Resonance Effects and the Vocalization of Speech

Abstract

Studies of the respiratory and laryngeal actions required for phonation are central to our understanding of both voice and voice disorders. The purpose of the present article is to highlight complementary insights about voice that have come from the study of vocal tract resonance effects.

Introduction

This paper overviews the following areas: (1) special resonance effects that have been found to occur in the vocal productions of professional performers; (2) resonance and anti-resonance effects associated with nasalization, together with clinical considerations associated with the diagnosis and/or treatment of hyponasal and hypernasal speech; and (3) studies of resonant voice and what they tell us about both normal and disordered speech production.

The respiratory and laryngeal actions required for phonation have been extensively studied and modeled (e.g., Hoit, Plassman, Lansing, & Hixon, 1988; Story & Titze, 1995; Alipour, Berry, & Titze, 2000; Hunter, Titze, & Alipour, 2004). Findings from these investigations are central to our understanding of both voice and voice disorders, and they continue to inform research in both areas. Also integral to the study of voice is an understanding of vocal tract resonance. Briefly, resonance concerns acoustical effects that the vocal tract exerts on sounds that propagate through it. It is affected by every part of the vocal tract from the vocal folds to the lips and external nares. Phonation creates a set of acoustic harmonics that are selectively enhanced by vocal tract resonances. The exact form that this enhancement takes can affect both listener perceptions and self-perceptions of vocal quality. It can also form the basis for vocal training.

The purpose of this paper is to provide an overview of vocal tract resonance effects and vocalization, with an emphasis on three areas. First, we overview resonance effects that have been found to occur uniquely – or especially prominently – in the vocal productions of professional performers. Next, we overview resonance and anti-resonance effects specifically associated with the nasalization of phonetic segments. We then consider the contribution of nasalization to perceptions of speech sound quality, and we discuss the clinical evaluation and treatment of hyponasal and hypernasal speech. Finally, we overview findings regarding resonant voice, which is an area of study that has contributed to the understanding of both normal and disordered speech production.

Professional Voice

Listener perceptions of age, gender, and overall vocal health have all be found to be influenced by the long-term spectral characteristics of a talker’s speech (Cleveland, 1976; Mendoza, Valencia, Muñoz, & Trujillo, 1996; Hartl, Hans, Vaissiere, Riquet, & Brasnu, 2001; Linville, 2002; Sjölander, 2003). Here, we report on a series of studies that have found that the long-term spectrum also contributes importantly to listener perceptions regarding the vocalizations of professional performers.

The Singer’s Formant Cluster

The long-term average spectra (LTAS) of operatic singing voices – especially male voices – have been found to have a broad peak in the spectral region around 3 kHz (Bartholomew, 1934; Sundberg, 1974, 1986). Sundberg termed this peak the “singer’s formant” (Sundberg, 1974) and, later, the “singer’s formant cluster” (Sundberg, 2003, Sundberg, Gu, Huang, & Huang, 2012). A particular benefit of this formant cluster is that it helps a performer’s voice stand out clearly when singing against a background of orchestral music. This is so because orchestral instruments have relatively reduced power in the spectral region where the formant cluster is located (Sundberg, 1974).

The singer’s formant cluster has been found to be more pronounced in trained singing voices than untrained voices, and to be only minimally present in conversational speech (Miller & Schutte, 1983; Schutte & Miller, 1985; Bloothooft & Plomp, 1986; Sundberg, 1987). It has also been found to be stronger when a trained performer sings in solo mode compared to a choir mode (Rossing, Sundberg, & Ternström, 1986).

Lowering the larynx lengthens the pharyngeal cavity and increases its cross-sectional area just above the larynx. Sundberg (1974, 1987) found that when the cross-sectional area expansion becomes large relative to the size of the laryngeal tube opening, this vocal tract configuration promotes a tight grouping of formants F3, F4, and F5. This grouping, in turn, creates a broad spectral peak in the 3 kHz region¹. A peak of this kind can be seen when any of the vowels are sung, and it is especially prominent for the back vowels. The height of the peak has been found to increase sharply in all vowel contexts whenever a performer sings loudly (Cleveland & Sundberg, 1983; Sundberg, 1990).

A resonance effect of a different sort – one specifically associated with the first formant – has been found to have particular importance for female operatic singers, who often sing with high fundamental frequencies. When the frequency of F1 matches (or nearly matches) the fundamental frequency (f₀) of a soprano’s voice, that fundamental component is strongly amplified and the overall loudness of the singing voice is increased. Sundberg (1977) reports that accomplished sopranos will vary their jaw position so that the F1 frequency ‘tracks’ f₀ and consistently affords this amplification effect. This technique has two notable advantages. First, it increases the overall efficiency of vocal production, and second, it affords greater loudness consistency and control whenever a sung passage ranges across different vowels (Sundberg, 1977).

The Actor’s Formant

Actors must speak with substantial vocal power in order to be heard throughout a theater and, at the same time, must maintain precise control over the voice in order to convey subtleties of expression and emotion (Master, DeBiase & Madureira, 2010). Leino (1993) made an acoustical assessment of acting voices that varied widely in their perceived quality. Leino specifically compared the LTAS of minute-long speech samples produced by 48 male professional actors whose overall voice quality had been rated as “poor”, “rather poor”, “fairly good”, or “good.”² What he found was that the most highly rated voices had two characteristics in common. The first was that their spectra had less slope – i.e., fell off more gradually with increasing frequency – which meant that the higher speech formants had substantial power. The second shared characteristic of these voices was that their spectra had a prominent peak in the 3 – 4 kHz region, a peak that Leino termed the “actor’s formant.” Together, the actor’s formant and the gentle spectral slope characteristic resulted in the highly-rated voices having 10–15 dB more power than other voices in the spectral region around 3500 Hz.

The actor’s formant is higher in frequency than the singer’s formant cluster and also somewhat weaker on average. Another notable difference between the two is that the actor’s formant is prominent in a performer’s speech, whereas a substantial singer’s formant cluster is rarely seen with speech. One thing the actor’s formant and the singer’s formant cluster do have in common is that they can both be strengthened by vocal training (e.g., Leino & Kärkkäinen, 1995; Frisell, 2007).

Vocal training could give rise to an actor’s formant if it either strengthened F4 individually or lowered the frequency of F5 to create an F4 – F5 formant cluster. Leino, Laukkanen, and Radolf (2009) found evidence of the latter change when they made a detailed study of the voice of an acting student in training. An articulatory model further showed that the clustering of F4 and F5 could be achieved by narrowing the oral cavity at the front and widening it at the rear, while simultaneously narrowing the epilarynx to some degree.

As noted above, one of the regular demands on actors is that their speech must project well enough to be heard clearly throughout a performance space. Pinczower and Oates (2004) asked a group of professional actors (n=13) to deliver a speech as if performing before a small audience in a small room (a condition they termed ‘comfortable projection’) and as if performing in a large theater (‘maximal projection’). A subsequent acoustical analysis found that the actor’s formant was significantly more prominent in the maximal projection condition.

Master, DeBiase, Chiari, and Laukkanen (2006) also asked professional actors to vary their vocal projection as if performing in spaces of increasing size (three spaces, termed ‘small,’ ‘medium,’ and ‘large’). For comparison, they then asked a group of nonactors to carry out this same task. Both groups increased the sound pressure level (SPL) of their voices monotonically across the small, medium, and large room conditions, and they did so by nearly identical amounts (mean SPL values differed by less than half a dB in any condition). However, the actors group was found to have a significantly stronger actor’s formant in every projection condition tested. A subsequent perceptual rating test found, further, that the actors’ speech samples provided a greater sense of projection in every space, and sounded louder in every space as well.

Actresses must also produce speech that projects well in different spaces, and Master et al. (2012) report that their strategies for doing generally do not involve the actor’s formant. Instead, they found that actresses make significant laryngeal adjustments which allow them to maintain a relatively low speech fundamental frequency, even when speaking at a high intensity.

Nasalization

When a nasal vowel is produced, the velum (soft palate) is lowered and air flows out through the talker’s nose as well as the mouth. When a nasal consonant is produced, the velum is again lowered and the oral cavity is fully occluded, so air flows out through the nose only (Huffman & Krakow, 1993). In either case, the spectrum of the resulting speech sound may be perceptibly altered by any or all of the following acoustical effects.

The nasal cavity has resonances at multiple frequencies that are determined by its overall size, and by the degree of opening at the nares (Johnson, 1997). Spectral components that coincide with these resonances will be enhanced. The first nasal resonance is especially prominent. For an adult talker, it typically has a center frequency around 250 Hz, which strengthens the first harmonic or second harmonic of the voiced sound source (Schwartz, 1968; Chen, 1995; 1997).

Acoustical reflections from the oral and nasal cavities can interact destructively within the vocal tract and create anti-resonances (spectral zeroes) at multiple frequencies. The strongest anti-resonance weakens components in the region of F1 which can add to the distinctiveness of the first nasal formant, which is typically lower in frequency (Schwartz, 1968; Johnson, 1997; Macmillin, Kingston, Thorburn, Dickey, & Bartels, 1999).

The nasal cavity has a large, pliant surface area which absorbs sound and produces both heat loss and rapid acoustical damping. This, in turn, broadens the bandwidths of the speech formants, especially the bandwidth of F1 (Maeda, 1993; Johnson, 1997; Pruthi & Espy-Wilson, 2004).

Language Variation and Individual Differences

Nasalized vowels and non-nasalized (oral) vowels are regularly produced by speakers of both English and French, but nasalization is a phonemic feature of vowels in French only (Haspelmath, 2005). In English, nasalized vowels are an allophonic variant that occurs exclusively when the vowel of a syllable is coarticulated with a neighboring nasal consonant. Styler (2017) compared the acoustic signatures of nasalized vowels produced by native speakers of French and by native speakers of English,³ and he found a significant language difference, especially regarding a factor called “spectral tilt”.⁴ Specifically, he found that the French speakers produced more dramatic tilt, possibly to enhance the overall audibility of this factor.

Styler also found that there was a great deal of talker-to-talker variability regarding the relative weighting of the various cues to vowel nasalization in both language communities, so much so that he speculated that some form of speaker normalization may be needed for the accurate perception of vowel nasalization, much as a normalization step is widely thought to be needed to compensate for substantial speaker differences in the F1 – F2 vowel space (Peterson and Barney, 1952; Ladefoged & Broadbent, 1957, Syrdal & Gopal, 1986).

Clinical Considerations

Nasality is a phonetic characteristic of normal speech; however, velopharyngeal dysfunction can lead to altered perceptions of this phonetic characteristic. For example, Tardif, Bertie, Marino, Pardo and Bressman (2018) report that hypernasality contributes to the listener’s perception of monotonous speech. Hypernasality “occurs when there is sound energy in the nasal cavity during production of voiced, oral sounds.” Hyponasality “occurs when there is not enough nasal resonance on nasal sounds due to a blockage in the nasopharynx or nasal cavity” (ASHA). ⁵

An important clinical decision regarding nasality is whether to treat it medically or to manage it behaviorally. Conditions that cause abnormal nasality are most evident in the diagnostic categories related to oral facial anomalies, maxillofacial malformations, syndromic conditions, hearing loss, and organic or functional changes associated with neurogenic communication disorders such as motor speech impairments or degenerative diseases.

A distinction between lack of adequate function of the soft palate for speech production, velopharyngeal incompetence, as different from the lack of adequate musculature, velopharyngeal insufficiency, is both practical and necessary to make diagnostically (Barsties and DeBodt, 2015). Visual examination of the velopharyngeal port via flexible nasopharyngoscopy allows the evaluation of strength, function, and structure of the velum as well as a visual inspection of the coupling of the nasopharynx and the oropharynx during speech.

Management of nasality is accomplished through four primary areas: 1) surgical management, 2) prosthetic management, 3) pharmacologic management and, 4) behavioral speech therapy (ASHA).

Surgical management

Hypernasality.

Anatomical and structural differences can be managed surgically via pharyngeal flap, pharyngeal wall augmentation, sphincter for pharyngoplasty, or a Furlow Z-palatoplasty. In the case of a nasal fistula, variations of flap surgery can be beneficial.

Hyponasality.

Surgical intervention is likely the treatment of choice to correct structural changes that obstruct the airway and that may restrict the phonetic component function of the nasopharynx. Common procedures include tonsillectomy/adenoidectomy, removal of nasal polyps, revision of the deviated septum, reconstruction of the stenotic nares, or management of choanal atresia.

Prosthetic management

Prosthetics are incorporated to correct resonance imbalances due to hypernasality when there are no surgical alternatives or if the individual is discomforted by surgical options. Prospects for prosthetic management may include the creation of a palatal obturator or a speech bulb, or fabricating a palatal lift to elevate the velum to offset limited velar movement (velopharyngeal incompetence). In cases of velopharyngeal insufficiency or incompetence a nasal obturator may be fabricated to reduce airflow through the nasopharynx.

Pharmacologic management

Medications such as antihistamines and steroids may be utilized to decrease swelling of or inflammation in the nasal cavity that are concomitant with allergies or other airborne irritants affecting the mucosa that lead to hyponasality.

Behavioral speech therapy

Speech therapy will not eliminate a resonance disorder that is caused by a structural anomaly. Issues associated with nasality that can be attributed to misarticulation or with secondary facial characteristics such as nasal air emission can be managed behaviorally. Options for behavioral speech therapy include phoneme-specific nasal air emission (PSNE) or phoneme-specific hypernasality with normal velopharyngeal function. Intervention techniques include visual feedback, auditory biofeedback, instrumental management through the use of the Nasometer (Wermker, Jung, Ulrich, & Kleinheinz, 2011), and standard phonetic speech therapy techniques associated with acoustic software applications.

Nasometry.

Abnormal nasality as perceived in human speech can be evaluated subjectively. Sometimes severity is indicated on an ordinal scale ranging from zero to indicate normal nasality to three or five with higher values suggestive of pervasive hypernasality with nasal air emission (NAE), nasal snort, or facial grimacing.

Efforts to quantify nasality have been facilitated by the development of the Nasometer.⁶ The Nasometer is a tool that endeavors to provide an objective measure of nasalance based on a ratio calculation between input differences of microphones placed near the oral and the nasal cavities. Nasalance is defined as “the ratio of amplitude of the acoustic energy at the nares, A_n, to the amplitude of the acoustic energy at the mouth, A_m.” The nasalance score is expressed as a percentage.

$$\frac{{\text{A}}_{\text{n}}}{\left({\text{A}}_{\text{m}}+{\text{A}}_{\text{n}}\right)}\times 100\%$$

Mean scores and ranges for multiple languages and dialects for a variety of verbal stimuli have been reported in the research. The important element for correlating a nasalance score with a perceptual estimate of nasality is to recognize that nasalance is the quantification of acoustic differences in sound pressure level that is recorded at the level of the nose and mouth, while nasality is determined by a trained listener’s judgement of the functionality of velopharyngeal coupling.

Resonant Voice

The American Speech-Language-Hearing Association states that: “A voice disorder occurs when voice quality, pitch, and loudness differ or are inappropriate for an individual’s age, gender, cultural background, or geographic location…. A voice disorder is present when an individual expresses concern about having an abnormal voice that does not meet daily needs—even if others do not perceive it as different or deviant”.⁷ Vocal tract resonance effects contribute to a properly produced and sounding voice.

In defining a “resonant voice”, there is a difference depending the approach (i.e., a voice clinician, a voice trainer, or a voice scientist). From the clinical perspective, a resonant voice is perceptually defined as “easy to produce and buzzy in the facial tissues” using “forward focus”, while in physical terms, it acts as “a reinforcement of the source by the vocal tract” (Titze & Verdolini Abbott, 2012). This is not to imply that vocal tract resonance isn’t occurring in regular speech but that a resonant voice is achieved when additional vocal tract shape modifications provide a greater overall resonance effect; these modifications usually consist of a more opened pharynx and a more occluded mouth opening than normal (Titze, 2001).

With coaching and practice, a resonant voice can have more radiated sound than non-resonant voice, so much so that it can be sensed in the facial tissues. This sensation can then provide a feedback cue during therapy sessions (Van Stan, et al., 2015), which can be particularly important for someone facing voice problems.

Biomechanically, resonant voice has been shown to be associated with a vocal fold configuration which is just barley adducted/abducted, and hence neither hypoadducted nor hyperadducted (Verdolini, Druker, Palmer, & Samawi, 1998; Peterson, Verdolini-Marston, Barkmeier, & Hoffman, 1994).

Acoustically, previous research has shown that the voice’s radiated output can be maximized by adjusting the harmonics and formants relationship (Titze, 2004, Titze, 2008). For example, by enhancing first formant tuning, the radiated voice spectra in the range of 2.0 to 3.5 kHz can be accentuated (Smith et al, 2005), which can increase perceived loudness. This, in turn, can lower the effort needed to produce voice in noisy situations (Schneider & Sataloff, 2007). This is the primary purpose of many resonant voice therapies (Edwin, et al, 2017) and the basis of several vocal exercises (Titze, Riede, & Popolo, 2008) where enhanced vocal resonance is desired (e.g. vocal performance, vocal limitations/disorders).

For most speech production, the fundamental frequency is much lower than the first vowel formant. But in cases where the fundamental frequency is in proximity to a vocal resonance – either in some compromised voices or during formant tuning in performance voice – increasing the coupling of the vocal source and the vocal tract can lead to vocal instabilities, such as voice breaks, subharmonics, and frequency jumps (Popolo, Titze, & Hunter, 2011). Acoustic and perceptual measures of these instabilities may provide a basis for tracking the extent of vocal fatigue and other vocal disorders (Halpern et al, 2009).

A voice user who inefficiently produces vocal resonance often experiences disproportionate vibratory sensations in the throat. Thus, for maximum efficiency in both speaking and singing, appropriate oral resonance should be fostered to boost vocal output while decreasing vocal stress (Verdolini et al., 1998).

Semi-occlusion of the Vocal Tract

Because the proper use of vocal resonance for sound propagation will both affect and be affected by voice disorders, it is general practice – regardless of the type of voice disorder or the reason for the limitation – that vocal therapy should include attempts to optimize vocal resonance through occlusion or semi-occlusion of the vocal tract (SOVT). This occlusion can take many forms, from lip-trills with the nose plugged to nasals with the lips closed. For example, patients using the Accent Method (AM) are asked to produce consonants such as voiced fricatives, which provide oral semi-occlusions in rhythmic vocalizations (Kotby & Fex, 1998), while in Lessac-Madsen Resonant Voice Therapy (LMRVT), they produce semi-occluded consonants such as fricatives and the nasals /m/, /n/, and /ŋ/ (Barrichelo-Lindstrom & Behlau, 2009; Orbelo, Li, & Verdolini Abbott, 2014; Verdolini, 2000). In Vocal Function Exercises (VFE), patients are asked to go through daily non-speech exercises that use semi-occlusion at the lips (Stemple, 2005). Furthermore, creating the semi-occlusion with the use of external devices such as phonation into straws or tubes to provide additional resistance during phonation to assist in training resonance voice has been well-documented (Habermann, 1980; Laukkanen, 1992; Story, Laukkanen, & Titze, 2000; Titze, 2002; Simberg & Laine, 2007). These techniques even include phonating through a tube in water (Enflo, et al, 2013). The benefits of vocal resonance therapies can be found throughout the literature (Laukkanen, Lindholm, & Vilkman, 1995; Titze & Laukkanen, 2007; Titze & Verdolini Abbott, 2012), including improved vocal quality (Edwin, et al, 2017) and facilitated learning of resonant voice techniques (Kapsner-Smith et al, 2015).

In summary, training the sound source to tune well to the vocal tract has been the basis of many successful vocal therapy and vocal enhancement programs. Many of these programs are based on semi-occluding the vocal tract in order to facilitate learning how to have a more resonant vocal production.

Summary

The source-filter theory emphasizes the fact that a talker who wishes to produce an intelligible and natural-sounding vowel or other vocalic segment must complete a two-step process (Hunter & Ludwigsen, 2017). Step one is phonation, which creates a tonal vibration in the larynx. This tone has a rich spectrum (components at every harmonic frequency), and a characteristic spectral shape (component amplitudes decreasing monotonically with increasing frequency). In step two, the phonated waveform is passed through the vocal tract and its spectrum is substantially modified according to the vocal tract’s resonant characteristics. The present article focused on step two of this process. Resonance effects uniquely associated with the voices of professional performers were reviewed. There are specific resonances that aid both singers, where they help the voice stand out against a background of accompanying music, and actors, where they help the voice project well and sound more satisfactory to a listening audience. Resonance and anti-resonance effects specifically associated with nasalization were reviewed, as were the clinical evaluation and treatment of hyponasal and hypernasal speech. Finally, we provided an overview of studies of resonant voice, as they pertain to the understanding of both normal and disordered vocal production, and to the enhancement of a talker’s vocal efficiency.