Variation in nasality between nasal and nasalized vowels in in Brazilian Portuguese: a pilot study
Luciana Ferreira Marques
University of Colorado, Boulder
Nasality is one of the most characteristic and controversial aspects of Portuguese. This feature has been studied from different perspectives. Phonologically, studies have proposed that nasality in a vowel is a binary feature that can emerge as an output of a nasalization rule, in which the nasalized vowel shares the property [+nasal] with the following nasal consonant. This is the case of a nasalized vowel, as in [ˈkɐ̃.mɐ] cama “bed”. It is regarded as a coarticulation process, and it is rather different from nasality from phonemic nasal vowels, in which the vowel is also nasalized as part of a nasalization process, but the conditioning nasal consonant may be deleted on the derivation process, yielding a full nasal vowel, as in [ˈkɐ̃ⁿ.tʊ] canto “singing” (Camara Jr, 1977). Regardless of the phonological status, the feature [+nasal] is absolute and the vowel can have it or not (Bisol, 1998).
Phonetically, that does not seem to be the case. The degree of nasality between phonemically nasal and coarticulatory nasalized vowels can vary, depending on the degree of the velopharyngeal port opening (Cagliari, 1977). Aerodynamic studies show that nasal vowels have a higher nasal airflow than nasalized vowel (Medeiros, 2011). Therefore, nasality should not be treated as an all-or-none phenomenon, as it intends strict phonological approaches. Because of that, phonemically nasal and coarticulatory nasalized vowels should also be treated differently, unlike most treatments given by mainstream phonological theories (see for example Mateus, 1982).
Coarticulation, such as the one occurring in [ˈkɐ̃.mɐ], is thought to aid in the perception of speech, because it can be controlled according to the phonemic distinctions that a language makes (Manuel, 1990). Furthermore, it can also be controlled as to provide information to the listener about the upcoming sounds. Based on these premises, one should expect the nasality in vowels to be informative for speech perception, as well as controlled not only for that purpose, but also because Brazilian Portuguese is generally known to have contrastive nasal vowels. This means that nasal coarticulation must be controlled to maintain the phonemic distinctions between nasal and nasalized vowels. However, given the fact that both nasalized and phonemic nasal vowels have the same phonological feature [+nasal], how can a speaker of Brazilian Portuguese differentiate between these two types of vowels and anticipate whether the upcoming sound in the utterance is a nasal consonant or not? Perhaps the aforementioned difference in degree of nasality from experimental evidence is a cue to it.
Even though a perceptual study is not the focus of this paper, it certainly has implications for it. In order for nasality in a vowel to be informative of the upcoming sound, there must be some kind of difference between nasalized and nasal vowels in Brazilian Portuguese. Thus, this study hypothesizes that there is such a difference, and it is due to the effect of coarticulation on the nasalized vowel. If coarticulation is constrained to enhance contrast, we should expect a difference in nasality between nasal and nasalized vowels in Brazilian Portuguese. To test this hypothesis, the methodology presented in Chen (1997) was used. The author shows that the difference between the first formant (F1) amplitude (A1) and the nasal peak amplitude (P0) is an indirect measure of velopharyngeal port opening. The smaller the difference, the more open the port, and therefore the more nasal a vowel is.
This methodology was tested on a set of words containing the low vowel /a/ (nasal [ɐ̃]) in the three contexts, all produced in sentences by two native speakers of Brazilian Portuguese. It is expected, then, that the low nasal vowel in Brazilian Portuguese have the smallest A1-P0 difference, followed by nasalized vowel. The oral vowel will have the highest A1-P0 difference because the velopharyngeal port in this vowel is expected to be closed during all its production.
Nasality in Acoustic terms
While most studies on quantifying nasalization rely on aerodynamic data, Chen (1997) proposes an acoustic correlate for nasality that can be fairly easy to measure. The acoustic consequences of velopharyngeal coupling in the spectrum are reduction of first formant (F1) amplitude, and the emergence of a low-frequency prominent peak in the spectrum of a nasal vowel. Chen calls the amplitude of the first formant A1, and the amplitude of the prominent nasal peak P0. The difference between these two consequences is the acoustic correlate of nasality. The author states that nasal vowels would have the smallest difference between A1 and P0. Thus, for oral vowels, one should assume the difference is higher. However, it is difficult to assume anything for nasalized vowels. The picture below, taken from Chen (1997), illustrates this point:
Fig 01 Spectral slices from oral and nasal vowels. The peaks used in A1-P0 and A1-P1 measures are labeled. Taken from Chen (1997).
The P0 peak could be either the first or the second harmonic, depending on which one has the greatest amplitude. Chen applies this technique to English and French. English does not have distinctive nasality, so the comparison was made between nasal and oral vowels. The author measured A1-P0 and A1-P1 (amplitude of a higher peak used when the peaks for A1 and P0 coincide) for the vowels /i/, /u/, /ɛ/, /ʌ/, /æ/ and /ɑ/. In general terms, A1-P0 was successful for low vowels /æ/ and /ɑ/, whose difference between oral and nasal was significant, ranging from 6 to 8 dB. The remaining higher vowels had better results with A1-P1, ranging from 10 to 15 dB.
For French, Chen measured only nasal vowels at different points to observe the increase in nasality each time. The author measured A1-P0 and A1-P1 for the vowels /ɛ̃/, /ɑ̃/, /ɔ̃/ and /œ̃/. It was hypothesized that nasality increases as the velopharyngeal port opens, thus the differences in amplitude would decrease each time. The results showed that A1-P0 and A1-P1 values at the beginning and at the end of the vowel were significantly different, ranging from 3 to 9dB for A1-P0 and 9 to 12 dB for A1-P1.
The results in Chen (1997) are very robust and promising. The author provides a reliable indirect way to measure velopharyngeal port opening. However, no comparison between oral and nasal vowels in French was made. Thus, the present paper can test the validity of the A1-P0 measure in new contexts, by comparing nasality measures in three different contexts. Before doing that, it is necessary to give an overall description of oral, nasal and nasalized vowels in the language under analysis, Brazilian Portuguese.
Oral, nasal and nasalized vowels in Brazilian Portuguese
Perhaps the most accurate phonetic description of vowels in Brazilian Portuguese is given in Silva (1998). The following is a summary of the characteristics of Brazilian Portuguese vowels based on Silva’s descriptions.
1. Oral vowels
Brazilian Portuguese has seven phonemic oral vowels, according to the following table:
Table 01 Brazilian Portuguese oral vowels.
They are identical in all dialects of Brazilian Portuguese. The back vowels are all rounded, and there is no length distinction. However, there is a stress-related reduction in the inventory. In pre-stressed syllables, mid-low and mid-high vowels are neutralized, and the actual production will depend on dialectal variation. In post-stressed final positions, the inventory is reduced to three vowels [ɪ], [ɐ] and [ʊ].
Acoustically, oral vowels are characterized by their first and second formants (F1 and F2). The table below shows the average values for F1 and F2 in Brazilian Portuguese for female speakers (Rauber, 2008):
Table 02 Average F1 and F2 for female speakers in Hertz.
These values are similar to formant values found by other studies, such as Cagliari (1977) and Sousa (1994).
2. Nasal vowels
Brazilian Portuguese has contrastive nasality. However, there is not a one-to-one correspondence between oral and nasal vowels. This language has five phonemic nasal vowels, as shown in the table below:
Table 03 Nasal vowels in Brazilian Portuguese.
Unlike the low oral vowels, nasal vowels show more variability in the different dialects of Brazilian Portuguese. The vowel /ã/ can have different transcriptions. Medeiros (2007) transcribes this vowel as [ɜ̃]. The mid-high vowels /ẽ/ and /õ/ become diphthongs /ẽɪ̃/ and /õɪ̃/ in some dialects. Also, in some dialects non-final nasal vowels are sometimes followed by what is called a nasal appendix, a consonantal sound that has the same place of articulation as the following consonant. Silva (1998) argues that, because this appendix does not appear systematically, it should not be considered part of the vowel. This is the approach used in this paper.
Acoustically, nasal vowels show the following average F1 and F2 (Medeiros, 2010).
Table 04 Average F1 and F2 in Hertz.
Medeiros analyzed only three vowels because the mid-high vowels become diphthongs in the Paulista dialect, which she analyzed. In terms of frequency, a lowering of F1 is observed for the low vowel, while it increases for the other two. F2 is a little lower comparatively for all nasal vowels.
Sousa (1994) points out that, besides changes in formant values, nasal vowels also show the appearance of a nasal formant around the frequency of F1. This could correspond to P0 in Chen (1997), since formants are essentially frequencies with enhanced amplitudes. In terms of amplitude, Sousa points out that, in general, nasal vowel formants present lower amplitudes when compared to oral vowels. These results are consistent with the methodology in Chen (1997).
Aerodynamically, nasal vowels are characterized by an increase in nasal airflow, without necessarily decreasing the amount of oral airflow, when compared to oral vowels. Low nasal vowels have higher nasal air flow than high nasal vowels (Medeiros, 2011).
3. Nasalized vowels
Nasalized, or coarticulatory nasal vowels are different from nasal vowels in that they must have a preceding or following fully realized and non-tautosyllabic nasal consonant, independently of stress. Thus, phonologically, they are oral vowels nasalized by coarticulation. There is not an inventory of nasalized vowels in Brazilian Portuguese because, in principle, every oral vowel can be nasalized.
Acoustically nasalized vowels are characterized by formant and amplitude changes similar to phonemic nasal vowels. However, changes in vowel quality from nasal to nasalized vowels do not happen; the only change in quality occurs from oral to nasalized vowels, as it occurs from oral to nasal vowels. That is because the velopharyngeal port is open in both types of vowels.
Aerodynamically, nasalized vowels also have higher nasal airflow than oral airflow, but lower nasal airflow than phonemic nasal vowels. As with nasal vowels, low nasalized vowels have higher nasal airflow than high nasalized vowels (Medeiros, 2011). The author states that this is the main difference between nasal and nasalized vowels. In terms of perception, then there should not be much difference between these two types of vowels.
This study was realized under the premise that coarticulation is informative and can be controlled for certain linguistic purposes. Manuel (1990) demonstrates that coarticulation is constrained depending on the inventory of distinctive sounds in a given language. “Languages will tend to tolerate less coarticulation just where extensive coarticulation would lead to confusion of contrastive phones (p.1).” Because coarticulation involves differences in amount of overlap in articulatory gestures, and sounds must be composed of certain articulatory gestures, too much overlap due to coarticulation can actually cause a change in the output of the target vowel, approximating it to a neighboring vowel in a language’s vowel space. Thus, speakers must control coarticulation in order to keep the contrasts in their language inventory.
The author designed an experiment that compared the amount of anticipatory vowel-to-vowel coarticulation in two languages with smaller vowel inventories and a language with larger vowel inventories. She found that the two languages with smaller vowel inventories have greater anticipatory vowel-to-vowel coarticulation than the language that has a bigger vowel inventory. Therefore, there must be some effort from the speakers to constrain the coarticulation in order to maintain the vowel contrasts in their languages.
If indeed there is a constraint in coarticulation due to vowel inventories, then it is reasonable to believe that languages with contrastive nasality in vowels would also constrain nasal coarticulation in certain contexts where confusion may arise because of this. Brazilian Portuguese is known to have contrastive nasal vowels, as well as heavily coarticulated nasalized vowels. Based on this, if there is any evidence of coarticulation control in Brazilian Portuguese, we should expect it to emerge in a comparison between phonemic nasal and coarticulatorily nasalized vowels. Nasal coarticulation must be controlled to maintain the phonemic distinctions between nasal and nasalized vowels.
The method consisted of measuring the difference between A1 (F1 amplitude) and P0 (amplitude of either H1 or H2) in nasal, nasalized and oral vowels in BP in three different time points, at the beginning, midpoint and at the end of the vowel. A1 and P1 (higher harmonic amplitude) difference was not measured because the target vowel is low, so the A1-P0 difference is more appropriate (please refer to section 2 for more details).
The data consisted of disyllable words with nasal, nasalized and oral low front vowel /a/ (nasal [ɐ̃]), with the stress and target vowel on the penultimate syllable. While Brazilian Portuguese has five distinctive nasal vowels, only the vowel /a/ was considered because its formant pattern allows for more consistent analyses. F1 in this vowel does not coincide with any of the first two harmonics, which happens to higher vowels. Any change in these harmonics would be due only to the influence of nasality. Thus, A1 and P0 were more reliably measured, with less change for corrections. The consonant following the target vowel was controlled for manner and place of articulation, covering the three stop places of articulation occurring in BP. The table below shows the words used for the experiment, with the transcription, type of vowel and meaning to show the contrast between these vowels1 :
Table 05 Words used in the experiment.
All words were inserted in the carrier sentence digo _________ claramente. (“(I) say ______ clearly”), and were uttered four times by two adult female speakers of Brazilian Portuguese. The recordings were made using Praat (Boersma and Weenik, 2013).
Once the sentences were recorded, each word and target vowel was labeled with text grids and measured automatically by a Praat script routine developed at the Phonetics Laboratory at University of Colorado at Boulder. Each vowel’s A1 and P0 were taken at three different time points, one at the beginning of the vowel, one at the midpoint and one at the end of the vowel. The difference was automatically calculated by the script routine. Corrections were made based on the visual inspection of each token’s spectral slice.
The working hypothesis in this study is that there is a difference in acoustic nasality between oral, nasal and nasalized vowels in Brazilian Portuguese, with nasal vowels being more nasal than nasalized vowels, and oral vowels being less nasalized overall. This would imply that there is some form control from the speaker to mark the difference between these two vowels, even though from a phonological point of view these vowels have the same [+nasal] feature. The alternative would be that these vowels do not differ significantly from each other, which means that there is no constraint on coarticulation. The table below shows the average acoustic nasality divided by vowel type in the three different time points, as well as the total average difference:
Table 06 Acoustic nasality by vowel type in d.
As stated in the methods section, the smaller the A1-P0 difference, the more nasal a vowel should be, because in turn, the more open the velopharyngeal port should be. By looking at the total column, it is possible already to infer that there is a difference in acoustic nasality between the oral vowel and the other ones, since the difference in the last two is negative, while the difference in the first one is positive. However, it is difficult to affirm that there is a difference between nasal and nasalized vowels. A one-factor ANOVA test was used to determine whether or not there is a difference between each vowel type with regard to nasality. The results show that vowel type influences the amount of nasality, F (2,2) = 41.9, p < 0.001. Now, we need to know whether there is a difference between each vowel type. This was done by running t-tests comparing oral vs. nasal, oral vs. nasalized and nasal vs. nasalized vowels, using the total amount of acoustic nasality. A comparison between oral and nasal vowels yielded a significant difference, t (137) = 8.29, p < 0.001, as well as a comparison between oral and nasalized vowels, t (138) = 7.99, p < 0.001. This result confirms the expectation that oral vowels would be different from nasal and nasalized vowels in Brazilian Portuguese. A comparison between nasal and nasalized vowels, however, did not show a significant difference, t (141) = 0.36, p = 0.72. This lack of significance is evidence that there is no difference between nasal and nasalized vowels in Brazilian Portuguese, and that there does not seem to be a speaker control in the amount of nasality, which would enhance contrast between these two types of vowels. However, based on the aerodynamic difference in terms of nasal airflow between nasal and nasalized vowels found in other studies (Medeiros, 2011), it is worth analyzing the data divided by time point, so as to observe the evolution in acoustic nasality along the vowels’ duration.
The graph below shows the amount of nasality in each vowel type by time point (1 = start; 2 = mid; 3 = end):
Fig 02 Amount of nasality by time point.
The graph above shows that oral vowels are less nasalized overall, as predicted. It also shows, however, that nasal and nasalized vowels follow different trends. Both of them start very nasal, at point 1. The nasal vowel becomes less nasal than the nasalized vowel at point 2, only to increase nasality again at point 3. The trend does not happen the same way for the nasalized vowel. While in general terms the nasal and the nasalized low vowel in Brazilian Portuguese do not seem to differ, it seems like the differ based on the time point. A t-test using only time point 3 confirms this difference, t (44) = 2.92, p < 0.01. Thus, the prediction that nasal and nasalized vowels in Brazilian Portuguese are different with respect to nasality is supported, even if only at the end of them. The fact that the nasalized vowel does not increase in nasality at the endpoint is odd, because there is a nasal consonant right after it. Thus, it seems that speakers still maintain different levels of nasality between these two types of vowels likely to enhance contrast. This agrees with the idea that coarticulation is constrained based on the phoneme contrasts that occur in a given language (Manuel, 1990). Since Brazilian Portuguese has contrastive nasality, it makes sense that the nasalized vowel would be less nasal because phonologically it is not a nasal vowel, but an oral vowel heavily coarticulated. Nevertheless, the nasalized vowel has less nasal coarticulation, which in fact aids in the perception of the nasal versus nasalized low vowel /ã/ in the language, since it contributes to differentiating nasal from nasalized vowels.
Discussion and Conclusion
This paper aimed at analyzing the difference in acoustic nasality between stressed oral, nasal and coarticulated nasal vowel /a/ in Brazilian Portuguese. It was expected that there would be a significant difference in nasality between all three types of vowels, due to the role of coarticulation control as aid in vowel perception. The acoustic parameters to indirectly measure acoustic nasality were the amplitude A1 of the first formant minus the amplitude P0 from the nasal peak in the low harmonics area. This is an indirect measure of the degree of velopharyngeal port opening. The measures were taken at three different time points for each token (start point, midpoint, endpoint). The differences in the amount of nasality between these vowels provided evidence for the constraints in articulation of these sounds and to the enhancement of these sounds perception. 1
The results showed that there is an overall influence of the type of vowel in the amount of acoustic nasality it has. An analysis of overall nasality yielded a significant difference in the degree of nasality between oral and the other two types of vowels. This demonstrates that the low oral vowel /a/ is less nasal than the other two vowels, as expected. An analysis by time point demonstrated that nasal and nasalized vowels differ at the endpoint, the nasal vowel being acoustically more nasal than the nasalized vowel. The results seem to be consistent with a possible control of nasal coarticulation to aid speech perception.
The fact that there is a nasal consonant following the nasalized vowel should make this vowel, if not more, at least equally nasal to the phonemic nasal vowel. Phonemic nasal vowels originally had a nasal consonant at their right edge, which conferred them the nasal quality they have today. This is essentially the same context where nasalized vowels occur. Thus, both vowels are acknowledged to be strongly coarticulatorily nasalized: diachronically in the case of phonemic nasal vowels, synchronically in the case of nasalized vowels. Phonologically this would be consistent with the spread of the feature [+nasal], in that this feature applies to the whole vowel, regardless of the vowel’s phonemic status.
However, if we consider that coarticulation is controlled to enhance the contrast between these two types of vowels, it follows that the nasalized vowel would be less nasal because it has less nasal coarticulation, and languages with a certain phonemic contrasts tend to have less coarticulation with that contrastive feature. Being less nasal at the end point would cue the speaker that there is a nasal consonant coming next in the speech stream, not an oral consonant. This nasal coarticulatory pattern could be a Brazilian Portuguese specific pattern to resolve the issue of nasality perception. Thus, this difference in nasality implies that coarticulation can be controlled for the benefit of the perceiver, and it is not a mere byproduct of the speech stream that the perceiver factors out when processing the speech signal.
Further research should be able to expand these results to other vowel qualities. Phonemic nasal vowels should be more nasal than coarticulatory nasalized vowels regardless of the quality, due to their status as phonemes in Brazilian Portuguese. Other research should consider the nasal consonant’s syllable position. Nasalized vowels have a following consonant in a different syllable than phonemic nasal vowels had historically. Perhaps the nasal consonant belonging to the same syllable would cause the vowel to be more nasalized than it belonging to a different syllable. In this case, less nasality in the nasalized vowel could cue not to a nasal consonant, but actually to a different syllable, so that nasality could actually function as a cue to a boundary, not a segment. Further studies on the perception of these vowels can shed light on these issues.
1There is no velar nasal consonant in BP; thus, for the nasalized vowel, a palatal nasal consonant was used, so it best approximates an articulation in the back of the mouth.
Bisol, L. Nasalidade, um velho tema. DELTA, Vol. 14, special issue, 1998.
Boersma, P. and Weenink, D. Praat: doing phonetics by computer [Computer program]. Version 5.3.04, retrieved from http://www.praat.org/, 2012.
Cagliari, L. An experimental study of nasality with particular reference to Brazilian Portuguese. Unpublished doctoral dissertation, University of Edinburgh, 1977.
Camara Jr, J. Para o estudo da fonêmica portuguesa. Rio de Janeiro: Padrão, 2 ed, 1977.
Chen, M. Acoustic correlates of nasalized vowels. Journal of the Acoustical Society of America, Vol. 102, No. 4, 1997.
Manuel, S. The role of contrast in limiting vowel-to-vowel coarticulation in different languages. Journal of the Acoustical Society of America, Vol. 88, No. 3, 1990.
Medeiros, B. Vogais nasais do português brasileiro: reflexões preliminaries de uma revisita. Revista de Letras, Vol. 72, 2007.
Medeiros, B. Nasal coda and vowel nasality in Brazilian Portuguese. Selected Proceedings of the 5th Conference on Laboratory Approaches to Romance Phonology. Sommerville: Cascadilla Proceedings Project, 2011.
Rauber, A. An acoustic description of Brazilian Portuguese oral vowels. Diacrítica, Ciências da Linguagem, Vol. 22, No. 1, 2008.
Silva, T. Fonética e Fonologia do Português: roteiro de estudos e guia de exercícios. São Paulo: Contexto. 6 ed, 1998.
Sousa, E. Para caracterização fonético-acústica da nasalidade no português do Brasil. Unpublished master thesis. Universidade Estadual de Campinas, 1994.
Este obra está bajo una licencia Creative Commons Reconocimiento-NoComercial-SinObraDerivada 3.0 Unported.