Nasalance, Nasal Airflow and Perceived Nasality in Carnatic Singers and Non-singers

Abstract

This study aims to compare the nasalance and nasal airflow between professional singers trained in Carnatic classical singing and non-singers. We also aimed to correlate perceived nasality with objective measurements of nasalance and nasal airflow. A total of 40 female participants (20 to 50 years) were involved in this study. The first group comprised 15 female professional Carnatic singers with a minimum of 10 years of classical training. The second group consisted of 25 non-singer females. These participants were compared on nasalance, nasal airflow and perceived nasality across three sets of stimuli (vowels, oral non-words, and nasal non-words) and three pitch-conditions (low, mid, and high). Correlations were also made between objective measures of nasalance and nasal airflow and perceived nasality. Mixed ANOVA showed a significant (P < 0.05) interaction in nasalance between pitch conditions and groups. Group differences were also observed in the nasalance scores of vowels, oral non-words, nasal non-words. There was a significant difference (P < 0.05) in nasalance with ascending pitch in singers and non-singers. A comparison of aerodynamic analysis of vowels, oral non-words, and nasal non-words between singers and non-singers suggested that nasal airflow was higher in singers. Perceptual nasality was significantly lower (P < 0.05) in Carnatic singers than non-singers. Correlations between objective measures of nasalance and nasal airflow and perceived nasality were not observed. Despite an increased airflow, nasality was lower in trained Carnatic singers than non-singers. Current findings suggest that vocal training impacts nasalance, nasal airflow and perceived nasality.

Introduction

Vocal resonance, enriches the timbre and intensity of sound made during phonation [1], and is crucial for singing and speaking. Nasal resonance, often considered desirable in singing, contributes to the quality and richness of the voice [2]. While nasal resonance occurs naturally in normal speakers, controversies exist regarding its use during singing, suggesting that singers possess refined control over velar adjustments [2]. Optimal vocal resonance, reflected like the singer’s formant [3], is a goal for classical singers who aim to enhance and amplify the fundamental pitch of their voice. Professional singers effectively utilize nasal resonance to enrich their voice, particularly in genres with high vocal demands, without imposing undue strain on the laryngeal system [4].

The use of nasal resonance during the production of non-nasal phonemes and the extent and manner of velar adjustments required to achieve desired vocal qualities are still not fully understood. Individual variability in using the velopharynx while singing have been reported [5–7]. Some have suggested that singers allow a small degree of velopharyngeal opening, leading to forward-focused resonance [7]. Nasalance studies have shown differences between amateur and trained singers, with several factors influencing nasality [8]. Despite decades of research into this phenomenon, it is still unclear whether nasalance results in perceptible nasality [9].

As pitch increases, there is an associated velar opening, but controlling nasalance and nasal airflow during singing remains important [6, 10]. Existing evidence suggests that trained singers can make skillful phonatory, resonatory, and articulatory adjustments different from non-singers [11, 12]. Velopharyngeal opening is negatively correlated with the perceived nasality in the singers singing in upper registers [5, 6, 13]. In contrast, comparisons between singers and non-singers have also revealed no differences in nasalance between singers and non-singers [14]. There is scarce literature on the correlation between nasalance and nasal airflow and the effect of singing training and experience on controlling nasalance with an open or lowered velum.

Carnatic singing, is a classical South Indian musical tradition, which necessitates rigorous training to attain a resonant and sonorous vocal quality known as the "open throated" technique [15]. Carnatic singing is characterized by predominantly low-pitched renditions while incorporating intricate vocal nuances in higher registers [15] and has a wide vocal range [16]. Earlier studies on Carnatic singers have reported their voice characteristics [15–18], awareness about vocal health [19–21], and documented improvements with therapeutic techniques [4, 22]. However, studies focusing on control of nasalance and nasal airflow in Carnatic singers are scarce. By investigating these aspects, contributions can be made to vocal pedagogy and may have implications for vocal training and preventing vocal injuries in Carnatic singers. Therefore, this study aims to compare the nasalance and nasal airflow between Carnatic singers and non-singers and to correlate perceptual measure of nasality with objective measurements of nasalance and nasal airflow.

Method

Study Design

This study involved a comparative study design. Two groups were formed based on the independent variable (Carnatic classical singing) and were compared on three major dependent variables (nasalance, nasal airflow and perceived nasality) across three sets of stimuli (vowels, oral non-words and nasal non-words) and three pitch-conditions (low, mid and high).

Participants

The present study considered two groups of non-smoking adult participants (20–50 years). Group A comprised 15 female professional Carnatic singers, with a minimum of 10 years of Carnatic classical training. Group B consisted of 25 age-matched female non-singers, with no formal vocal training or singing exposure. All participants were screened by a speech-language pathologist to assess their oral structure and function. Participants with an upper respiratory tract infection at the time of recording were excluded from the study.

Instrumentation

Nasalance was measured using Nasometer II (6400) (KayPENTAX (Lincoln Park, NJ)). Airflow measurement was obtained using the Aeroview system (v.1.5.0) (Glottal Enterprises Inc, Syracuse, NY). Before recording, both instruments were calibrated as per manufacturer recommendations. We also used a digital voice recorder (Olympus LS100 (Olympus Imaging America Inc., PA)) for voice recordings.

Procedure

All recordings were obtained in a quiet recording room. Before the actual recording, we involved the prospective participants in a practice session to familiarize them (especially the non-singers) with pitch-conditions. We used an audio–video material, wherein one of the authors (a professionally trained singer) modelled the tasks. In this model, pitch frequencies of 240 Hz, 360 Hz, and 482 Hz were used to represent low, mid, and high pitches, respectively, for vowels. Similarly, for oral non-words, pitch frequencies of 245 Hz, 373 Hz, and 497 Hz were used, while nasal non-words were associated with pitch frequencies of 250 Hz, 376 Hz, and 500 Hz. The participants were asked to produce and match the pitch for each pitch-condition. Participants who scored > 90% in this training were only considered for subsequent procedures.

Stimuli and Tasks

Both groups performed two tasks under three conditions (Table 1). Task I involved 10-s phonation of the vowels /a/, /i/, and /u/. This task was divided into three pitch-conditions: low, medium, and high. Task II required participants to sing oral (/pa:va/, /pi:va/, and /pu:va/) and nasal non-words (/ma:va/, /mi:va/ and /mu:va/) with ascending pitch. Similar to the first task, this task also involved three different pitch-conditions involving three frequencies (low, mid, and high).

Table 1.

List of stimuli for tasks I and II

	Task I				Task II
	Vowels				Non-words
Low-pitch	Mid-pitch	High-pitch	Low-pitch		Mid-pitch		High-pitch
/a/	/a/	/a/	/pava/a	/mava/b	/pava/a	/mava/b	/pava/a	/mava/b
/i/	/i/	/i/	/piva/a	/miva/b	/piva/a	/miva/b	/piva/a	/miva/b
/u/	/u/	/u/	/puva/a	/muva/b	/puva/a	/muva/b	/puva/a	/muva/b

^aOral non-words

^bNasal non-words

Nasalance Measurements

Participants were seated comfortably, and one of the authors explained the procedure. The headset was then positioned appropriately. The Olympus LS100 audio recorder was placed at an angle of 45 degrees from the mouth at a distance of 15 to 30 cm. The participants completed the first task (vowel phonation) before attempting the second task (oral and nasal non-word singing).

Aerodynamic Measurements

Participants were seated comfortably, and one of the authors explained the procedure to them. Then, the Rothenberg mask was tightly applied to cover both nasal and oral orifices to prevent air leakage. Before proceeding to Task II, all participants completed Task I in all three conditions. We continuously monitored the nasal airflow of the participants during recordings. Retrials were conducted for singing and phonation errors, and only the correct versions were considered for data collection. After completing both tasks, we extracted and stored the nasal airflow traces for further analysis.

Data Analysis

We examined only selected portions for mean nasalance and airflow from the data. We considered steady-state portions of vowel phonation from the first task at all three conditions. Similarly, from the second task, we considered the steady portion of the vowel, which follows the initial consonant (/p/ and /m/) for all three conditions. We calculated the mean nasalance value (%) and average nasal airflow (ml/sec) from the selected portions and subjected the obtained values to further statistical analysis.

Auditory-Perceptual Measurement

Sustained phonation of the vowel /a/ and singing task of the oral non-word /pava/ and nasal non-word /mava/ were subjected to auditory-perceptual measurement of nasality. Five Speech-Language Pathologists rated nasality for these stimuli on a 100 mm Visual Analog Scale. Each stimulus at all three pitches was rated separately.

Statistical Analysis

Obtained data were subjected to statistical analysis using SPSS version 26, R platform (v.4.3.0) (R Core Team, Vienna, Austria), and R Studio (v.2023.03.1 + 446) [23]. We examined the distribution of data based on which appropriate statistical analysis was carried out. An α level of 0.05 was maintained throughout this study.

Results

Nasalance Measurement

Data obtained from nasalance measurement followed normal distribution on administering Shapiro–Wilk’s test (P > 0.05). Hence, parametric tests were applied for nasalance measures.

Descriptive Statistics

Descriptive statistical measures were derived for vowels, oral, and nasal non-words for nasalance values (Table 2). The mean scores of the nasalance were higher in Group A than Group B except at high-pitch in all the vowels and oral non-words. It was also observed that the vowel /i/ had high nasalance in both groups. Unlike vowels and oral non-words, the nasal non-words showed no clear trend as the pitch varied in both groups. However, the nasalance value was high for /miva/ than the other nasal non-words.

Table 2.

Descriptive statistical measures for nasalance (%) values

Task I (Vowels)		Group Aa	Group Bb	Task II (Oral non-words)		Group Aa	Group Bb	Task II (Nasal non-words)		Group Aa	Group Bb
Stimuli	Condition	Mean (SD)	Mean (SD)	Stimuli	Condition	Mean (SD)	Mean (SD)	Stimuli	Condition	Mean (SD)	Mean (SD)
/a/	Low	34.33 (21.22)	31.28 (12.13)	/pava/	Low	28.20 (21.92)	26.28 (12.37)	/mava/	Low	49.33 (14.91)	54.96 (14.24)
	Mid	37.07 (19.87)	36.28 (15.56)		Mid	30.93 (19.11)	28.04 (13.60)		Mid	51.13 (14.62)	58.00 (15.10)
	High	30.73 (16.68)	42.68 (14.38)		High	29.13 (16.68)	36.20 (16.48)		High	41.87 (12.41)	62.24 (17.65)
/i/	Low	51.60 (28.40)	44.36 (17.62)	/piva/	Low	42.13 (23.63)	34.64 (16.49)	/miva/	Low	79.53 (10.93)	74.40 (10.38)
	Mid	52.93 (25.26)	43.60 (17.64)		Mid	46.13 (23.40)	39.76 (19.11)		Mid	74.80 (9.83)	74.96 (9.58)
	High	44.33 (24.93)	49.24 (18.75)		High	40.33 (22.86)	47.04 (18.28)		High	64.87 (13.39)	76.40 (13.42)
/u/	Low	29.20 (20.57)	23.48 (9.64)	/puva/	Low	22.33 (15.89)	19.24 (13.05)	/muva/	Low	61.20 (15.64)	57.96 (15.68)
	Mid	35.93 (19.30)	24.44 (13.17)		Mid	31.87 (17.78)	20.76 (14.32)		Mid	56.47 (15.03)	58.36 (16.97)
	High	31.07 (19.98)	32.88 (15.35)		High	25.60 (15.86)	33.96 (16.92)		High	45.20 (13.36)	60.44 (21.31)

^aSingers

^bNon-Singers

Nasalance Analysis

Nasalance in Vowels

Mixed ANOVA with two-way repeated measures ANOVA showed a significant difference between vowels (2, 76 = 34.26; P < 0.001, η_p² = 0.474). It also showed a significant interaction between pitch and groups (2, 76 = 9.05; P < 0.001, η_p² = 0.192) and pitch and vowel (4, 152 = 3.13; P = 0.01, η_p² = 0.076). This finding indicated that singers and non-singers behave differently as the pitch changes and that the nasalance of the vowels is different across the pitch. Hence repeated measure ANOVA was applied for both groups separately to find the difference in nasalance of vowels with a change in pitch.

Group A (Singers)

Repeated measures ANOVA for vowels across pitches showed that the vowel /i/ was the only vowel with a statistically significant difference in the nasalance across the different pitches (2, 28 = 3.71; P = 0.037, η_p² = 0.21). Post hoc multiple comparisons with Bonferroni adjustments suggested that only the mid and high-pitch for the vowel /i/ (P = 0.05) were significantly different.

Group B (Non-singers)

The vowels /a/ (2, 48 = 13.76; P < 0.001, η_p² = 0.364) and /u/ (2, 48 = 7.8; P = 0.001, η_p² = 0.245) were significantly different across the different pitches in non-singers. Post hoc multiple comparisons with Bonferroni adjustments indicated that nasalance values of /a/ were significantly different across low-pitch and high-pitch (P = 0.001) as well as mid-pitch and high-pitch (P = 0.004). Similarly, nasalance values of /u/ were significantly different across low-pitch and high-pitch (P = 0.01) and mid-pitch and high-pitch (P = 0.013) conditions.

Analysis Across Pitch-Conditions

Since significant differences among vowels were observed in the mixed ANOVA, the pitch was kept constant while performing a mixed ANOVA with repeated measures across the three vowels. Groups did not significantly affect vowel nasalance. Hence, both groups were combined for analysis across pitch-conditions. The results showed a significant difference in the nasalance of vowels at all three different pitches (low-pitch (2, 78 = 47.23; P < 0.001, η_p² = 0.548); mid-pitch (2, 78 = 21.59; P < 0.001, η_p² = 0.356); and high-pitch (2, 78 = 22.39; P < 0.001, η_p² = 0.365).

Post hoc multiple comparisons with Bonferroni adjustments suggested that in low-pitch, there was a significant difference in nasalance between vowels /a/ and /i/ (P < 0.001); /i/ and /u/ (P < 0.001) and /a/ and /u/ (P = 0.005). Similarly, in mid-pitch, there was a significant difference in nasalance between vowels /a/ and /i/ (P = 0.001); /i/ and /u/ (P < 0.001) and /a/ and /u/ (P = 0.028). Further, in high-pitch, there was a significant difference in nasalance between vowels /a/ and /i/ (P = 0.001); /i/ and /u/ (P < 0.001) and /a/ and /u/ (P = 0.028).

Nasalance in the Oral Non-words

Mixed ANOVA with two-way repeated measures ANOVA showed a significant difference between oral non-words (1.647, 62.577 = 37.069; P < 0.001, η_p² = 0.494) (with Greenhouse–Geisser correction) and between pitch-conditions (1.513, 57.476 = 5.586; P = 0.011, η_p² = 0.128) (with Greenhouse–Geisser correction). We also found a statistically significant interaction between pitch and groups (2, 76 = 7.195; P = 0.001, η_p² = 0.159). This shows the singers and non-singers behave differently as the pitch changes and that the nasalance of the oral non-words is different across pitch-conditions. Hence repeated measure ANOVA was carried out for both groups separately to find the difference in nasalance of each oral non-word with changes in pitch.

Group A (singers)

The results of repeated measure ANOVA showed no significant difference in the nasalance scores of the oral non-words across the three different pitches in singers.

Group B (Non-singers)

The results of repeated measures ANOVA of oral non-words at low, mid and high-pitch in non-singers showed a significant difference observed for the three oral non-words (/pava/ (1.290, 30.959 = 10.014; P = 0.002, η_p² = 0.294) (with Greenhouse–Geisser correction); /piva/ (1.472, 35.324 = 6.161; P = 0.01, η_p² = 0.204) (with Greenhouse–Geisser correction); /puva/ (1.605, 38.519 = 15.419; P < 0.01, η_p² = 0.391) (with Greenhouse–Geisser correction)) across the three different pitch-conditions.

Post hoc multiple comparisons with Bonferroni adjustments suggested a significant difference in nasalance between low and high (P = 0.009) and mid- and high-pitch-conditions (P < 0.001) for the oral non-word /pava/. Similarly, post hoc multiple comparisons with Bonferroni adjustments revealed that for the oral non-word /piva/, there was a significant difference in nasalance between low and high (P = 0.021) and mid- and high-pitch-conditions (P = 0.012). We also found a significant difference in nasalance between low and high (P = 0.001) and mid- and high-pitch-conditions (P = 0.001) for the oral non-word /puva/.

Nasalance in the Nasal Non-words

Mixed ANOVA with two-way repeated measures ANOVA was carried out for the analysis of nasal non-words showed significant differences across three pitch-conditions (2, 76 = 4.87; P = 0.01, η_p² = 0.114) and three nasal non-words (2, 76 = 111.67; P < 0.001, η_p² = 0.746). There was also a significant interaction between pitch-conditions and groups (2, 76 = 16.25; P < 0.001, η_p² = 0.299) and nasal non-words and group (2, 76 = 4.44; P = 0.015, η_p² = 0.105). We also noted a significant interaction between pitch-conditions and nasal-nonwords (4, 152 = 5.19; P = 0.001, η_p² = 0.120). Because of the interaction between pitch-conditions and groups, a pairwise comparison was made for both the pitch-conditions and nasal non-words separately in singers and non-singers.

Relating to pitch-conditions, post hoc multiple comparisons with Bonferroni adjustments identified a statistically significant difference in nasalance between low and high (P = 0.046) and mid- and high-pitch-conditions (P = 0.026). Similarly, post hoc multiple comparisons with Bonferroni adjustments showed that the mean scores of nasalance between the nasal non words /mava/-/miva/ (P < 0.001); /miva/-/muva/(P < 0.001); and /mava/-/muva/ (P = 0.018) were significantly different. Because of the interactions between pitch-conditions and groups as well as nasal non-words and groups, for further analysis, the groups were considered separately.

Group A (Singers)

Repeated measures ANOVA in singers showed that the nasalance of each nasal non-word varied significantly across the three different pitch-conditions (low (2, 28 = 73.67; P < 0.001, η_p² = 0.84); mid (2, 28 = 28.83; P < 0.001, η_p² = 0.673); high (2, 28 = 47.19; P < 0.001, η_p² = 0.771)). Further post hoc multiple comparisons with Bonferroni adjustments revealed that in low-pitch condition, the nasalance scores between the nasal non-words /mava/ and /miva/ (P < 0.001); /mava/ and /muva/ (P = 0.005); and /miva/ and /muva/ (P < 0.001) were significantly different. In both mid-pitch and high-pitch-conditions, the nasalance scores between nasal non-words /mava/ and /miva/ (P < 0.001) and /miva/ and /muva/ (P < 0.001) showed significant differences.

Since mixed ANOVA had shown a statistically significant interaction between pitch-conditions and groups, a repeated measure ANOVA was carried out separately for singers across pitch-conditions in each nasal non-word. It was found that all nasal non-words (/mava/ (2, 28 = 10.84; P < 0.001, η_p² = 0.437); /miva/ (2, 28 = 23.55; P < 0.001, η_p² = 0.627); and /muva/ (2, 28 = 14.28; P < 0.001, η_p² = 0.505)) varied significantly across pitch-conditions. Post hoc multiple comparisons with Bonferroni adjustments revealed that for the nasal non-word /mava/ there was a significant difference between low and high (P = 0.006), mid and high pitch (P = 0.008) conditions. Similarly, for the nasal non-word /miva/, there was a significant difference between low and high (P < 0.001), mid and high pitch (P < 0.001) conditions. Further, for the nasal non-word /muva/, there was a significant difference between low and high (P < 0.001), mid and high pitch (P = 0.007) conditions.

Group B (Non-singers)

On administering repeated measures ANOVA in non-singers we found that the nasalance of each nasal non-word varied significantly across the three different pitches (/mava/ (2, 48 = 51.99; P < 0.001, η_p² = 0.684); /miva/ (1.626, 39.022 = 36.514; P < 0.001, η_p² = 0.603) (with Greenhouse–Geisser correction); and /muva/ (2, 48 = 27.43; P < 0.001, η_p² = 0.533)). Further post hoc multiple comparisons with Bonferroni adjustments revealed that in all three pitch-conditions, the nasalance scores between the nasal non-words /mava/ and /miva/ (P < 0.001) and /miva/ and /muva/ (P < 0.001) were significantly different.

Since mixed ANOVA showed a significant interaction between pitch-conditions and groups, a repeated measure ANOVA was done separately for non-singers across pitch-conditions within each nasal non-word. However, only /mava/ (2, 48 = 6.53; P = 0.003, η_p² = 0.214) was found to differ significantly across pitch levels. Post hoc multiple comparisons with Bonferroni adjustments suggested that for the nasal non-word /mava/, there was a significant difference only between low- and high-pitch (P = 0.012) conditions.

Nasal Airflow Analysis

Data obtained from airflow measurement followed non-normal distribution on the administration of Shapiro–Wilk’s test (P < 0.05). Hence, we applied non-parametric tests for airflow measures.

Comparison of Nasal Airflow of Vowels Across Groups

Descriptive statistical measures were derived for vowels, oral non-words and nasal non-words for nasal airflow values (Table 3). Mann–Whitney U test was done to compare the nasal airflow between the two groups. However, we did not find a statistically significant difference between the groups across the three pitch-conditions for vowels. Similarly, no significant difference between the groups in oral non-words at different pitches was observed. However, a significant difference was only found in the low-pitch /muva/ (P < 0.01) for nasal non-words. Although insignificant, contexts with the vowel /i/ showed more nasal airflow than other vowels in both groups in different conditions.

Table 3.

Descriptive statistical measures and comparison of nasal airflow (ml/sec) values across groups

Task	Stimuli	Condition	Group Aa		Group Bb		Z	P-value
			Median	IQR	Median	IQR
I (Vowels)	/a/	Low	27.30	93.30	17.90	42.02	− 1.467	0.142
		Mid	31.80	74.55	34.20	55.95	− 0.237	0.812
		High	47.30	149.80	43.10	64.05	− 0.014	0.989
	/i/	Low	27.70	154.86	19.80	50.86	− 0.894	0.371
		Mid	60.30	238.70	33.20	66.90	− 1.076	0.282
		High	44.80	259.20	55.90	99.65	− 0.433	0.665
	/u/	Low	20.10	38.58	10.20	13.35	− 0.852	0.394
		Mid	7.82	10.97	11.20	20.07	− 0.321	0.748
		High	34.80	139.21	14.20	46.82	− 0.740	0.459
II (Oral Non-words)	/pava/	Low	17.00	76.42	10.80	15.14	− 1.523	0.128
		Mid	34.10	96.41	19.50	23.92	− 1.383	0.167
		High	73.60	173.53	31.30	45.42	− 1.020	0.308
	/piva/	Low	13.20	103.09	24.50	45.48	− 0.265	0.791
		Mid	12.30	163.11	16.70	59.76	− 0.265	0.791
		High	23.30	86.87	38.80	76.17	− 0.545	0.586
	/puva/	Low	6.64	31.64	7.83	14.13	− 0.629	0.530
		Mid	8.61	11.97	12.60	26.29	− 1.131	0.258
		High	14.00	31.44	13.30	27.05	− 0.349	0.727
II (Nasal Non-words)	/mava/	Low	117.00	224.20	93.20	95.10	− 1.020	0.308
		Mid	90.20	125.00	121.40	163.34	− 0.489	0.625
		High	114.00	282.60	147.00	147.74	− 0.377	0.706
	/miva/	Low	180.00	239.00	140.00	133.50	− 1.299	0.194
		Mid	235.00	132.00	176.00	138.63	− 1.662	0.096
		High	163.00	232.70	184.00	170.70	− 0.168	0.867
	/muva/	Low	194.00	126.00	96.40	111.20	− 3.031	0.002*
		Mid	129.00	134.00	119.00	156.48	− 0.279	0.780
		High	60.50	130.70	107.00	206.70	− 0.545	0.586

^aSingers; ^bNon-Singers, *significant at P < 0.05

Nasal Airflow of Group A (Singers)

Using Wilcoxon Signed Rank test, the nasal airflow within both groups were analyzed across vowels, oral non-words and nasal non-word stimuli. However, we found only minimal variations in nasal flow among singers for vowels, oral and nasal non-words.

Vowels

Results showed that most comparisons across vowel and pitch did not differ statistically. Only mid-pitch /i/ Vs mid-pitch /a/ (Z = − 2.158, P = 0.031) and high-pitch /u/ Vs mid-pitch /u/ (Z = − 2.669, P = 0.008) showed a significant difference in nasal airflow of singers.

Oral Non-words

Results showed that most comparisons across oral non-word and pitch did not differ significantly. This finding indicates that singers maintained the nasal airflow for oral non-words across the pitch-conditions without much variation. Only low-pitch /puva/ Vs low-pitch /piva/ (Z = − 2.726, P = 0.006), mid-pitch /puva/ Vs mid-pitch /pava/ (Z = − 2.385, P = 0.017), high-pitch /puva/ Vs high-pitch /pava/ (Z = − 2.442, P = 0.015) and high-pitch /puva/ Vs mid-pitch /puva/ (Z = − 2.215, P = 0.027) showed a statistically significant difference in nasal airflow of singers.

Nasal Non-words

Results showed that most comparisons across nasal non-word and pitch did not differ significantly. This finding suggested that singers maintained the nasal airflow for nasal non-words across pitch-conditions without much variation. Only low-pitch /miva/ Vs low-pitch /mava/ (Z = − 2.101, P = 0.036), mid-pitch /miva/ vs /mava/ (Z = − 2.329, P = 0.020) mid-pitch /muva/ Vs mid-pitch /miva/ (Z = − 2.499, P = 0.012), high-pitch /miva/ Vs high-pitch /muva/ (Z = − 2.442, P = 0.015) and high-pitch /muva/ Vs low-pitch /muva/ (Z = − 2.385, P = 0.017) showed a statistically significant difference in nasal airflow of singers.

Nasal Airflow of Group B (Non-singers)

Similar to group A, the Wilcoxon Signed Rank test was utilized to analyze the nasal airflow within group B across vowels, oral non-words and nasal non-words (Table 4). Most comparisons across vowels, oral and nasal non-words and pitch differed significantly among non-singers. Variation in nasal airflow was thus deemed higher for non-singers.

Table 4.

Conditions with statistically significant differences in nasal airflow (ml/sec) values for Group B (Non-singers)

Task I comparisons				Task II comparisons
Stimuli	Z	P-value	Stimuli	Z	P-value	Stimuli	Z	P-value
Low-pitch /u/—Low-pitch /i/	− 3.35	0.001	Low-pitch /piva/—Low-pitch /pava/	− 3.108	0.002	Low-pitch /miva/—Low-pitch /mava/	− 3.767	< 0.001
Mid-pitch /u/—Mid-pitch /i/	− 3.996	< 0.001	Low-pitch /puva/—Low-pitch /piva/	− 3.565	< 0.001	Low-pitch /muva/—Low-pitch /miva/	− 2.785	0.005
Mid-pitch /u/—Mid-pitch /a/	− 3.646	< 0.001	Mid-pitch /puva/—Mid-pitch /piva/	− 2.301	0.021	Mid-pitch /miva/—Mid-pitch /mava/	− 2.543	0.011
High-pitch /u/—High-pitch /i/	− 2.812	0.005	High-pitch /puva/—High-pitch /piva/	− 3.108	0.002	Mid-pitch /muva/—Mid-pitch /miva/	− 2.112	0.035
High-pitch /u/—High-pitch /a/	− 2.65	0.008	High-pitch /puva/—High-pitch /pava/	− 2.166	0.03	High-pitch /miva/—High-pitch /mava/	− 2.166	0.03
Mid-pitch /a/—Low-pitch /a/	− 3.431	0.001	Mid-pitch /pava/—Low-pitch /pava/	− 3.027	0.002	High-pitch /muva/—High-pitch /miva/	− 2.112	0.035
High-pitch /a/—Mid-pitch /a/	− 2.301	0.021	High-pitch /pava/—Mid-pitch /pava/	− 2.422	0.015	Mid-pitch /mava/—Low-pitch /mava/	− 2.892	0.004
High-pitch /a/—Low-pitch /a/	− 3.592	< 0.001	High-pitch /pava/—Low-pitch /pava/	− 3.7	< 0.001	High-pitch /mava/—Mid-pitch /mava/	− 2.274	0.023
High-pitch /i/—Mid-pitch /i/	− 3.269	0.001	High-pitch /piva/—Mid-pitch /piva/	− 3.108	0.002	High-pitch /mava/—Low-pitch /mava/	− 3.162	0.002
High-pitch /i/—Low-pitch /i/	− 3.431	0.001	High-pitch /piva/—Low-pitch /piva/	− 2.139	0.032
High-pitch /u/—Mid-pitch /u/	− 3.646	< 0.001	Mid-pitch /puva/—Low-pitch /puva/	− 2.906	0.004
High-pitch /u/—Low-pitch /u/	− 3.511	< 0.001	High-pitch /puva/—Mid-pitch /puva/	− 2.18	0.029
			High-pitch /puva/—Low-pitch /puva/	− 3.484	< 0.001

Auditory-Perceptual Measurement

For analysis of auditory-perceptual measurement, the average perceptual nasality was calculated from the ratings of five SLPs for the vowel /a/, the oral non-word /pava/ and the nasal non-word /mava/ at three different pitches. Then the mean perceptual nasality was compared across the groups (Fig. 1). The findings showed that nasality was higher in non-singers than singers and increased with pitch-conditions. Pairwise comparisons across pitch-conditions revealed statistically significant differences in nasality in both groups (Table 5). Further, between-group comparisons suggested significant differences between the groups on perceived nasality for mid-pitch and high-pitch /pava/ and on high-pitch /mava/ (Fig. 1).

Fig. 1.

Mean Nasality of singers and non-singers across different stimuli and pitch-conditions

Table 5.

Pairwise comparisons of perceived nasality across various pitch conditions within both the groups

Condition	Group Aa		Group Bb
	Z value	P-value	Z value	P-value
Mid-pitch /a/—Low-pitch /a/	− 3.302	0.001	− 4.377	< 0.001
High-pitch /a/—Low-pitch /a/	− 3.41	0.001	− 4.374	< 0.001
High-pitch /a/—Mid-pitch /a/	− 3.355	0.001	− 4.375	< 0.001
Mid-pitch /pava/—Low-pitch /pava/	− 3.182	0.001	− 4.289	< 0.001
High-pitch /pava/—Low-pitch /pava/	− 3.411	0.001	− 4.373	< 0.001
High-pitch /pava/—Mid-pitch /pava/	− 3.298	0.001	− 4.376	< 0.001
Mid-pitch /mava/—Low-pitch /mava/	− 3.422	0.001	− 4.011	< 0.001
High-pitch /mava/—Low-pitch /mava/	− 3.409	0.001	− 4.374	< 0.001
High-pitch /mava/—Mid-pitch /mava/	− 2.557	0.011	− 4.347	< 0.001

^aSingers; ^bNonsingers; All comparisons were significant at P < 0.05 levels

Correlation of Nasalance and Nasal Airflow with Perceived Nasality

Perceived nasality was correlated with nasalance (Fig. 2) and nasal airflow (Fig. 3) using Spearman’s correlation. Both groups showed no correlation between these measurements. However, nasal airflow and perceptual measures were weakly correlated for both groups. Although statistically insignificant, singers had a more negative correlation between nasal airflow and perceived nasality than non-singers.

Fig. 2.

Correlation between perceived nasality and nasalance across different stimuli and pitch-conditions

Fig. 3.

Correlation between perceived nasality and nasal airflow across different stimuli and pitch-conditions

Discussion

Nasalance in the Vowels

We found a significant interaction in nasalance between pitch and groups, indicating that the effect of frequency variation on nasalance differs between the groups for vowels. Additionally, there was a group difference in nasalance scores for vowels, necessitating separate analysis of pitch for each group.

Group A

The nasalance scores of the vowel /i/ varied between mid and high pitches, with mean scores decreasing at high pitch. Previous studies have shown that as pitch increases, the velopharynx narrows and nasality decreases due to changes in vocal intensity and velar configuration [10, 24, 25]. However, studies have found that certain singers, such as tenors, use nasalance to change timbre at higher pitches [6, 7, 26]. Overall, current findings suggest that Carnatic singers effectively adjust their velar configuration to restrict excess air passage through the nasal cavity at high-pitch.

Group B

Similar to the findings of the present study, researchers have found that untrained singers exhibit increased nasalance for the vowel /i/ compared to other vowels [8, 14]. In contrast, trained singers can control their nasality through variable velar patterns [5]. A mixed ANOVA with repeated measures in this study showed significant differences in nasalance between vowels at all pitches, with the highest nasalance observed for the vowel /i/, followed by /a/ and /u/. This finding is consistent with existing literature [8, 27, 28]. Taken together, these findings suggest that untrained singers may have an inefficient velar adjustment at higher pitches.

Analysis Across Pitch-Conditions

Variations in nasalance across pitch-conditions were found in this study. Increased oral impedance to high vowels results in greater nasality than back vowels. The difference in nasalance between the groups in this study may be attributed to vocal training, which equips singers to control nasality through velopharyngeal opening and airflow. This mechanism allows trained singers to produce resonance rather than nasality, as perceptual nasality in singing is often considered unappealing. Trained singers are known to utilize velopharyngeal opening (VPO) to ease register transitions [29]. Literature also evidences that, while singing across passaggio, male singers utilize nasalance to maintain oscillatory regularity [30]. Hence it can be thought that even the Carnatic singers utilize the VPO to maintain the pitch.

Nasalance in the Oral Non-words

In the oral non-words, mixed ANOVA showed a significant interaction across the pitch and groups. Hence, differences in the nasalance scores across the pitch in singers and non-singers were explored separately.

Group A

The nasalance scores of oral non-words did not vary significantly across pitches in this study. This finding may be due to the requirement for tighter velopharyngeal closure for voiceless consonants [31]. The lack of significant difference in nasalance scores across pitch-conditions in this study suggests that singers used a consistent velar configuration [24].

Group B

In non-singers, nasalance scores of oral non-words showed significant differences across pitches, with pairwise comparison revealing substantial differences between mid- and high-pitch and low- and high-pitch-conditions. This finding of the current study is consistent with previous studies in singers [8], and cleft population [32], that lowering pitch decreases nasalance scores. The increase in nasality in non-singers as pitch increased from low to high may be attributed to a lack of appropriate training and inherent vocal abilities.

Nasalance in the Nasal Non-words

In this study, a mixed ANOVA performed on nasal non-words showed significant differences in pitch and vowels. The pairwise comparisons revealed statistically significant differences between mid and high-pitch and low and high-pitch for both the groups, with singers exhibiting lower mean nasalance scores. Carryover nasal airflow from the preceding consonant /m/ is more remarkable in singing than in speaking [7]. This co-articulatory nasalization [33] results in high nasalance values for nasal non-words. However, the low mean nasalance scores in singers in the present study suggest that they regulate nasalance even in nasal contexts with velar control.

In this study, nasalance scores in singers were generally lower than in non-singers. This finding may be due to experienced singers regulating nasal airflow and using nasal resonance to brighten their voices [34]. An interaction between pitch and vowels was also observed in the present study, with significant differences in nasalance scores for nasal non-words /miva/ and /muva/ across pitch-conditions. Similar to available literature, we also found that the vowel /i/ had the highest nasalance among all nasal non-words in all pitch-conditions [8, 27, 28]. This may be due to decreased oral intensity and increased nasal intensity in high front vowels [35]. Further analysis was performed separately for singers and non-singers due to significant interactions between pitch and groups and vowel and groups.

Group A

Current findings showed significant differences in nasalance scores across vowels and pitch in singers for nasal non-words. This result may be due to the influence of vowels on nasalance scores [8, 35]. Similar to existing literature, a consistent decrease in nasalance scores with high-pitch was observed for singers in current study as well [7, 10]. This may be attributed to the forward focus used by classically trained singers to increase resonance, reducing vocal fold impact stress [36–39]. Recent evidence also suggests that using forward focus leads to lower nasalance scores [40].

Group B

Phonetic environment and co-articulatory nasalization could influence the untrained velum and explains the significant nasalance score differences between vowels in non-singers involved in the present study. Across different pitches, only the nasal non-word /mava/ varied significantly, with nasalance values increasing as pitch increased [8]. This may be due to an increase in intensity at higher scales.

Comparison of Nasal Airflow Analysis Between the Groups

A comparison of aerodynamic analysis between singers and non-singers involved in this study suggested that overall nasal airflow is higher in singers. This may be due to the greater use of VPO by singers during singing [6] Carnatic singers use nasal murmurs as exercises to initiate phonation. VPO has been found to exist in classical singers through various methods, including acoustic effects [2, 29], imaging [13, 41, 42], and nasoendoscopy [5, 6]. A narrow VPO can be regarded as a slit, and the acoustic impedance of VPO increases as frequency increases [6]. Thus, this study confirms that singers carefully adjust VPO opening to colour their timbre without nasal quality [6]. Thus, this study opposes the view that VP is closed during classical singing [43].

Even though nasal airflow is higher in singers, perceived nasality has no significant difference compared to non-singers [6]. This suggests that trained Carnatic singers can control their velum movement or make appropriate physiological adjustments/compensation of the vocal tract to project their voice. These physiological adjustments/compensations need to be explored further using appropriate speech physiological measurements. In this study, nasal airflow was more varied in Group B. This finding could be because non-singers cannot control their velum and maintain constant airflow while singing. However, trained singers master finer control of velum with training thereby limiting the variability in nasal airflow for different stimuli and pitch.

Vowels

We found no statistically significant differences between Carnatic singers and non-singers for vowels. Carnatic singers involved in this study consistently maintained the nasal airflow across pitch and vowels which is thought as a result of singing training. Singers adjust their velum to avoid excessive contraction of pharyngeal and velar muscles [6, 13]. Perceptually, voices are rated as beautiful and powerful when sung with a lowered velum and dull when sung with a raised velum [6, 13, 16, 17]. Nasal resonance is accepted and used in singing, and singing teachers and voice clinicians should emphasize on forward rather than backward vocal quality [40].

Oral Non-words

Findings for oral non-words in this study showed no statistically significant difference between the groups. However, both groups had higher nasal airflow for the stimulus /piva/ across pitch-conditions. This outcome is consistent with earlier findings [8, 14], which suggested that anterior constriction and greater oral impedance in the vowel /i/ result in greater nasal airflow. This finding may also apply to the production of /piva/, where the combined effect of higher intra-oral pressure of /p/ and oral impedance and tongue constriction of /i/ result in increased nasal airflow. The production of stop consonants is considered the point of maximum VP closure and maximum intra-oral pressure.

Nasal Non-words

Similar to vowels and oral non-words, the nasal non-word /miva/ had a greater nasal airflow in both groups. We found a significant difference between the groups for low-pitch /muva/. As pitch increased from low to high, a decrease in nasal airflow was observed for /miva/ and /muva/, suggesting that singers restrict their nasal airflow as the pitch increases [13]. In contrast to non-singers, Carnatic singers showed minimal variation in nasal airflow across pitch and vowels. This finding suggests that singers can maintain nasal airflow through controlled velar adjustments with varying pitch-conditions [7].

Perceptual Analysis

We found that non-singers had higher perceptual nasality than singers. This finding from the present study is correlated with the effect of vocal training on nasality, i.e., the singers use their VPO to enhance their singing voice without changing their perceptual nasality [9]. As pitch increases, perceived nasality also increases in both groups.

Correlation of Nasalance and Nasal Airflow with Perceived Nasality

Spearman’s correlation coefficient showed no correlation between perceived nasality and nasalance measurements and nasal airflow. Birch et al. [6], found no correlation between the nasal airflow and the perceived nasality in singers. Their findings revealed that the degree of perceived nasal quality is not related to a velopharyngeal opening. In contrast, the findings of the present study reveal that the objective measurements of nasalance and nasal airflow; and the perceived nasality are not correlated. However, studies have reported that nasalance and perceived nasality are often correlated in other populations such as repaired cleft palate [44].

Limitations and Future Directions

The restriction to achieve a wide oral opening using the Rothenberg mask poses a significant limitation to this study, as it prevents the exploration of certain vocal coloring techniques employed by singers. Simultaneous recordings of nasalance and airflow measurements would have yielded more meaningful inferences. Future studies can replicate the current methodology with a larger sample size, while also investigating the effects of vocal training on vocal range, quality, and endurance.

Conclusion

The findings of this study revealed that Carnatic singers utilize nasal airflow to enhance resonance through sympathetic vibrations in the nasal cavity, as indicated by increased airflow and nasalance scores. Simultaneously, these individuals effectively regulate and modify the velum and airflow, decreasing perceived nasality. The assumption that a positive correlation exists between VPO, nasalance, nasal airflow, and perceived nasality may not always hold true among Carnatic singers.

Author Contributions

TJ contributed to conceptualization, planning, data collection and analysis, and report writing. VVS, VRT contributed to data collection and analysis, and report writing. JJB contributed to data analysis, and report writing.

Funding

No funding was received for conducting this study.

Declarations

Disclosure

The authors have no conflicts of interest to declare.

Ethical standard

The procedures of this study adhered to ethical regulations set by the institutional ethical committee. We obtained informed consent from all the participants before initiating the procedures of this study. The authors have no ethical conflicts to disclose.