grammar, and lexicon, Japanese and Ryukyuan are not mutually intelligible.

Map 6 Map of the Ryukyu Islands, with the village Onna on the main island of Okinawa.

Since Ryukyuan is a severely endangered language, ryūka is also under threat. Today, it is more common for people in the Ryukyu Islands to compose waka than ryūka, since only the older generation can speak genuine Ryukyuan, whereas the younger generation speak Japanese only.²

Waka and ryūka

The terms waka and ryūka literally mean uta, ‘song/poem’, in Japan and in the Ryukyu Islands, respectively. It was clear from previous research that the vocal expressions which result from the performance of waka and ryūka lend themselves to study as performance templates.³ Our main objectives were to describe the respective templates and to study differences and similarities. This includes the study of the performance of long and short syllables, the performance of irregular poetic lines, and the relationship with spoken intonation. The forms of representation chosen were Melodyne graphs, Praat graphs, and graphic transcription with a focus on the realization of syllables and on approximative pitches.

Japanese and Ryukyuan share major similarities regarding their prosodic features, such as syllable structure, accent, intonation, and rhythm. Both Japanese and Ryukyuan are lexical pitch-accent languages where all the content words have one of two accent types, with a few dialects as exceptions. Likewise, some intonation characteristics of Ryukyuan resemble those of Japanese.⁴

Waka and ryūka are short poems representing Japan and the Ryukyu Islands. They are similar in the mora counting principle, and in the total number of morae (metric units): a waka is composed of 31 morae, while a ryūka has 30 morae. However, there are at least two significant differences between the two: they differ in the number of phrases and in their respective compositions. Waka are based on an odd number of morae divided into five phrases: 5–7–5–7–7 morae, while ryūka are based on an even number: 8–8–8–6 morae, consisting of four phrases. Another way to express this is that the metrical form of waka terminates by lengthening (5⇒7 morae), whereas that of ryūka terminates by shortening (8⇒6 morae). It is interesting to note that both final lengthening and final shortening are commonly observed in human speech to indicate terminality.

Waka and ryūka also differ with regard to music. While ryūka developed along with the accompanying musical instrument, sanshin, and with a number of fixed melodies, waka have less often been connected with fixed melodies. Both, however, have a ‘recitation-like’ form of performance, and it is this form that will be analysed here.

Analysis 25 Two waka performances

Waka (tanka) is a traditional form of Japanese short poetry that is still often performed orally. Karuta is a memory game, in which cards displaying one half of a waka are distributed with the reverse side up and the players compete by combining the first and the last parts of the waka correctly. In this process the waka are normally read out aloud, and the material consists of two complete collections of such readings on a cassette tape, by males in both collections.⁵

Naniwazu is a waka poem which is not included in the collection itself but is used to start the karuta game (Example 115). In the two recordings, it is read without interruptions (the other 100 waka have a pause in the middle, long enough for the listener to fill in the second half before the correct continuation follows).

When waka/tanka are printed in Japanese, they are usually written in one line. There are exceptions, though, such as the three-line tanka of the poet Ishikawa Takuboku (1886–1912). In translation into Western languages, they are commonly written in five lines of 5–7–5–7–7 morae, but before about 1950 usually in four lines. These are, however, matters of typography. The usual way of defining a musical phrase in orally transmitted vocal expressions is that of a metric unit that has a recognizable ending, usually recurring more than once, consisting of, for instance, a long tone and/or a pause. In the performances studied here, these musical phrases coincide with each of the five lines. There is also a fairly long break after the first three lines so that, in the karuta game, they are divided into 5–7–5 and 7–7. In this case, the five lines also coincide with linguistic phrases. Linguistic phrases will simply be referred to as phrases, whereas musical phrases will be specified where this is deemed necessary. In Example 116, the waka Naniwazu is divided into five phrases and the morae are numbered.

Naniwazu 1

Characteristics:

• Total length ≈ 18 seconds.
• 1st part equalling 5 + 7 + 5 morae: ≈ 9 seconds.
• Dominating pitch C4 ≈ 259–263 Hz.
• Lowest pitch (starting pitch) B2 ≈ 124 H.
• 2nd part equalling 7 + 7 morae ≈ 6.5 seconds.
• Dominating pitch C4 ≈ 258–263 Hz.
• Lowest pitch (starting pitch) E3 ≈ 163 Hz.

There are two musical phrases, A and B. Actually, both are mainly performed at the same pitch, with interspersed shorter morae performed at a lower pitch. It is also clear that some morae are performed long and some much shorter. This is demonstrated in Examples 117–118. The distribution of short and long morae in lines of different length serves to give the performance a certain musical metre. It has been suggested that in performance, waka of 5- and 7-morae lines fall into 4-beat measures.⁶ In Example 118, stressed morae are preceded by vertical lines that serve as bar-lines. Whether or not this creates 4-beat measures is hard to say, but there is definitely a slow and steady pulse.

Example 117 Melodyne graph for waka Naniwazu 1 by a male performer. The vertical lines show ends of lines (metric units). Vertical: pitch, horizontal: time (1 column = 1 second). • 28 Waka Naniwazu 1

Example 118 Morae performed long and short in Naniwazu 1. Vertical lines in the graphic description approximately denote 4-beat measures, i.e. an even musical metre. Vertical lines in the text denote breaks between poetic lines.

• The first musical phrase A consists of the first 3 lines of the waka (5–7–5), and B of the final two lines (7–7).
• The first parts of A and B are nearly identical; but while the first part of A corresponds to a line of 5 morae, a line of 7 morae is compressed into what is basically the same time-frame in B. This is achieved by squeezing two morae into the space of one, or three morae into the space of two.
• The second and third lines (7 + 5) in A are performed one unit (= bar) longer than the second line in B (7). In both musical phrases, 5 morae are performed long. They are distributed as follows:
  Long morae:

  A)	B)
  Line 1: 2nd and 5th	Line 4: 3rd and 5th
  Line 2: 7th	Line 5: 3rd, 5th and 7th
  Line 3: 4th and 5th

The combination of pitches and rhythm is assumed to constitute the performance template in this style of waka performance. This means that all waka will follow the same basic pattern when performed in this style. A preliminary listening test supports this interpretation. Still, there may be variations depending on which sounds occur in various waka. This involves factors like long vowels and morae consisting of a single sonorant consonant (the ‘moraic nasal’ N). If speech intonation plays a part, it is not obvious. Furthermore, there are waka that differ from the norm regarding the number of morae in a line.

Naniwazu 2

Characteristics:

• Total length ≈ 15.5 seconds.
• 1st part equalling 5 + 7 + 5 morae: ≈ 7 seconds.
• Dominating pitch B♭3 ≈ 231–235 Hz.
• Lowest pitch (starting pitch) E♭3 ≈ 154 Hz.
• 2nd part equalling 7 + 7 morae ≈ 7.5 seconds.
• Dominating pitch B♭3 ≈ 233–237 Hz.
• Lowest pitch (starting pitch) D♭3 ≈ 139 Hz.

The Melodyne graph clearly shows the two parts (Example 119). In this case, the initial phrase has a shorter long tone on ni; and the very last mora of the second part, -na, is held for a relatively longer duration. Apart from this, the structure is similar: the pitch is rather stable, and the first mora of each line is performed short and at a low pitch (the graph did not register these for lines 2 and 5, but they are clearly audible). The structural sketch (Example 120) demonstrates how more morae are squeezed in, compared to example 118, while the basic principle is the same.

The combination of pitches and rhythm is taken to constitute the performance template in this style of waka performance. This means that all waka performed in this style will follow the same basic pattern. This interpretation is supported by listening. A comparison with other waka performances shows that the sonorant consonant N is treated as one mora while long vowels are treated as two morae. Variations occur in cases of irregular lines, for instance lines with 6 morae instead of 5, or lines with 8 morae instead of 7.

Example 119 Melodyne graph for waka Naniwazu 2 by a male performer. The vertical lines show ends of lines.

Example 120 Approximate pitches and short and long morae in the Naniwazu 2. Sections with several morae squeezed in are underlined.

Soundwave, amplitude, and duration of waka performance

Example 121 shows the soundwave and amplitude for the waka performances. The duration of each phrase is measured from the soundwave (Example 122). The following observations are made:

• Both performances are performed as two phrases of approximately the same loudness (cf. amplitude value in dB). In the performance in Example 121: top, the first musical phrase ends with decreasing loudness, while in Example 121: bottom, it is the second musical phrase that ends with decreasing loudness.
• The duration of phrases does not reflect the number of morae as speech does. Japanese is a mora-timed language for which each mora takes about the same time in speech, so that, for instance, a phrase containing 7 morae has a longer duration in speech than a phrase with 5 morae. This contrasts with stress-timed speech rhythm, where stress is isochronous. Germanic languages such as English and German are examples of stress-timed languages.⁷

Example 121 The soundwave, transcription, and amplitude for the waka Naniwazu 1 (top) and waka Naniwazu 2 (bottom) by male speakers. The vertical lines indicate the division between 5–7–5–7–7. The arrow A indicates decrease of amplitude.

The duration of each phrase is shown graphically in Example 122. For the speaker of Naniwazu 2, the durations of 5- and 7-mora lines are nearly equal in the two pairs 5–7 and 5–7, while the last 7-mora line is much longer. In order to accommodate this, words in 7-mora lines are compressed while words (vowels) in 5-mora lines are lengthened. The two speakers (Naniwazu 1 and 2) differ slightly as to exactly which vowels are lengthened.

Example 122 The duration in milliseconds for each 5 and 7 phrases/lines for the two waka performances. N1 = Naniwazu 1 and N2 = Naniwazu 2.

The waka performance template

Melody

• Level melody without speech intonation, but with a tonal centre (see Example 117).
• One dominating high pitch.

Rhythm

• Regular pulse.
• Even musical metre.

Form

• Short (15–18 seconds).
• Binary: Two musical phrases divided by a pause.

Phrasing

• Prosodic and musical phrases are aligned.
• Musical phrases end on lengthened morae (but other morae are also lengthened).
• A line of 7 morae is compressed into the same time-frame as a line of 5 morae. This is done by contraction: squeezing two morae into the slot of one, or three morae into the slot of two.
• Variations occur in cases of lines of irregular length.

Initial/final formulae

• Lines start with a glissando from a lower pitch on the first mora.

Analysis 26 Three ryūka performances

This analysis builds on performances of three ryūka attributed to the eighteenth-century ryūka poetess Onna Nabii, who lived in a village on Okinawa (Figure 7). Ryūka performance appears to be based on morae rhythm and does not show a clear metric pattern.

Ryūka kanjaganu

This ryūka follows the general pattern of four lines of 8–8–8–6 morae.⁸ It should be observed that in Ryukyuan, the moraic nasal N counts as one mora (Example 123).

A Melodyne graph of the Ryūka kanjaganu performance shows two clear sections: A of 8–8 morae and B of 8–6 morae (Example 124). The melodic phrases A and B have different pitches, with B significantly lower than A. The melodic phrases start with a rising interval and end with a falling interval, and they have a short final tone. There are many small intervals in the course of a performance; these small intervals are not recorded in detail. The distribution of morae and approximate pitches is shown in Examples 124–125.

Figure 7 A scene from the movie about Onna Nabii, showing the recitation of Ryūka Unnadaki and a transliteration with Japanese characters.

Example 124 Melodyne graph for Ryūka kanjaganu by a female performer, Kyoko Gushiken. The vertical lines show ends of lines.

Example 125 Distribution of morae in the Ryūka kanjaganu. Double vertical lines show the beginnings of major sections (preceded by upbeats), whereas single vertical lines show the ends of poetic lines.

Ryūka Unnadaki

This ryūka by Onna Nabii is featured in a film about her life (Examples 126–127).⁹

Characteristics:

• Total length ≈ 10 seconds (9.9).
• 1st part equalling 8 + 8 morae: ≈ 5 seconds (5.1).
• Initial pitch ≈ A#4: 461 Hz.
• Dominating pitch C5: 529 Hz – C#5: 545 Hz – C5: 529 Hz.
• The performance starts and ends on approximately the same pitch.
• 2nd part equalling 8 + 6 morae ≈ 4.5 seconds (4.6).
• Initial pitch ≈ F#4: 370 Hz.
• Dominating pitch A#4: 471 Hz – B4: 508 Hz – A4: 444 Hz.
• The performance pitch is where the first half starts and ends about a semitone lower.
• The penultimate mora comes out one octave too high in the graph, i.e. twice the frequency: A#5: 909 / 2 Hz = A#: 454 Hz (marked with an arrow).
• Two other high pitches in that line start high and drop down, but this is not shown in the graph (appoggiatura, marked with arrows in the graph of Example 127).

The basic structure of this performance is similar to that of the previous one (Ryūka kanjaganu). However, the first melodic phrase ends at the starting pitch instead of going down (A in Example 128). As in Ryūka kanjaganu, there is a tendency towards a very small pitch difference in the last two morae in a line. In this case, however, the pitch rises slightly instead of falling.

Example 127 Melodyne graph for Ryūka Unnadaki by a female performer. Arrows mark registrations that came out wrong in the graph (compare Example 128). The vertical lines show ends of lines.

Example 128 Distribution of pitches and morae in Ryūka Unnadaki. Double vertical lines indicate the beginnings of major sections (preceded by upbeats); single vertical lines signal the ends of poetic lines.

Ryūka Unnamachi

This ryūka by Onna Nabii is featured in the same film (Example 129).¹⁰

Characteristics:

• Total length ≈ 9 seconds.
• 1st part equalling 8 + 8 morae: ≈ 5 seconds.
• Initial pitch ≈ A4: 444 Hz (rising).
• Dominating pitch C5: 517 Hz – C#5: 545 Hz – C5 535 Hz.
• The performance starts and ends at approximately the same pitch.
• The pitch on the last two morae falls to ≈ A4: 452 Hz.

• 2nd part equalling 8 + 6 morae ≈ 4 seconds.
• Initial pitch ≈ F#4: 360 Hz.
• Dominating pitch A4 452 Hz – B4 483 Hz – falls to G#4 425 Hz at the end.
• The performance pitch is approximately where the first mora of the first half is performed, and ends about a semitone lower.

The basic outline agrees with the previous ryūka. The last two morae in the first and third lines are performed at a slightly higher pitch (Examples 130–131). As in the other performances, the second musical phrase starts at a low pitch, but in this case the second line also starts low on the first short mora (chi in chijinu-).

Example 130 Melodyne graph for Ryūka Unnamachi by a female performer (the same as for Ryūka Unnadaki). The vertical lines show ends of lines (metric units). • 29 Ryûka Unnamachi

Example 131 Distribution of pitches and morae in Ryūka Unnamachi. Double vertical lines indicate the beginnings of major sections (preceded by upbeats); single vertical lines show the ends of poetic lines.

Soundwave, amplitude, and duration of ryūka performance

Example 132a–b shows the soundwave and amplitude for the ryūka performances of Ryūka Unnadaki and Ryūka kanjaganu. The duration of each phrase/line is measured from the soundwave.

Example 132 Soundwave, transcription, and amplitude curve for two ryūka performances. a) Ryūka kanjaganu: the vertical lines indicate the division between 8–8–8–5. The arrow A indicates amplitude decrease. b) Ryūka Unnadaki: the vertical lines indicate the division between 8–8–8–6. The arrows A and B indicate amplitude decrease.

In performance a), there is some amplitude decrease (amplitude value in dB) in the second group of 8 morae and a more pronounced decrease at the very end of the performance. In performance b), the total melody is clearly divided into two musical phrases that differ in loudness. Difference with regard to loudness is also seen in the width of the soundwave. The first part is much louder than the second. Each musical phrase is performed with rapidly decreasing loudness for the last sonorant, here a prolonged [m:] indicated by the arrow A and [N:] indicated by the arrow B.

The decrease in amplitude is paralleled by a decrease in duration. In Example 133, the performances of Ryūka Unnamachi and Ryūka kanjaganu are compared in this respect. It shows that the duration decreased in the second half of the performances.

Example 133 The duration in milliseconds for each 8 and 6 phrases/lines for the two ryūka performances. uN = Ryūka Unnamachi, kaN = Ryūka kanjaganu.

The ryūka performance template

Melody

• Two musical phrases, of which the second is lower than the first, corresponding to downstep in speech intonation (see Example 124).
• Speech intonation is present in the rising start of musical phrases and in the falling endings.

Rhythm

• Each phrase constitutes one (slow) pulse.
• There is no regular rhythmic subdivision within a phrase.

Form

• Short (9–12 seconds).
• Binary: Two musical phrases divided by a pause, corresponding to lines of 8 + 8 and 8 + 6 morae, respectively.

Phrasing

• Musical phrases end on shortened morae and with decreasing amplitude.

Initial/final formulae

• Musical phrases start with a glissando from a lower pitch on the first or second morae.
• Musical phrases generally end with a falling interval, and with the final mora short and unstressed.

Word variations

• The moraic nasal N has the duration of one mora.

Speech and vocal expression

The performances of waka and ryūka display some notable differences compared to the spoken forms of Japanese and Ryukyuan as regards rhythm and intonation. Waka performances deviate from speech, both in rhythm and in intonation.

Rhythm

In performance, the number of morae in the phrases/lines is not reflected in duration, as is usually the case in speech. As seen in Examples 121–122, the entire sequence of 5–7–5–7–7 phrases is divided into 5–7 | 5–7 | 7, each line with an increasing duration. Segmental durations in 7-mora phrases are shorter than those in 5-mora phrases. The waka verse form itself is a short–long verse: short (5 morae) – long (7), short (5) – long (7), long (7). When performed, this pattern is further reinforced by the distribution of short and long tones. There are some cases in which the second mora of a bi-moraic foot is lengthened, as argued by Poser: nani: (line 1), kono: and hana: (both in line 5) and there are no examples of lengthening of the first mora pair.¹¹

In contrast, the ryūka verse form is composed of a long (8) – long (8), long (8) – short (6) pattern. This seems to be reinforced in the performance by a subdivision of the two musical phrases that coincides with the poetic lines, as shown in Example 128. Ryūka employs shortening to mark finality, whereas waka employs lengthening for finality. In ryūka, the lengthening tends to be on the last vowel or on a moraic nasal of the musical phrase.

A notable difference between waka and ryūka performances is seen in their pitch patterns, as demonstrated in the Melodyne graphs (Examples 117 and 119 for waka, Examples 124, 127, and 130 for ryūka). Although both waka and ryūka performances have two musical phrases, which are separated by a pause, there is a difference in the way they are arranged. In waka performance, the two musical phrases are produced at the same pitch register, while in ryūka performance the second musical phrase is produced at a lower pitch register than the first. One of the characteristics of Japanese and Ryukyuan intonation is placing a final group of words at a lower pitch register. Furthermore, each musical phrase starts with a pitch rise and terminates with a pitch fall in ryūka, exactly as in Ryukyuan intonation. The onset pitch rise is also present in Japanese intonation, but it is not reflected in waka performances. Example 134 shows the intonation contour for the sentence Manamigaru naafankaija ichura, ‘Manami goes to Naha’, as spoken by a male speaker of the Shuri dialect. The utterance can be seen to be divided into three intonation units that descend gradually in pitch register. Each unit starts with a pitch rise and ends with a pitch fall. Furthermore, the rapid pitch fall with disappearing tone found in the ryūka citation is also a reflection of Ryukyuan intonation. Note that this final weakening is not reported for Japanese intonation.¹²

Other acoustic features such as loudness (shown in the amplitude) were not found to be consistent for ryūka. In one speaker’s ryūka (see Example 132b), the second musical phrase was consistently performed with lower amplitude; but this was not found in a ryūka performed by another female speaker (Example 132a).

Example 134 The spoken phrase meaning ‘Manami goes to Naha’ (male speaker, Shuri dialect). Three intonation units are separated by the vertical lines and numbered 1–3. Each unit starts with a pitch rise and ends with a pitch fall. (Adapted from Nagano-Madsen 2015)

Waka and ryūka summary

The characteristics of waka and ryūka performances are summarized in Table 8.

Performances of traditional poems in Japan and the Ryukyu Islands, waka and ryūka, were compared for their melodic patterns, rhythmic patterns (duration), and amplitude. The waka performances revealed a total deviation from Japanese speech rhythm and intonation. The melody of waka performance has a high-level pitch in the same pitch register throughout, with an extreme final lengthening for the two sections. Mora-timed rhythm, reflecting the number of morae in an utterance, was not observed.

As for the melody of ryūka performance, a high level pitch register, as observed for waka, was also present. However, the melody of ryūka also exhibits some similarities to the phrase- and sentence-level intonation in Ryukuan speech. Since Ryukyuan (and also Japanese) is a lexical pitch-accent language in which each content word has its own accent, it is not possible to formulate a melody that fits all the ryūka. However, the components of Ryukyuan intonation at the phrase and sentence levels are uniform, regardless of the pitch-accent components of each word. These phrase and sentence-level components of intonation are copied in the performance of ryūka. They are: 1) an initial pitch rise for each musical phrase of ryūka, 2) a final pitch fall found at the end of a musical phrase, 3) the second unit of ryūka is produced at a lower pitch register, and 4) a rapid pitch fall and intensity decrease at the end of ryūka. All the four components are found in corresponding intonation units of Ryukyuan. Though the first three characteristics are equally present in the intonation of Japanese, they are not reflected in the waka performance.