University of Texas, Austin

The object of this essay is to describe some of the current research in Luso-Brazilian (LB) phonology. Rather than build a bibliographic background, I shall introduce this topic by showing various techniques current analysts utilize to investigate Brazilian Portuguese nasality⁵⁴. The reason for this unconventional approach is that more than at any other time, «phonology», the study of speech sounds, is comprised of an extensive array of distinct academic specializations. It is quite revealing to see the many levels and angles of analysis which can be applied to such a topic. For example, linguists who adhere to generative theories approach the study of LB phonology as an opportunity to verify or develop claims regarding universal grammar. Sociolinguists study the same sound sequences; their focus, however, might be on which portion of a given population has a distinct pronunciation and what the social implications might be. Concurrently, phoneticians are interested in the aerodynamics of speech patterns independent of universal grammar or social implications.

While in some circles these various branches of linguistics enjoy a certain level of autonomy, this luxury is not available to LB phonologists. In order to compile a complete picture, researchers in LB need to draw from all sources. This will become evident as we investigate the nasalization of /a/ when followed by a tautosyllabic nasal consonant. I shall draw from experimental phonetics as the core element of this analysis and then add footnotes and commentary to show how related branches of phonology might approach this issue.

One of salient phonological characteristics of Brazilian Portuguese (BP) is its strong vowel nasalization. In addition to describing vocalic segments as ± oral or ± nasal⁵⁵, nearly all researchers in BP take advantage of a useful dichotomy between nasal vowels and nasalized vowels (Lacerda & Head 1963). Generally speaking, there are five phonemic nasal vowels /ĩ õ ũ/. At the same time, any vowel may be nasalized when followed by a tautosyllabic nasal element⁵⁶. The object of this study is to gain a better understanding of nasalization in BP by comparing the acoustic changes in vowel quality and segment duration of minimal pairs containing /a/ and /a + N/.

It has been a challenge for LB phonologists to provide an accurate account of Portuguese nasalization and their theoretical approach also modifies which patterns are highlighted. For example, early structuralists attempt to describe nasalization by categorizing its phonemic value. Consequently, one theory accepts the archiphonemic analysis which suggests a sequence of an oral vowel followed by a nasal consonant, such as, campo /káNpo/ «field» (Mattoso Câmara 1953, 1972; Morais Barbosa 1965; Reed & Leite 1947). Other structuralists suggest a nasal allophone for each consonantal phoneme (Hall 1943, Feldman 1967). Under this analysis [^mp is an allophone of /p/ showing a prenasalization feature, e. g., campo [k^mpu] «field». Finally, others posit a single nasal phoneme for both oral and nasal vowels (and in fact «nasalized» vowels), such as, cata

––––––––   854   ––––––––

/káta/ «search» vs. canta /kta/ «she sings» (Back 1973; Head 1964; Lopes 1976; Ludtke 1953)⁵⁷. It was during this time when Lacerda & Hammarstrom (1957: 123) proposed a phonetic transcription for Portuguese, based on the International Phonetic Alphabet. For nasalization they suggest:

Current generative approaches focus on morphological relationships between pairs such as Pão-panificadora «bread-bakery» and bem-benefício «good-benefit». As such, these descriptions generally include a Vowel + Nasal sequence, which allows one to produce underlying forms such as /pan + e/ -> páo, /ben + e/ -> bem (Azevedo 1981; Brakel 1985; Mira Mateus 1982).

At times acoustic researchers have offered phonetic cues about nasality which affect phonological descriptions⁵⁸. In one instance, Delgado Martins (1975: 6) reports that the duration of nasal vowels is about equal to that of the oral vowels; however nasalized vowels are another matter and are not mentioned. López (1979: 47) states that the most consistent consequence of the nasalization of /a/ in BP is a weakening in amplitude and a rise in the first formant frequency. These acoustic parameters are elaborated in Head (1964: 204):

In the interest of caution, however, Almeida (1976: 357) finds this description somewhat simplistic. In his experience it is more difficult to find unequivocal signs of nasality.

Lipski (1975) reports strong nasalization of oral vowels in open syllables. This may be seen in the disparate pronunciations of cama «bed» between Spanish and Portuguese. In Spanish one hears a clear oral vowel [ká-ma], but in Portuguese speakers exhibit heavy nasalization, saying [k·ma]⁵⁹. As Lipski (1975: 67) explains:

Besides changes in vowel quality, nasalization may affect the duration of the oral vowels, the nasal consonant, or both. Researchers in BP nasalization propound whether nasalized words still retain a nasal consonant or whether this consonant is subsequently deleted after vowel nasalization. This debate is highly subjective; opinions are based on the degree of phonemic vs. phonetic precision, phonetic environment, register, and/or dialect, etc. These parameters are also subsequently interpreted according to whichever linguistic theory is ascribed to⁶⁰.

Independent of the analyst's objective, experimental data are required in order to interpret tendencies regarding the nasalization of Portuguese vowels. Unfortunately, the existing data up to this point are fragmentary and scarce.

One of the challenges in gathering data for this experiment is to control the various intervening variables. The analysis touches on oral vs. nasal vowel quality, consonant vs. vocalic duration, and speaking velocity. The emphasis here is to provide acoustic parameters distinguishing oral and nasalized pronunciations; as such, the examination is limited to minimal pairs containing /a/ and /a + N/. The tokens are:

tato	«touch»	tanto	«so much»
dado	gave (part.)	dando	«giving»
caça	«s/he hunts»	cansa	«s/he tires»

Underlying this experiment is the assumption that phonologically reduced forms contrast with full citation forms as speech becomes more relaxed or faster (cf. Shockey 1987: 217). With this in mind, tokens representing unreduced citation forms were compared with reduced pronunciations of the same word. Data collection was monitored by the following control factors.

Informants: In order to control pitch level for accuracy of vowel quality, all subjects are adult male speakers between the ages of 18 and 27 and all are from the city of São Paulo. Thirty Brazilian informants, all students at the Universidade de São Paulo, were taped in the

––––––––   855   ––––––––

recording studio of the Communications Department at that institution⁶¹.

Procedure: The data were collected from sentence and word lists which contained the tokens chosen for the experiment. The subjects were asked to read the sentences and words, which were provided on cards in four ways: (1) sentence form; (2) reduced form; (3) carrier phrase; and (4) citation form. (See the Appendix for a list of the sentences used for each token). All subjects followed the same procedure and order. They first read the sentences which contained the unmarked embedded tokens. Subjects were asked to speak in a normal speaking style as if they were talking to a friend. They were then asked to read the same set of sentences twice in a row as fast as they could. These tokens represent the reduced forms. Next the subjects were given cards which contained the individual words and were asked to read them at a normal pace within the carrier phrase, «Diga a palavra —— comigo» (Say the word with me). Finally, to obtain the citation forms, the subjects were asked to read carefully the same list of words, numbering each (e. g., um-dodo, dois-tanto... etc.) in order to control the pace and to avoid the effects of overlapping adjacent tokens.

The analysis of the data was performed at the Phonology Laboratory of the University of California, Berkeley. Using the linguistic program «Sadie» running on a Hewlett Packard 836-236 computer, spectrograms were made of the tokens from which vowel formant frequencies and segment durations were determined. The vowel formant frequencies were measured at the so-called «steady-state» of the formant⁶²s. Sadie displays a harmonic spectrum, from which formant peaks are marked and frequency is displayed. In measuring the segment duration of each token, Sadie allows one to mark the token at any point and the duration from that mark is given in milliseconds. The oral tokens contain CVCV alternations between /t/, /d/, /k/, or /s/ and the vowels. These transition points are readily visible. Aspiration was included with the duration of the consonantal element. The nasal tokens were more difficult to divide into segments. The transition from an oral vowel to a nasal element builds gradually and segments cannot be assigned to definite boundaries. Consequently, I separated the nasal consonant from the vocalic element only if there was a marked reduction in amplitude (as evidenced in the wave form), or if there was a distinct change in the vowel formant frequencies during the consonantal nasal pronunciation (such as a loss of the formants)⁶³.

It may be appropriate to justify some of the procedural strategies and to describe some of the challenges involved. Normally, reduced forms occur in relaxed, natural, conversational settings. In comparison, a contrived list of sentences and words seems very artificial. The disadvantages of word lists are prominent: there are spelling pronunciations, fewer false starts (natural occurrences in conversation), nervousness in knowing that there is a tape recorder, an effort to pronounce a word «correctly», etc. However, there are also disadvantages in collecting natural speech. The identification of register is subjective. There are varying phonological effects from adjacent elements which cannot be controlled. It is nearly impossible to have more than one informant say the same token in the same setting, let alone thirty. Inevitably one must compromise. Nevertheless, the expectation here was that the large sample size would offset some of the adverse effects of not recording natural speech.

The major challenge was to collect reduced forms. Shockey (1987) presents data regarding the effects of rate on the reduction processes and the methodology here is essentially an adaptation of hers. Shockey asked the informants to repeat two sentences ten times at a slow, then a normal, and than a very fast rate. Since the informants in my analysis had four levels of analysis, only two repetitions were required.

Analysis: Significant changes in vowel quality, for both the oral and nasal tokens, may be effectively compared using a 2 x 4 two-way Analysis of Variance on the first and second vowel formant frequencies. The ANOVA, a common statistical test, is ideal for the examination of vowel quality⁶⁴. In the comparison of two populations, an ANOVA is designed to compare whether the variance within a given group is significantly greater than the variance between groups. For example, in this experiment we want to determine whether the variance in vowel height between Portuguese oral and nasalized vowels is significant. There are two fundamental reasons why it does not suffice simply to compare the average formant

––––––––   856   ––––––––

frequencies of the oral and nasalized vowels. First, without an F-ratio (the ratio comparing two independent variances) it is merely conjecture to say that average scores from two groups are different enough to be significant or that the difference is not due to chance. Secondly, even if the average scores between the two groups appear to be wide, it may be that the standard deviation in one or both of the groups is so great that it is impossible to draw any conclusion from the average score.

In this experiment a 2 x 4 two-way ANOVA is appropriate because we are looking at two factors: Factor A = speaking velocity (four levels), and Factor B = Oral vs. Nasal (two levels). In partial replication of Disner (1988), I ran two separate ANOVAs; one for the distinction of the first formant frequencies for Portuguese /a-/, a second for the second formant frequencies of /a-/.

The segment duration is analyzed by comparing the proportional spacing in milliseconds of all segments in each token and evaluating their percentage of change among the four speaking velocities.

Vowel Quality: Table 1 presents results regarding the vowel quality of oral and nasalized /a/ as related to the four speaking velocities.

The average formant frequencies of the oral tokens are similar to those from previous BP studies (Nobre & Ingemann 1987). The most visible difference between the oral and nasalized averages is found in the decreased first formant frequencies of the nasalized vowels. The lower F1 averages confirm that vowel raising is brought on by nasalization, similar to that of phonemic nasality. In addition to vowel raising, the second formant frequencies of the nasalized vowels are somewhat higher than the oral vowels, suggesting a more fronted pronunciation⁶⁵.

With reference to speaking velocity, the data suggest a slight trend toward centralization as velocity increases; both the oral and nasalized vowels exhibit a lower F2 frequency from Citation to Reduced forms.

Results from the ANOVA confirm these general impressions inferred from the data⁶⁶.Table 2 presents the results of ANOVA 1 which compares the variance between the first formant frequencies of the oral and nasalized vowels.

The F-ratios corroborate a significant difference in vowel height between the oral and nasalized pronunciations of /a/ (Factor B). The different formant frequencies brought on by speaking velocity are also significant (Factor A) and quite possibly are related to the centralization of the oral tokens. Table 3 displays the results of ANOVA 2 comparing the variation found in the second formant frequencies.

Due to the fact that both the oral and the nasalized pronunciations exhibit a shift toward lower F2 frequencies as speaking velocity increases, it is not surprising to see that the F-ratio for Velocity (Factor A) reaches significance. The overall higher F2 frequency of the nasalized vowels proves to be significant as well, suggesting that the nasalized vowels are not only raised but are more fronted than the oral pronunciations.

Duration: The average segment durations of the oral tokens, tato, dado, caça and the nasalized tokens, tanto, dando, and cansa are presented in Table 4.

A preliminary comparison of the vowel durations shows that the nasalized vowels are all shorter than the oral ones. It would be inaccurate, however, to claim that nasalization is the only determining factor. Vowels are generally shorter in closed syllables (Lehiste 1970) and the duration of vowels also systematically decreases as the number of segments increases (Klatt 1976; Lindblom, Lyberg & Holgren 1981). Consequently, one needs to exercise caution when comparing the raw scores between the oral and nasalized tokens. However, despite these possible intervening variables, not only are the nasalized vowels shorter than the oral ones, but the nasalized tokens were pronounced more rapidly than the oral ones. As such it seems viable that Portuguese speakers identify nasalization with voice quality more than duration.

When looking at the raw data it is difficult to see the effect that speaking velocity has on the segmental relationship of the tokens, due to the fact that the reduced words were pronounced faster. It is more revealing to compare the effect that velocity has on the proportional division of the oral and nasalized tokens. Figure 1 displays the percentage of change of each segment from tato, dado, and caça in the transition from citation form to reduced form.

The average segment division in citation form is moderately symmetrical. The tonic syllable and the atonic syllable are nearly identical; the tonic, naturally, is slightly longer.

However, BP unstressed syllables are characteristically weak in continuous speech. Consequently, the atonic syllable is proportionately small in reduced form (e. g., the final vowel only covers about 10%). This pattern is evident in all of the tokens containing oral vowels. This is especially evident in tato which most informants pronounced by eliminating or devoicing the atonic /o/, [tát]-[tát^h].

Voiced consonants are customarily shorter in duration than their voiceless counterparts

––––––––   858   ––––––––

and occlusives are generally shorter than fricatives (Delgado Martins 1975). Accordingly, the initial and intervocalic duration of /d/ in dado is shorter than /t, s/ in the other tokens.

The nasalized tokens are more difficult to compare because as speaking velocity increases, each token reduces differently. The raw scores showing the average segment durations suggest that cansa and dando are distinct in their reduction processes. BP cansa exhibits nasality in its vowel quality at the expense of the nasal consonant. Without an occlusive consonant to promote the articulation of a nasal consonant, /kánsa/ is pronounced [ksa] in continuous speech. In fact, even citation form of cansa reveals little evidence of a nasal consonant. In the case of dando and the /nd/ heterosyllabic cluster, BP frequently drops the occlusive element. Although this is morphologically limited to present participles, there may also be some phonetic basis. Nasal consonants are longer when followed by a voiced obstruent, thus becoming more perceptible and less likely to be deleted (Ohala & Greenlee 1980). Whether phonetic or morphophonemic, BP /dándo/ is regularly articulated [dnu]. The obstruent deletion also creates new syllable boundaries whereby the nasal element is now in syllable initial position. These divisions are more evident in Figure 2 which displays the proportional relationship of the nasalized tokens, tanto, dando, and cansa as related to increased speaking velocity. Figure 2 also illustrates a very brief atomic vowel in tanto which, similar to the oral token tato, eliminates or devoices atonic /o/, [tⁿt]-[tⁿt^h]

Naturally the conclusions to a study such as this one are going to be classified phonetically. However, it bears repeating that our conclusions concerning LB phonology are enhanced when phonetics, sociolinguistics, universal grammar, applied linguistics, first and second language acquisition, contrastive analysis, and dialectology can all be integrated. This is not to say that LB studies should be less poignant or replete with pseudo-correlations, but it would be a shame if our research were only valid within the limited sphere in which it was conceived⁶⁷.

Although the data in this experiment are limited to a small set of oral and nasalized minimal pairs, we can draw experimental conclusions regarding BP vowel quality and segment duration as influenced by speaking velocity:

1. BP /a/ exhibits significant vowel raising when nasalized by a subsequent tautosyllabic nasal element;

2. Increased speaking velocity promotes centralization of tonic /a/ in both oral and nasalized environments;

3. The duration of nasalized /a/ is shorter than its oral counterpart. In fact the nasalized words are shorter than the oral ones, suggesting that nasalization may be signaled more by voice quality than duration;

4. Tonic and atomic syllables are symmetrically balanced in careful pronunciations, but as speaking velocity increases the BP atomic syllables become proportionately shorter;

5. BP cansa exhibits a nasalized vowel with no subsequent nasal consonant, [ks];

6. BP dando deletes the obstruent /d/, resulting in a syllable initial nasal consonant [dnu];

7. BP tanto features a short nasal consonant. However, due in part to the voiceless obstruent, the atomic vowel is extremely reduced, [tⁿt]-[tⁿt^h].

What makes these conclusions interesting with regard to BP phonology is to see how the various disparate specializations might implement them for their own purposes. Obviously the easiest application will be for phoneticians who can utilize these data in the analysis of nasality and nasalization of other BP vowels. For example, although most descriptions of BP do not report significant

––––––––   859   ––––––––

changes in vowel timbre between the other oral and nasal vowels, this still needs to be verified experimentally. I especially suspect changes in vowel quality and segment duration brought on by the heavy diphthongization which accompanies /e + N/, as in [spri] sempre «always» which is typical in São Paulo. There may also be significant variations in the pronunciation of nasalized vowels which are also raised in unstressed position, as in [ĩsínu] ensino «I teach».

Theoretical phonologists can implement these conclusions to reevaluate the terms used to describe BP sounds. For example, the changes in vowel quality and duration suggest that the label «± nasal» is insufficient in providing an accurate assessment of BP nasality. This is not just a case of inventing new terms and binary combinations, but will require a greater understanding of what features are affected by nasalization.

The conclusions above also raise questions as to what degree the changes in vowel quality and duration are language specific or part of a general aerodynamic universal. Consequently, contrastive analysts in BP phonology can utilize these data for comparison with similar Romance languages or among As I mentioned in the introduction, the study of phonology has now advanced to include many specializations. Researchers in BP phonology are involved in all branches of this discipline and our challenge now is to be able to advance the various disciplines while being able to synthesize information from related academic specializations.

non-native speakers of Portuguese.

The following list contains the sentences in which the oral and nasalized tokens were embedded to evaluate the effects of nasality on the duration and vowel quality /a/ in Brazilian Portuguese. Here the tokens are italicized, but they were unmarked in the actual experiment.

1. O Carolinhos sabe tanto como José.

2. Entre os sentidos o tato pode ser o mais esquecido.

3. A gente se cansa disso.

4. Esse cão de caça corre muito.

5. Esse rapaz está dando problemas.

6. Esse presente foi dado de boa vontade.

Almeida, Antonio. 1976. «The Portuguese Nasal Vowels: Phonetics and Phonemics». In Jurgan Schmidt-Radefelt (ed.), Readings in Portuguese Linguistics Amsterdam: North-Holland Publishing. 348-96.

Azevedo, Milton M. 1981. A Contrastive Phonology of Portuguese and English. Washington, D. C.: Georgetown University Press.

Back, Enrico. 1973. «São fonemas as vogais nasais do português?» Constructura 4: 297-317.

Beyl, David W. 1975. «A Distinctive Feature Analysis of the Portuguese Phonological System». Georgetown University Working Papers on Language and Linguistics 11: 84-105.

Bibliografía de Lingüística Portuguesa. 1987. Lisboa: Núcleo de Estudos de Lingüística Contrastiva da Faculdade de Ciências Sociais e Humanas da Universidade Nova de Lisboa e Litoral, Edições.

Bibliografía Dialectal Galego-Portuguesa. 1976. Lisboa: Centro de Lingüística da Universidade de Lisboa.

Brakel, C. Arthur, Jr. 1985. «Reflections on the Analysis of Exceptions to the Rule of Iberian Portuguese Vowel Reduction». Hispanic Linguistics 2: 63-85.

Delgado Martins, María R. 1975. «Vogais e consoantes do português». Boletim de Filologia 24: 1-11.

——. 1982. Aspects de l'accent en portugais. Hamburg: Buske.

Dietrich, Wolf. 1980. Bibliografia da Lingua Portuguesa do Brasil. Tübingen: Tübinger Beitrage zur Linguistik, Verlag Günter Narr.

Disner, Sandra Ferrari. 1988. «On Describing Vowel Quality». In John J. Ohala and Jeri J. Jaegger (eds.), Experimental Phonology. Orlando: Harcourt Brace Jovanovich, Publishers. 69-79.

Duarte, Yara and Raquel Teixeira. 1979. «O processo de nasalização das vogais em português sob o enfoque da fonologia gerativa». Letras de Hoje 37: 20-35.

Feldman, David. 1967. «A Comparison of the Segmental Phonemes of Brazilian Portuguese and American Spanish». Linguistics 29: 44-57.

Godínez, Jr., Manuel. 1978. «A Survey of Spanish and Portuguese Phonetics». Working Papers in Phonetics 44. University of California, Los Angeles.

Guy, Gregory Riordan 1981a. «Linguistic Variation in Brazilian Portuguese: Aspects of Phonology». Dis. University of Pennsylvania.

——. 1981b. «Parallel Variability in American Dialects of Spanish and Portuguese». In David Sankoff and Henrietta Cedergren (eds.), Variation Omnibus. Carbondale: Linguistic Research. 85-95.

Hall, Robert A., Jr. 1943, «The Unit Phonemes of Brazilian Portuguese». Studies in Linguistics 1: 1-6.

Hatch, Evelyn and Hossein Farhady. 1982. Research Design and Statistics: For Applied Linguistics. Rowley, MA: Newbury House Publishers, Inc.

Head, Brian E 1964. «A Comparison of the Segmental Phonology of Lisbon and Rio de Janeiro». Dis. University of Texas, Austin

——. 1987. «Propriedades fonéticas e generalidade de processos fonológicos: O caso do "r caipira"». Cadernos de Estudos Lingüisticos 13: 5-39.

Klatt, Dennis H. 1976. «Linguistic Uses of Segmental Duration in English: Acoustic and Perceptual Evidence». Journal of the Acoustical Society of America 59: 1208-221.

Lacerda, Armando de and Goran Hammarstrom 1952. «Transcrição fonética do português normal». Revista do Laboratório de Fonética Experimental, Coimbra 1: 119-35.

Lacerda, Armando de and Brian Head 1963. «Análise de

––––––––   861   ––––––––

sons nasais e sons nasalizados do português normal». Revista do Laboratório de Fonética Experimental, Coimbra 6: 1-17.

Lehiste, Ilse. 1970. Suprasegmentals. Cambridge, MA: M. I. T. Press.

Lemle, M. and Anthony Naro. 1977. Competências básicas do português. Rio de Janeiro: MOBRAL.

Lindblom, Björn, Bertil Lyberg and Karm Holgren 1981. Durational Patterns of Swedish Phonology: Do They Reflect Short-Term Motor Memory Processes? Bloomington, IN: Reproduced by the Indiana University Linguistics Club.

Lipski, John M. 1975. «Brazilian Portuguese Vowel Nasalization: Secondary Aspects». Canadian Journal of Linguistics 20: 59-77.

Lopes, Edward (Eduardo). 1976. Fundamentos da lingüística contemporanea. São Paulo: Edta. Cultrix.

López, Barbara Strodt. 1979. «The Sound Pattern of Brazilian Portuguese (Carioca Dialect)». Dis. University of California, Los Angeles.

Ludtke, Helmut. 1953. «Fonemática portuguesa». Boletim de Filologia 13: 273-88; 14: 198-217.

Mattoso Câmara, Joaquim Jr. 1953. Para o estudo da fonêmica portuguesa. 2nd ed. (1977). Rio de Janeiro: Organização Simőes.

——. 1972. The Portuguese Language. Translated by Anthony J. Naro. Chicago: University of Chicago Press.

Morais Barbosa, Jorge de. 1965. Etudes de phonologie portugais. Lisboa: Junta de Investigaçőes do Ultramar

Mira Mateus, María Helena. 1982. Aspectos de Fonologia Portuguesa. 2nd ed. Lisboa: Instituto Nacional de Investigação Científica.

Naro, Anthony J. 1976. Tendências atuais da lingüística e da filologia no Brasil. Rio de Janeiro: Livraria Francisco Alves Editora S. A.

Nobre, María Alzira and Frances Ingemann. 1987. «Oral Vowel Reduction in Brazilian Portuguese». In Robert Channon and Linda Shockey (eds.), In Honor of Ilse Lehiste. Dordrecht-Holland/Providence-U. S. A.: Forts Publications, 195-206.

Ohala, John J. and Mel Greenlee. 1980. «Phonetically Motivated Parallels between Child Phonology and Historical Sound Change». Language Sciences 2: 283-308.

Pardal, Ernesto D'Andrade.1977. Aspects de la phonologie (generative) du portugais. Lisbone: Centra de Linguística da Universidade de Lisboa.

Parkinson, Stephen. 1983. «Portuguese Nasal Vowels as Phonological Diphthongs». Lingua 61: 157-77.

Quilis, Antonio. 1979. «Comparación de los sistemas fonológicos del español y del portugués». Revista Española de Linguística 9: 1-22

Redenbarger, Wayne. 1977. «Lusitanian Portuguese [] as an Advanced Tongue Root and Constricted Pharynx Vowel». In Peter Hagiwara (ed.), Studies in Romance Linguistics. Rowley, MA: Newbury House. 26-36.

——. 1978. «Portuguese Vowel Harmony and the Elsewhere Condition». In Margarita Súñer (ed.), Contemporary Studies in Romance Linguistics. Washington: Georgetown University Press. 258-78.

——. 1981. Articulator Features and Portuguese Vowel Height. Cambridge, MA: The Department of Romance Languages and Literatures of Havard University.

Reed David W. and Yolanda Leite. 1947 [19431. «The Segmental Phonemes of Brazilian Portuguese: Standard Paulista Dialect». In Phonemics: A Technique for Reducing Languages to Writing. By Kenneth L. Pike, 194-202. Ann Arbor. University of Michigan Press.

Simões, Antônio Roberto Monteiro. 1987. «Temporal Organization of Brazilian Portuguese Vowels in Continuous Speech: An Acoustical Study». Dis. University of Texas, Austin

Shaw, Inés Senna. 1986. «Vowel Nasality in Brazilian Portuguese: An Experimental Approach with Focus on Derivational and Inflectional Alternations». Dis. University of Kansas.

Shockey Linda. 1987. «Rate and Reduction: Some Preliminary Evidence». In Robert Channon and Linda Shockey (eds.), In Honor of Ilse Lehiste. Dordrecht-Holland/Providence-U. S. A.: Foris Publications. 217-24.

Votre, Sebastião. 1979. «Aspectos da variação fonológica no Rio de Janeiro: Processos de redução de travamento silábico». Letras de Hoje 37: 36-52.

——. 1981. «Phonological and Syntactic Aspects of Denasalization in Spoken Brazilian Portuguese». In David Sankoff and Henrietta Cedergren (eds.), Variation Omnibus. Carbondale: Linguistic Research. 97-103.

Wright, James T. 1986. «The Behavior of Nasalized Vowels in the Perceptual Vowel Space». In John J. Ohala and Jeri J. Jaeger (eds.), Experimental Phonology. Orndo: Harcourt Brace Jovanovich, Publishers. 45-67.

Hispania [Publicaciones periódicas]. Volume 72, Number 4, December 1989