The Acoustics of Kabuki Theaters

Summary The study presents a room acoustical investigation of a representative sample of eight Kabuki theaters as the most important public performance venues of pre-modern Japan. Room acoustical parameters according to ISO 3382 were measured for the unoccupied and simulated for the occupied condition. In comparison with European proscenium stage theaters, they have lower room heights in the auditorium, with usually only one upper tier, and no high stage house for movable scenery. The lower volume per seat results in lower reverberation times, The wooden construction and the audience seating arrangement on wooden straw mats on the ﬂoor instead of uphol-stered seats leads to a mostly ﬂat frequency response up to 4 kHz, resulting in an excellent speech intelligibility, as documented by values for deﬁnition (D 50 ) and the speech intelligibility index (STI). The acoustical conditions support the dynamic acting space created by pathways extending the stage from the front through the audience to the rear of the auditorium. They allow great contrasts in the perceived acoustical proximity depending on the selected acting position, and support a high degree of immersion of the audience into the dramatic action.


Introduction
The Kabuki is the most important genre of traditional Japanese public theater.D uring the Edo period (1603-1868), it became the primary form of public entertainment for the growing merchant class in the urban centers of Japan, with aparticular type of performance venue. Only after the Meiji Restoration of 1868, characterized by astate driven"modernization through westernization" affecting all aspects of society,theaters inspired by Western models were built in major cities such as Tōkyōa nd Osaka. At the same time, the industrialization brought city culture to more rural areas and led to an increase in the construction of Kabuki theaters outside the cities. Until today,the Kabuki is avital form of art, with about 20 active theaters throughout Japan.
The earliest records of Kabuki date back to the beginnings of the Edo period, describing female dance performances accompanied by flutes and drums, which took place on available Shrine stages, as well as on temporary open-air stages in Kyōto. These stages were inspired by existing stages for Nōt heater of the time, featuring a roofed stage, while the audience wasseated in front of the stage in open air.I n1 624 the first permanent theater in Edo (Tōkyō) wase stablished, called Saruwaka-za (later renamed Nakamura-za). It still had no roof above the audience seats, which were placed in front of the stage (hiradoma).
Permanent roofs started to appear from 1670, butitwas only after the issuing of fireregulations in 1723, that tiled roofs were required by the government, which needed new supporting structures. This process wasanimportant step towards the development of the physical theater in its final form [1]. In 1724, the three big theaters in Edo, namely the Nakamura-za, the Ichimura-za, and the Morita-za were all completely equipped with plastered walls and tiled roofs. Around the same time, ap athway called hanamichi with about 1.5 mi nw idth, which had started to develop from the end of the 17th century as at emporary extension of the stage, found its finala nd permanent position at stage right [2,3]. Starting in 1736, the practice of dividing the pit into rectangular areas of different prices (masu)w as introduced. Around 1772 an arrower secondary pathway (kari-hanamachi)was introduced at stage left, and the two were connected by atertiary path at the back of auditorium (ayumi). Gradually,the theater buildings for Kabuki devel- oped their characteristics distinguishing them from the Nō stage theyhad originated from. The roof above the stage, typical of the Nōstage, disappeared from the Kabuki theaters from around 1796. By 1830, the Kabuki theater (or shibaigoya as theyare usually referred to in Japanese)had reached its mature form [4].
Kabuki performances present ad ramatic plot from a standard repertoire of plays created in the 18 th and 19 th century.Staging historic events centered around the Samurai class or the life of the commoners of the feudal age, these plays consist of ac haracteristic form of singing, as well as acting and dancing accompanied by music on and off stage. At the core of aKabuki performance are the so called mie poses, in which the actor stays in acertain pose at the eshichisan point of the hanamichi for am oment to emphasize the action of the plot. These poses, as well as the beginning and the end of ap lay are accompanied by rhythmical motives, played on woodblocks (ki or tsuke) placed at stage left. As mall ensemble of one or more stringed instruments (shamisen), flutes as well as percussion instruments, placed behind as latted wall on stage right, contribute music and sound effects. Plays can also consist of anarrator sitting on aplatform on stage accompanying himself on the shamisen,orpassages of acting and dancing without dialog accompanied by asmall orchestra of shamisen musicians on stage, which is visible to the audience. The shamisen is plucked with ap lectrum and, together with the percussion instruments, forms ar hythmically accentuated background music, while the singers deliversustained legato notes to it.
In the current study,w ep resent the results of several room acoustical measurement campaigns [5] [6], with in-situ measurements and room acoustical simulations of a representative sample of remnant Kabuki theaters. The main goal of the study wastodescribe the range of room acoustical conditions of this performative genre, with a special focus on the particular features of these venues in contrast to theater buildings in the European tradition.

Description of the theaters
All theaters investigated are two-storeywooden structures. Theyexhibit the typical architectural features of this building type (Figure 2),a nd all of thema are still used for performances of traditional Kabuki plays. Twoo ft hem, the Hōō-za and the Kanamaru-za, were built in premodern Japan, during the late Edo period (1603-1868), while six of them were built in the subsequent Meiji period . Geographically,the theaters are located in three of the four main islands of Japan, including the islands of Shikoku and Kyūshūinsouthern Japan, and Honshu as the largest and most populous island in central Japan.
The Kanamaru-za, located on the island of Shikoku and completed in 1835, resembles the Kabuki theaters at the heyday of their development. In terms of size, dimensions, and stage machinery,i tm atches the dimensions of the three big Edo theaters [7]. The proximity to the Kompira Shrine, considered one of the most sacred places of worship in Japan, seems to be the reason for finding such a remarkable example of Edo period architecture in the rural area of the Kagawaprefecture. The other theater in the island of Shikoku, called the Uchiko-za, located in Uchiko town, Ehime prefecture, wasbuilt in 1916, celebrating the coronation of Emperor Taisho [8]. Four of the Kabuki theaters investigated are preserved in the Gifu prefecture in central Japan. The Hōō-za in Gero city is the oldest and also the smallest of the theaters studied. The original date of construction as an earby shrine stage is unknown (sometime in mid-Edo period)but it was relocated to the current site in 1827 and has been used as atheater since then. The Murakuni-za, opened in 1882 in Kakamigahara city,the Hakuun-za opened in 1890 in Gero city,and the Meiji-za, opened in 1895 in Kashimo village were constructed in the Gifu prefecture in the early years of the Meiji era, when commoners in rural areas of this prefecture came in contact with city culture through the emerging silk industry which resulted in af urther development of Kabuki performances and the increased construction of venues for entertainment [9].
The theaters on the island of Kyūshūw ere also constructed in the Meiji era. Theyi nclude the Ya chiyo-za, opened in 1910 in Ya magacity,Kumamoto Prefecture and the Kaho Gekijo, opened in 1921, located in Iizuka city, Fukuoka prefecture. Table Ishows the date of opening, the cubic volume, the capacity and the volume per person for the eight theaters considered in this investigation.

In-situ measurements
In the eight theaters of the current study (Table I),r oom acoustical measurements according to ISO 3382 were carried out [10], using al aptop-based measurement system 3 and sine sweeps to obtain impulse responses with alength   Figure 4). Fort wo of the theaters (the Meiji-za and the Hakuunza), an exemplary investigation wasc onducted, comparing the acoustical conditions for the most important acting positions in the Kabuki play.Besides astage-front and ar ear-stage position, which exist also in theaters of the European tradition, these include ap articular location on the Hanamichi pathway,where the most crucial parts of a Kabuki play such as the mie poses are presented. This is a point located seven-tenths away from the rear of the auditorium, or three-tenths away from the stage (shichi-san). Thus, in the measurements and simulations of these two venues, three source positions were investigated: • S A located on the center stage, 0.8 mbehind the front of the stage • S B located on the center stage, 5mbehind the front of the stage • S C located on the Hanamichi, at the so called shichi-san point Foreach source position, 12 receiverpositions were measured. Speech transmission index( STI)m easurements were carried out using ab roadband speaker with ad river of 12 cm diameter.
Room acoustical parameters according to ISO 3382 [10] were derivedfrom the impulse responses, including • the early decay time EDT as apredictor for perceived reverberance, • the sound strength Gasapredictor for perceived loudness, • the definition D 50 (early to total sound energy ratio)as predictor for speech clarity,and • the early lateral energy fraction J LF as apredictor for perceivedsource width.  The parameters derivedf rom the measurements were obtained using aMatlab script based on the ITAtoolbox [11]. The parameters derivedf rom the simulations were calculated in the software (see 3.2).

Simulations
Forthe acquisition of the geometry of the theaters, threedimensional point cloud data of the Meiji-za, Hakuun-za, Kanamaru-za and Uchiko-za waso btained using ac ommercially available laser scanner. 4 Fort he other theaters the geometry wasdetermined using alaser distance meter. Based on plan and section cut images exported from the laser scans, as well as on architectural drawings and pictures, computer models were created for the eight theaters, using SketchUp Make2017. As ag eneral guideline, we have attempted to keep a minimum structural size of 0.5 mi nt he room acoustical models, which has turned out to delivert he best simulation results [13, p.176], resulting in models with anumber of 100 to 300 faces. Scattering coefficients were set as suggested in [14] (a scattering coefficient at 707 Hz is specified and af requencyf unction of rising scattering values increasing with frequencyisextrapolated).
Forthe stage and the Hanamichi, absorption coefficients for wooden floor on joists were applied [15], while for the unoccupied and occupied Tatami, absorption coefficients were determined by ownmeasurements (Section 3.3).
The remaining surfaces include different, mostly wooden materials, whose absorption values are quite homogeneous butc annot be specified exactly by measurements in situ. Therefore, a"residual" surface wasassigned to all remaining surfaces and the values were fitted so that the resulting room average reverberation time would match the measured results within aJ ND of 5% as described in ISO 3382. In the model, an omnidirectional source and listeners were inserted at locations corresponding to the microphone positions in the in-situ measurements. The simulations of the speech transmission index (STI)w ere carried out using the source directivity of a male speaker [16], assuming an ormal vocal effort as defined in ANSI 3.5 [17] with abackground noise levelapplying the NC 25 curve. The simulations were further veri- fiedbycomparing the measured and the simulated STI values, which showed adifference of below0.05 in all cases. The simulations were conducted using ahybrid mirror image/ray tracing algorithm [18].

Measurements of absorption coefficients
Am ain difference between the Western theater and the Kabuki theater of Japan is the seating arrangement. The audience is not seated on chairs buto nr ice strawm ats called tatami.S ince absorption coefficients of audience seated on Tatami, especially with respect to historical seating density were not available, measurements of the sound absorption according to ISO 354 [19] for unoccupied Tatami as well as for audience sitting on Tatami were carried out in the reverberation chamber of TU Berlin (V = 200 m 3 ). Atest specimen consisting of six Tatami with a total surface area of S = 9.7 m 2 wasplaced on the floor of the chamber (type Amounting). The perimeter of the test specimen wassealed with an acoustically reflective frame made of 30 mm thick wood. Fort he measurements of the absorption coefficient in the occupied case, twoT atami with at otal surface area of 3.2 m 2 were placed in the corner of the reverberation room. To obtain absorption coefficients of an "infinite surface", the edges of the test specimen were covered with 500 mm high and 30 mm thick wood panels to avoid the increased aisle absorption, as suggested in [20]. To compensate for the increased sound absorption due to the 3dB higher sound pressure levelinthe edges, acorrection was applied as suggested by [21], enlarging the test surface by astrip of width b,where b = λ m /8.
According to [4], the three Edo theaters in 1841 accommodated fivepersons in one seating box (Masu), measuring 1.3 mby1.35 m. In later years, it wastried to increase the capacity of the theaters by reducing the size of one rectangle to 1. seems to be ap lausible average of the historical seating density in the Kabuki theaters. Therefore, on twoT atami, ten persons (five male, fivef emale)w ere seated in three rows of two, three, and twopersons resulting in acomparatively tight seating density of approximately three persons per m 2 .Another factor influencing the sound absorption is the clothing of the audience. Therefore, measurements were performed with persons wearing jackets and persons wearing no jackets. In Figure 5the sound absorption coefficients α s for Tatami as well as for persons sitting on Tatami with different clothing are shown. The values derivedf rom the measurements were converted to octave band values according to ISO 11654 [22] for the use in the simulations described in Section 3.2.

Reverberation times
The room average reverberation times values for the unoccupied case derivedfrom the measurements and the occu-

Early reflections
Individual early reflections, which can makean oticeable contribution to the acoustic characteristics of ar oom, arrive at the listener for times belowt he perceptual mixing time, which is between 50 and 100 ms for rooms of this size [24]. Forthis time window, Figure 7shows the typical pattern of early reflections appearing for different positions of the actor on stage, both from measurements in the Hakuun-za theater.W ith the source located at the center stage position (top), the direct sound is followed by stronger frontal first-order reflections from the floor (1),a nu pper reflection from the gable roof (3),al ateral reflection from the slanted walls on the side of the stage (4),a nd al ateral reflection from the sidewalls (5).A nother strong reflection arriving approximately 7msafter the direct sound (2) seems to be alateral second-order reflection from the side of the Hanamichi and the floor.
With the source located at the Hanamichi, strong frontal first order reflections can be identified coming from the floor (1),upper reflections from the twosides of the gable roof (2,3), as well as alate frontal reflection from the back wall of the stage (4).

Room acoustic parameters
The room average values of the reverberation time T 20 ,the early decay time EDT,the sound strength G, the definition D 50 and the speech intelligibility indexS TI for the unoccupied and the occupied case are shown in Table II. Va lues for D 50 between 0.68 and 0.91 and for the STI between 0.63 and 0.74 (both occupied)i llustrate the excellent speech intelligibility in all theaters. This is addi-  tionally supported by room average values for Gbetween 6.0 and 9.7 dB. The values for sound strength Ga ti ndividual listening positions in the eight theaters (occupied)a re between 3.4 and 12.9 dB. The decrease with increasing source-receiver distance is shown exemplarily for the Kaho Gekijo theater (Figure 8),with simulated values for the occupied condition compared to predictions by the classical diffuse field theory and Barrons'srevised theory [29,30]. Although the revised theory systematically overestimates the simulated  values by about 1dB, it offers aconsistently better fit than the classical theory,also in all other theaters considered.
The values for early lateral energy fraction J LF ,c alculated for the Meiji-za and the Hakuun-za theater (Table III)

Room acoustics and acting position
In contrast to the classical European proscenium stage the Kabuki theatre allows actors to takeupdifferent positions in front of, inside and behind the audience. By the example of twot heatres (Meiji-za and Hakuun-za), Table III and Figure 9i llustrate the acoustic effect of the different acting positions (main stage front, main stage back, Hanamichi pathway). With Speech Transmission Indices STI ≥ 0.65 and D 50,m values ≥ 0.73, speech intelligibility is always good regardless of the source location in both theaters. Nevertheless, there is an otable increase of both loudness (G m ), intelligibility (STI)a nd direct-to-diffuse ratio (asc haracterized by D 50 )w ith the speaker moving from stage back to stage front to the Hanamichi position. The big difference between the stage back and stage front positions is due to the absence of as tage canopy, which is whythe rear position is only supported by aweak ceiling reflection at the lower edge of the stage portal. As the most important acoustical cues for the perceiveddistance, these differences between the acting positions entail notable different sensations of proximity between actors and audience.
As an example, the spatial distribution of STI values for the three source locations (Figure 9) illustrates howdifferent parts of the audience are addressed by different positions of the actors and howt he sensation of being within the dramatic action evolves, when taking into consideration that the actors can freely move between these points, Figure 9. STI grid response from the simulation with the source located at the Hanamichi (top), at the back of the stage (middle), and the front of the stage (bottom)for the Meiji-za theater, assuming abackground noise of NC 25. and that several actors can be positioned at different locations at the same time.

Original Data
The original CAD-Models (.skp)ofthe eight theaters, including the source and receiverpositions used in the mea- Table II. Room acoustic parameters for the unoccupied (u) and the occupied (o) case. The values are averages for the octave bands suggested in [10]. The values in bold print are derivedfrom the measurements. Standard deviations are givenfor the occupied condition to illustrate the homogenity of the sound field in the audience area. surements and simulations are available as an electronic publication [28].

Discussion
The Kabuki as the most important traditional Japanese public theater form with its characteristic mixture of spokenand sung vocal passages with instrumental accompaniment has brought forth also aparticular architectural type of performance venue. It is au sually two-storeyb uilding with ar ectangular floor plan, and with the audience sitting on Tatami mats on the floor and on one surrounding gallery.Measurements and simulations of arepresentative sample of eight Kabuki theaters built between 1827 and 1921 (late Edo, Meiji and Taisho period)indicate the characteristic acoustical conditions of this genre. In comparison with European proscenium stage theaters such as the Viennese court theatres or the small Italian opera houses of the same epoch ( Figure 10, [25,27]), the Kabuki theaters are less reverberant relative to their size due to the relatively small volume per seat of 1-4 m 3 ,e xcept of one venue which is slightly larger.A lthough Kabuki plays combine elements of song, pantomime and dance with instrumental accompaniment, the acoustical conditions consistently seem to be designed for optimal speech intelligibility,w hich is indicated by early to late energy ratios (D 50,m )above 0.68 and speech transmission indices (STI) above 0.63 in the occupied condition, as well as ar ather flat frequency-dependent reverberation time for all theaters of the sample up to 4kHz. The conditions in terms of size and reverberation are most comparable with those of English theatres from this period such as Theatre Royal in Bristol or Wyndham'sTheatre in London [26]. One main difference when comparing the Kabuki theater to most stages in European tradition is "the unlimited freedom of its theatrical space where stage and auditorium merge and actor and audience sympathetically fuse into one" [7, p. 49]. This is achievedb ya ne xtension of the main stage by three pathways surrounding the audience, with the Hanamichi (stage-right)a st he most important. Together with the seating arrangement, which does, unlike European proscenium theaters, not predetermine the spectators' viewing direction, this creates ad ynamic perfor- mance space and ahigh degree of immersion into the dramatic action with respect to the social, visual and acoustical experience, as illustrated both by the overall change in room acoustical conditions and the spatial distribution of room acoustical parameters such as the speech transmission index(STI)for the different acting positions. These acoustic conditions characterize not only the experience of the audience of at heatrical genre of particular importance for Japanese culture, with its peculiar mixture of spoken theatre and music; theya lso theyc haracterize the cultural experience of aJapanese audience with the room acoustic conditions of music and theater performances in general, at at ime when Western concert culture came to Japan after the opening of the country following the Meiji Restoration of 1868. In contrast to aEu-ropean audience whose experience wass haped by av ariety of performance venues including large and reverberant spaces such as churches, large baroque festivalh alls, or even by Renaissance music theater in rooms with 2-3sr everberation [32], the Japanese audience only knew open-air performances, such as the older NōTheater,and the conditions of the Kabuki theater represented by the sample of rooms described here, with ad ensely packed audience and very clear acoustics. Western room acoustic standards for musical concerts with areverberation time of about 2swere thus highly unusual for aJapanese audience and could establish themselves only with agreat delay in the second half of the 20th century [33].