The study deals with variable phonetic traits of Brazilian Portuguese (BP) in contact with Italian immigration languages. It focuses on the realization of an alveolar tap (weak-r) in onset position where one would expect an alveolar trill (strong-r) or its fricative variants (glottal/velar fricative):

The emigration of Italians to Brazil between the second half of the 19th century and the beginning of the 20th century promoted the contact of Portuguese with Italian dialects.

Other phonetic traits of BP in contact with Italian immigration languages

The variable realization under investigation in this paper is the production of a weak-r where one would expect a strong-r. Its pattern of variation and change is compared with that of the two other processes, the vocalization of the alveolar lateral approximant in coda position (henceforth

A previous study (

Palatalization and vocalization generally occur below the level of social awareness in Flores da Cunha (

The article follows the theoretical-methodological assumption of

The main contribution of the article is the heuristic exploration of the assumptions of

Flores da Cunha is one of the 55 municipalities that make up the ICR-RS, a region located in the northeast of the state (

Flores da Cunha in Rio Grande do Sul and Brazil (

In Flores da Cunha, social practices identified with the Italian ethnic origins of the town are expressed mainly in the cuisine and in the culture of grapes and wine, traditions explored by local tourism in celebrations such as the Harvest Festival (

There are two contrastive rhotics in the Portuguese phonological system, /r/ and /ɾ/, called respectively strong-r and weak-r in this paper. The contrast is evidenced only in syllabic onset in intervocalic position (

(1)

/muro/

/kareta/

/karo/

/ɛra/

/muɾo/

/kaɾeta/

/kaɾo/

/ɛɾa/

In word-initial syllabic onset, only /r/ occurs: /roupa/

The contrast between /r/ and /ɾ/ is neutralized in syllabic coda, a position in which, according to Silva (

In syllabic onset, there is variation in the production of /r/ but not in the production of /ɾ/. The most frequent variants of /r/ in onset in BP are the voiceless velar fricative [x] and the voiceless glottal fricative [h], while the least frequent variant is the multiple alveolar trill (

Most studies on the production of a weak-r where one would expect a strong-r in BP in contact with Italian immigration languages deal with the realization in the framework of Labovian variationist theory (

Concerning social predictor variables, weak-r correlates with variables Sex, Age Group, Bilingualism, Place of Residence. Female, younger people favor the alveolar trill or its fricative variants (velar or glottal), while the older population favors the tap in Flores da Cunha (

Regarding linguistic predictor variables, weak-r correlates with variables Position of the Syllable in the Word and Number of Syllables in the Word. Weak-r is favored in medial word position in Caxias do Sul (2004), Flores da Cunha (

Some of the studies suggest that the decline in the production of weak-r by younger people relates to their daily social practices, as moving to and from different speech communities to study and work (

Therefore, in general terms, studies on the production of a weak-r where one would expect a strong-r in BP in contact with Italian immigration languages show that weak-r is favored by males, older, less educated bilingual speakers who live in rural areas. Conversely, the production of strong-r is favored by young, female subjects, maybe due to the stigmatization of weak-r. Medial word position favors weak-r, initial position disfavors it.

As stated in the introduction, generalized phonological processes in supralocal BP can be observed in BP in contact with Italian immigration languages but in moderate or low proportions. The vocalization of the alveolar lateral approximant in coda position and the variable palatalization of alveolar plosives before high front vocoids are two of them. Unlike the production of a weak-r where one would expect a strong-r, vocalization and palatalization are not stigmatized. Both processes are in progress.

Battisti & Dornelles Filho (

Battisti & Dornelles Filho (

Logistic regression of the application of the vocalization of /l/ according to the informant’s linguistic stabilization date (total data) (

Logistic regression of the application of the vocalization of /l/ according to the informant’s linguistic stabilization date by sex (

The analysis of the realization of weak-r according to the age of stabilization of the informants’ phonological system suggests that there would already be some application of the palatalization in the informants’ BP around 1920 (

As with vocalization and palatalization, what would be the pattern of linguistic variation and change in the production of weak-r in the BP of Flores da Cunha? This is what we seek to clarify in this article, testing the hypothesis that, since this variable is a stereotype, it would be in sharp decline in the community in a path of communal change, as mentioned in the introduction.

Logistic regression of the application of the palatalization of /t, d/ according to the informant’s linguistic stabilization date (total data) (

Logistic regression of the application of the palatalization of /t, d/ according to the informant’s linguistic stabilization date by sex (

Two analyses are presented in this paper: a real-time analysis (trend study)

The data used in the analyses were extracted from 24 sociolinguistic interviews, 12 belonging to the VARSUL (1990) collection, 12 belonging to the BDSer (2008-2009) collection.

The data collected from the interviews were coded in two spreadsheets, one for VARSUL data and one for BDSer data, considering the following variables: (a) response variable:

Predictor variables

Source: the authors.

Data were also coded for the Informant and Word variables, included as random variables in mixed-effects statistical models.

The VARSUL and the BDSer samples are balanced for sex and age, but not for level of education and degree of bilingualism (considering their use of Portuguese, the majority language, and Italian immigration languages), that is why level of education and degree of bilingualism do not correspond to predictor variables in the regression models. Information on education level and degree of bilingualism is used only when discussing the proportions of the production of weak-r, considering the following categories: (i) level of education: Elementary School-4 years, Elementary School-8 years, High School; (ii) degree of bilingualism: speaks (Italian immigration languages), understands (Italian immigration languages) and null (does not speak or understand Italian immigration languages).

The statistical analysis is performed with the R program (

The objectives of the real-time analysis are to identify the linguistic and social factors correlated to weak-r in each period under analysis and, in the comparison of the models, to clarify the conditioning factors of the process in the path of the variation and change. We test the hypotheses that the production of weak-r (a) has declined in about 20 years and (b) its application correlates with sex (male), age (elderly) and position of the syllable in the word in the BP of Flores da Cunha, both in the VARSUL and in the BDSer samples, as attested in previous studies (

The analysis of the realization of weak-r according to the age of the stabilization of the phonological system replicates the procedures of Battisti & Dornelles Filho (

The results of the analyses do not confirm the hypotheses tested. The real-time analysis (trend study) shows that the probability of occurrence of weak-r has not declined in about 20 years. It remains stable, which can be seen from the total proportions of rule application observed: 66% (707/1,072) in the VARSUL data, 65% (503/769) in the BDSer data.

In the analysis of the VARSUL (1990) data, the predictor variables Syllable Position with Respect to the Stressed Syllable and Number of Syllables in the Word did not present a significant p-value in the chi-square test; therefore, they were not included in the regression model. The predictor variable Syllable Position in the Phonological Word (henceforth Syllable Position), Sex, Age Group (fixed effects), in addition to Word and Informant as random variables, are included in the regression model. The results of the model

Estimates of the model parameters (logistic regression, generalized linear model with mixed effects) of the realization of a weak-r where one would expect a strong-r in the BP of Flores da Cunha – VARSUL.

Initial (ref. value) | 396/626 (63%) | ||||

Medial | 311/446 (70%) | 0.594 | 0.344 | 1.727 | 0.084 . |

Female (ref. value) | 268/611 (44%) | ||||

Male | 439/461 (95%) | 8.583 | 4.869 | 7.763 | 0.078 . |

40–59 years old (ref. value) | 291/319 (91%) | ||||

25–39 years old | 171/407 (42%) | –1.857 | 5.504 | -0.337 | 0.735 |

60 years old or older | 245/346 (71%) | 2.4103 | 4.621 | 0.522 | 0.602 |

Source: the authors.

MODEL 1. WEAK R ~ SYLLABLE POSITION + SEX + AGE + (1 | INFORMANT) + (1 | WORD).

N = 1,072.

Intercept = –3.0080.

The results in

In the analysis of the BDSer (2008-2009) data, as in the analysis of the VARSUL data, the predictor variables Syllable Position with Respect to the Stressed Syllable and Number of Syllables in the Word did not present a significant p-value in the chi-square test. Therefore, they were not included in the regression model. Again, predictor variables Syllable Position, Sex and Age Group are included in the model as fixed-effect variables, and Word and Informant as random variables. The results of the model are presented in

Estimates of model parameters (logistic regression, generalized linear model with mixed effects) of the realization of a weak-r where one would expect a strong-r in the BP of Flores da Cunha – BDSer.

Initial (ref. value) | 303/486 (62%) | ||||

Medial | 200/283 (71%) | 0.7628 | 0.4281 | 1.782 | 0.074 . |

Female (ref. value) | 255/460 (55%) | ||||

Male | 248/309 (80%) | 7.1773 | 3.7708 | 1.903 | 0.057 . |

40–59 years old (ref. value) | 85/226 (38%) | ||||

25–39 years old | 54/160 (34%) | 2.7074 | 4.6347 | 0.584 | 0.559 |

60 years old or older | 364/383 (95%) | 7.2247 | 3.6358 | 1.987 | 0.046 * |

Source: the authors.

MODEL 1. WEAK R ~ SYLLABLE POSITION + SEX + AGE + (1 | INFORMANT) + (1 | WORD).

N = 769.

Intercept = –5.3762.

The real-time analysis, therefore, does not confirm the hypothesis of a decline in the production of weak-r in about 20 years, nor the correlation with Sex and Age Group. The analysis shows that these variables interact in the two samples and that the interaction affects the estimates of the fixed variables in the models. The only hypothesis that seems to be confirmed is that of Syllable Position, with the Medial factor favoring the rule, but we take this result cautiously because the correlations attested are only slight.

The results here obtained seem to contradict Battisti & Martins (

We crossed the proportions of realization of weak-r with variables Level of Education and Degree of Bilingualism. The chi-square test showed a significant difference in the application of the rule by factor of both variables. The results for the VARSUL sample are in

Proportions of weak-r by Education – VARSUL data.

Source: the authors

Proportions of weak-r by Bilingualism – VARSUL data.

Source: the authors.

Proportions of weak-r by Education – BDSer data.

Source: the authors.

Proportions of weak-r by Bilingualism – BDSer data.

Source: the authors.

VARSUL informants with a high degree of education (

Altogether, these results (

Under the premise that 15 years old is the age at which the phonological system stabilizes and having calculated the year in which each of the 24 informants (12 VARSUL informants, 12 BDSer informants) was 15 years old, it was possible to go back to 1930. With this piece of information, a logistic regression analysis was performed to obtain the estimated proportion of the production of weak-r at the time when the phonological system of the informants stabilized (

Logistic regression of the production of weak-r according to the date of the informant’s linguistic stabilization (total data).

Source: the authors.

Total proportions are an average. Besides that, the VARSUL and BDSer samples were collected more recently, in 1990 and 2008-2009 respectively. Sex stratification of the regression according to the informant’s linguistic stabilization date (

Logistic regression of the production of weak-r according to the date of the informant’s linguistic stabilization (stratification by sex).

Source: the authors.

The results of the analysis by age of stabilization of the phonological system, therefore, do not confirm the hypothesis that the decline in the realization of weak-r would be sharp, following a pattern of communal change. The pattern is that of generational change i.e., a gradual decline in the proportions of rule application from generation to generation.

The analyses presented in this paper did not confirm both the hypothesis of a sharp decrease in the realization of weak-r in the BP of Flores da Cunha and the hypothesis of correlation of the production of weak-r with sex (male) and age group (elderly). The only hypothesis confirmed was that of the correlation of weak-r with syllable position (medial) in the phonological word.

The results here obtained contribute to clarifying the patterns of progression of sound changes in varieties of BP in contact with Italian immigration languages. They show that, as with vocalization and palatalization, which have been progressing in the BP of Flores da Cunha towards leveling with the supralocal BP, the realization of weak-r declined over a period of almost 70 years but stabilized between the years 1990 and 2008–2009, a pattern of progress that conforms to the gradual character of the generational change. That is, despite the tendency towards leveling of the BP of Flores da Cunha with supralocal patterns of Portuguese, the slow decline in the production of weak-r and its stabilization in the most recent period indicate the possibility that weak-r is being incorporated as a trait of the local accent.

Concerning the phonological system of Portuguese, the BP of Flores da Cunha and that of other Brazilian communities founded by Italian immigrants exhibit a different distribution of the alveolar tap than the one predicted (

The results of the real-time analysis are not conclusive. They point to a possible favoring role of the medial syllable in the application of the rule.

The regression analysis considering the age of stabilization of the phonological system made it possible to trace the pattern of variation and change of the realization of weak-r in the BP of Flores da Cunha since at least as far back as 1930. The analysis also captures the different behavior of male and female speakers in the process. As a heuristic resource, this type of analysis should not be taken as a method to faithfully portray the investigated reality, but as a device to approach inaccessible empirical realities, which can only be treated by hypothesis, as is the case of the long process of language variation and change.

A future development of the study reported here may be a work in line with the third wave sociolinguistics (

According to De Boni and Costa (

Battisti (

Altenhofen and Margotti (

VARSUL (Variação Linguística na Região Sul do Brasil) [Linguistic Variation in the Southern Region of Brazil] is a speech database shared by four higher education institutions in Southern Brazil: Universidade Federal do Rio Grande do Sul (UFRGS), Universidade Federal de Santa Catarina (UFSC), Universidade Tecnológica Federal do Paraná (UTFPR), Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS). The sociolinguistic interviews of VARSUL are conducted with participants of both sexes (male, female) with different levels of education (Elementary I, Elementary II, High School) and ages (25–50 years old, 50 years old or more), residing in the capitals and some cities from the countryside of Rio Grande do Sul, Santa Catarina and Paraná, states in the Southern Region of Brazil. The interviews of the base sample of VARSUL were carried out in the early 1990s. Information available at:

BDSer (Banco de Dados de Fala da Serra Gaúcha) [Speech Database of the Region of Serra Gaúcha], from Universidade de Caxias do Sul (UCS), in Rio Grande do Sul, includes sociolinguistic interviews from 4 municipalities in the ICR-RS: Caxias do Sul, Antônio Prado, São Marcos, Flores da Cunha. The interviews of BDSer were stratified according to the informants’ area of residence (rural, urban), sex (male, female), age group (15 to 25 years old, 30 to 45 years old, 50 to 65 years old, 70 years old or more), education (1st–4th grade of Elementary School, 5th–8th grade of Elementary School, 1st–3rd grade of High School, Higher Education – one or more years). The interviews in Flores da Cunha, used in this study, were carried out between 2008 and 2009.

Data provided by Município de Flores da Cunha (

The fricativization of /r/ has been occurring in BP since the 1970s, according to Langaro (

In quantitative Labovian analyses, the term

Phonetic-acoustic analyses such as that by Harrington (

The asymmetry of language transmission was first proposed by Labov (

The age of 15 years was arbitrarily taken to delimit youth because (i) individuals from 15 to 19 years old form one of the youngest age groups (adolescents) in studies of linguistic variation in apparent time (

Real-time analyses of the trend study type are designed “to make use of the present to interpret the past: to return to the scene of an earlier study and repeat it as closely as possible” (

Part of the data come from the sample of Azeredo (

In the study by Azeredo (

In mixed-effects logistic regression statistical analyses, both fixed-effects variables (such as those in Chart 2 in this article) and random variables (whose effects correspond to a range of values not controlled by the analyst) are included in the models. Incorporating predictor variables such as Informant and Word as random variables to the statistical models minimizes the effect of commonly repeated words or individual characteristics of the informants on the results achieved, as explained by Oushiro (

Even though level of education and degree of bilingualism were not predictor variables in the regression models, data were coded for the two variables. The variable degree of bilingualism has two levels in the coding of VARSUL data (speaks, understands) because all 12 VARSUL informants are bilinguals, but it has three levels in the coding of BDSer data (speaks, understands, null) because 2 of the 12 informants are not bilingual.

If a p-value is less than 0.1, it is flagged with a dot (.). If a p-value is less than 0.05, it is flagged with one star (*). If a p-value is less than 0.01, it is flagged with 2 stars (**). If a p-value is less than 0.001, it is flagged with 3 stars (***).

The authors have no competing interests to declare.