Theorizing an Acoustic...

The fuzzy boundary between instrument and performer has been a frequent theme in music research. Interpretations of the relationship between performer and instrument have varied, ranging from seeing the body and instrument as totally distinct entities to seeing the body as an instrument itself. This research seeks to contribute to the discussion around what separates the musician from the instrument.
In this project, field recordings from the Kmhmu-speaking people of northern Indochina were analysed in order to examine a close relationship between melodic vocalizing and instrumentation. In many pieces, the voice of the instrumentalist is intertwined with the sounding of the instrument. This process yields both a quick continual interchange between voice and instrument and a perceptual blending of the sounds of both. Working from a select number of recordings,1 this research has focused on performances featuring both instrument solos (especially with a flute) as well as simultaneous flute and vocal performances, in which both resonators appear to be sounding at the same time: that is, the performer both sings through and plays the flute at the same time.

A Theorized Sound System
In order to understand variances in sound production, we have broken the production of musical sound into three constituent components. For the Kmhmu instruments examined here, the three most important components appear to be: the vibrating material (here, primarily air, however, in the case of the jaw harp, also the instrument's lamella); the vibrator (the mechanism that causes the vibrating material to vibrate); and the resonator. These three components can be found either entirely in the human musician (as in the case of singing) or as shared by the musician and the instrument.

The Application of the Sound System
The instruments analyzed in this project consist of two winds, the hoor tlaa and the toot , and a form of jaw harp called the hroong. The the hoor tlaa and toot are both constructed from bamboo. The hoor tlaa is approximately two meters long and is end-blown while the toot is of varying lengths and is transverse. The toot can be played with air supplied by either the mouth or the nose. The hroong is made of wood or bamboo.


Fig. 1 Photo of woman playing the hoor tlaa in Houei Sai province, LaoP.D.R., 1991. Photo by F.P.

With the hroong, hoor tlaa and toot as examples, we applied the three-component system to the various methods of playing on each instrument. As described above, this system includes three components: vibrating material, vibrator, and the resonator, each feasibly found in either the performer or the instrument. Figure 3 is a diagram charting the three components of our sound system for each instrument. One can see that the source and the resonator of the music varies from instrument to instrument. This exemplifies the fluidity of the boundary between musician and performer; if the mechanisms of the production of sound can switch from human to instrument, then the role of each (musician and instrument) in the production of sound must also vary.

Fig. 2 Photo of woman playing the toot.

In listening to these selections, it is frequently difficult to be certain perceptually which pitches are created by the vocal chords and which are emitted from an instrument; they sound alike, vocal and instrumental timbre imitating each other. The purpose of this research is to acoustically analyze these select Kmhmu performances in order to ascertain what exactly is speech-like about the woodwinds and what is instrument-like about the voice. It was suspected that part of the perceptual confusion lies in the transition of the three components outlined below.

Components of a Sound System

Vibrating Material Vibrator (human) Resonator (human) Vibrator (non human) Resonator (non-human)
Voice air vocal chords vocal tract
Toot air fipple flute
Hoor Tlaa air lips hoor tlaa
Hrong lamella finger vocal tract
Voice and Toot air vocal chords vocal tract fipple flute
Voice and Hoor Tlaa air vocal chords vocal tract hoor tlaa

Fig. 3 The three components (vibrating material, vibrator, and resonator) of an abstract sound system. On this chart are graphed the three Kmhmu instruments I have looked at as well as solo performances of the voice and performances which intimately combine vocalizing and instrument-playing. This chart displays the transition of the components of sound production among the different instruments. When two independent sound producers are combined (for example with the voice and the toot or the voice and the hoor tlaa), then the individual characteristics merge, leading to traits such as double resonators, found both in the human and in the instrument. Defining the components of sound production and examining the results of two independent systems merging has not been fully discussed in arguments over the boundaries between musician and instrument.

The (Human) Voiceness of Musical Sound
To explain the perceived similarities in sound in these Kmhmu musical performances, this research began with the question of what exactly was voice-like in the instrumental tones, and what was instrument-like in the vocalizations, anticipating that the results of analysis would contribute to an understanding of the "fuzzy boundary" between instrument and performer. The bulk of the research has focused on the voice-ness of instruments‹what acoustic characteristics do these instruments have that might cause them to sound like the human voice.

For the purposes of this research, "voiceness" is defined as the ability to sustain a timbre (or vowel quality) across the entire pitch range of which a speaker is capable. We chose this definition because research has shown that while the general source-filter theory of sound production is as valid for vocal as for instrumental sound, the voice is set apart by the fact that the coupling between the source and filter for the voice (i.e., the source wave produced by air passing through the vocal cords and the vocal tract resonator) is among the loosest of all sound systems. What this means functionally is that for the voice, pitch and timbre (or vowel quality) are independent of each other, whereas in most instruments with a more tightly coupled system, pitch and timbre are interdependent. As an instrument passes through its pitch range, its timbre will change, depending on the tightness of coupling between the source wave and resonator. We might imagine a coupling continuum with tightly coupled systems lying at one end and loosely coupled systems at the other. At the tight end, an instrument would change timbre with each pitch, while at the loose end of the continuum, the timbre would stay the same throughout the full pitch range of the instrument. Since the voice exemplifies looseness of coupling, we have designated a stable timbre as characteristic of "voiceness," while a changing timbre is characteristic of "instrumentalness."
Methods
Given the criteria for "voiceness" defined above, our method of analysis was to see to what extent and how quickly an instrument's tone (including tones produced by the voice) changed timbre as a function of pitch. In its steady state, a tone's timbre can be represented by the relative amplitudes of the tone's harmonics, as indicated by the tone's spectral envelop. A spectral envelop is simply a line traced over the top of each harmonic visible in a spectrum of the tone at a steadystate point in time. Thus, an analysis of the spectral envelopes of specific pitches within these pieces was completed using Soundscope and Spectro 3.0. For both the voice and the instruments, pitches from the high, mid, and low points of the pitch range of each instrument (during a single performance) were extracted and acoustically analyzed using these two programs.
With the hoor tlaa, the spectral envelopes at a high (Fig. A1), mid (Fig. A2), and a low pitch (Fig. A3) were significantly different with the fundamental assuming a less and less important role in the tone's timbre. This indicates that the instrument's bamboo resonator and source wave produced by the lips are tightly-coupled, resulting in a high level of dependence between the source wave and the resonator on the resultant pitch.

Fig. A1: Hoor tla, High Pitch Fig. A2: Hoor tla, Mid Pitch Fig. A3: Hoor tla, Low Pitch
A comparison of the toot led to similar results; the spectral envelopes of a toot differed between a high pitch (Fig. B1) and a low pitch (Fig. B2), also indicating tight-coupling.
Fig. B1: Toot, High Pitch Fig. B2: Toot, Mid Pitch Fig. B3: Toot, Low Pitch
The human voice, however, maintains a steady spectral envelope across a wide range of pitches, supporting the belief that the human voice is loosely-coupled (there is independence between the source wave and the resonator, resulting in a parallel timbre, vocal-quality, or speech-ness at low pitches and at high pitches). Work on the jaw harp, included in Figure 3, is not yet complete; it is included, however, because it represents a transition point between human speech and musical instrument.
When these Kmhmu instruments are accompanied by a simultaneous vocal performance, a double resonator is utilized, as was briefly described in Figure 3. When the performer both sounds the instruments and vocalizes, the resultant sound is a product of the loosely-coupled vocal system producing a sound which is then re-filtered by a second resonator. This often leads to a final musical product that has its own acoustic properties that make it perhaps difficult to categorize perceptually.
In several segments of the hoor tlaa performance, there are moments where it is difficult to discern if the performer is merely vocalizing, if she is blowing into the bamboo, or if she is doing a combination of the two. This methodology describes a way to determine acoustically what may not be discernible perceptually.
Ambiguous excerpts from the hoor tlaa piece were analyzed using SoundScope. Examining, again, the performance at both a high and a low pitch, the spectral envelopes were studied. The spectral envelope of the high pitch (Fig. C) was similar to the envelope of a high pitch excerpted from a solely instrumental segment (Fig. D). The spectral envelope of the low pitch, however, varied in shape between the segment that was solely instrument (Fig. E) and the segment that was an unclear mixing of instrument and voice (Fig. F). This discontinuous parallelism in the shape of the envelopes suggests that the second excerpt is indeed a mixture of hoor tlaa and human voice. Had the second excerpt (the excerpt perceived as unclear) shared the same shape at the high pitch as it did at the low pitch with the first excerpt (the excerpt that was strictly hoor tlaa), then it might have suggested that the second excerpt was also purely instrumental, based on its resultant acoustic characteristics. However, the production of a new spectral envelope at this low pitch leads one to assume that the performance is the result of a sound system more loosely-coupled than the instrument alone. Since analysis of the instruments revealed both to be tightly-coupled, it was determined that the resultant pitches are produced by a double resonator, a combination of sounding through the instrument itself and through the performer's vocal tract.

Fig. C Fig. D

Fig. E Fig. F
Similar results were found with an excerpt of a combined toot and voice selection. Here, as with the hoor tlaa, performances on the toot were excerpted. Segments at like pitches were compared at different moments in the piece and changes in the spectral envelopes were then compared. As expected, the spectral envelopes of the pure toot changes across pitches; the shape of the envelope at a high pitch (Fig. G) differs from the shape of the envelope at a low pitch (Fig. H). However, the envelope of the simultaneous sounding of the instrument and vocalizations lead to a constant spectral envelope across pitches; the shape of the envelope at a high pitch (Fig. I) is the same as the shape of the envelope at a low pitch (Fig. J). This second set of figures display a characteristic common to loosely-coupled instruments (such as the human voice) and is not normally characteristic of tightly-coupled instruments (such as the flute). However, because these figures were taken from performances where the toot was blown at the same time as the performer vocalized, the produced sound is a result of a double resonator, making this normally tightly coupled instrument become loosely-coupled.
This research has focused on analyzing the performer's vocal performances and the performances of the hoor tlaa and the toot, two bamboo flutes common in the Kmhmu culture. Research on another instrument, the hroong or jaw harp, is as of yet inconclusive. I do plan to include this non-woodwind instrument in the research because of its relevance to the fuzzy boundary between human and instrument.

Fig. G Fig. H

Fig. I Fig. J
Results
I am currently theorizing a continuum that seeks to explain the perceptual similarities between the sounds of the vocalist and the sounds of the instrument. At one end of this continuum (A) lies the characteristics of absolute voice‹segments of the performance where it is assumed that the vocalist is not using the woodwind to produce, modify (except in an incidental way), or amplify her voice. At the other end of this continuum (Z) lies the characteristics of absolute instrument‹segments of the performance where it is assumed that the instrument is clearly being sounded by only blowing air, without any vibration of the vocal chords.

(insert graphic of continuum)

This continuum is based upon research on the spectral envelopes of these Kmhmu instruments and vocalists. By examining the acoustic structure of both instrument and voice, the research has noted the stability, and instability, of the spectral envelopes over a wide range of pitches. These changes or parallels in the spectral envelope reflect the interdependence of the three components in a sound system: source wave, stimulus, and resonator. A tightly-coupled instrument (such as the hoor tlaa or the toot) will have a varying spectral envelope across pitches, an acoustic reflection of the high level of dependence between the components. A loosely-coupled instrument (the human voice, for example) will retain a constancy in its spectral envelope across pitches, an acoustic reflection of the independence of the components.

Significance of Research
The utility of this research lies perhaps foremost in the methods used. By acoustically analyzing the shape of spectral envelopes at various pitches, one is able to relate perceived sound to acoustic phenomena such as coupling.3 Coupling may be one way to explain the perceptual similarities shared by voices and instruments.
Discerning the interrelatedness of these components may lead to an acoustically-based understanding of the speechness of sound. This methodology allows us to focus on the timbre (or vowel quality) of a musical performance. By noting the stability or change in timbre across pitches, we are able to discern how tightly the three components of a speech system (source wave, stimulus, and resonator) are related. This constancy or variability of timbre allows us one more method of support to the debate around the speechness of music.

The author gratefully acknowledges the funding of the National Science Foundation for funding this source for this research.