RUSSIAN ACCENT IN BRAZILIAN PORTUGUESE AFFECTS THE PERCEPTION OF VOICE PLEASANTNESS BY BRAZILIANS Sotaque Russo em Português Brasileiro Afeta a Percepção de Agradabilidade de Voz por Brasileiros

The field of foreign accents in Brazilian Portuguese is very poorly explored mainly in relation to perceptual impressions caused on listeners by foreign accents. The objective of this paper is to conduct a perceptual experiment to analyze the perception of voice pleasantness in relation to degrees of foreign accent. Our focus is on Russian accented Brazilian Portuguese. Despite the representativeness of the Russian community in Brazil, there are hardly any studies on Russian accented speech. To evaluate the perception of voice pleasantness in relation to the Russian accented speech, we have selected, from our database, speech samples in Brazilian Portuguese from 12 native Russian speakers and six Brazilians, all of them residing in São Paulo. The speech samples were incorporated into an online questionnaire. The analysis of the 129 answers given by the native Brazilian Portuguese speakers showed a strong negative correlation between the degree of Russian accent in Brazilian Portuguese and the degree of voice pleasantness. Experience of interaction with foreigners or knowledge of foreign languages, including Russian, did not influence the result. We conclude that higher degrees of Russian accented speech in Brazilian Portuguese affect the perception of speakers ́ characteristics by Brazilians in a negative way.

Brazil is investing efforts in the data collection for the Atlas Linguístico do Brasil (ALiB), the Linguistic Atlas of Brazil (ANDRADE MOTA; MARCELINO CARDOSO, 2000;RAZKY, 2014); perception of the regional dialects by Brazilian Portuguese speakers from other regions has also been studied (LOPES et al., 2014;PARAFITA-COUTO, 2017). The classical works of Labov emphasize the importance of studying the sociolinguistic aspects of accent variation (LABOV, 2001;2006). The research currently conducted in Brazil also draws attention to the influence of social variables on accent: Oushiro (2017) mentions factors such as gender, age, education, social class, mobility, and origin of parents; for the internal migrants, she emphasizes the importance of age of arrival and length of residence (OUSHIRO, 2020).
Foreign accent is another important issue since both speech perception and production in L2 are affected by the L1 system of non-native speakers (LLISTERRI, 1995). The theories about phonological deafness (POLIVANOV, 1931) or the L1 phonological system functioning as a "filter" (TRUBETZKOY, 1939) exist since the first half of the 20th century. The contemporary models include the Speech Learning Model (SLM) (FLEGE, 1995;2007); Perceptual Assimilation Model (BEST; TYLER, 2007); Phonological Permeability Hypothesis (CABRELLI AMARO; ROTHMAN, 2010); Articulatory Phonology (BROWMAN; GOLDSTEIN, 1989), among others. The crucial factors that affect the speech production in L2 are the quality and quantity of input in L2, and also the amount of L1 use (FLEGE, 1995(FLEGE, , 2007. Factors such as age of L2 learning, length of residence, access to formal instruction, and motivation are considered to be secondary (MOYER, 2008;PISKE;MACKAY;FLEGE, 2001). Our focus here is not on the models explaining the phenomenon of accent in L2 or factors contributing to L2 accent severity, but on the perception studies of the foreign accent.
When listening to someone speaking, listeners are not only informed but impressed as well (BOLINGER, 1986). As Laver (1994) points out voice is an index of the biological, psychological and social characteristics of the speakers. Based on the segmental and prosodic choices the speakers make, listeners attribute paralinguistic (psychological conditions) as well as extralinguistic features (physical, social, regional and economic conditions). Relevant works on the index value of speech are Sapir (1920), Pittam (1995), Fonagy (2001, and Kreiman and Sidtis (2011), just to mention some. A discussion on the expressivity of speech productions can be found in Barbosa (2009) and Madureira, Fontes and Camargo (2019).
The first work that investigated judgments of bilingual speakers´ speech productions in different languages was the one by Lambert et al. (1960). They recorded English and French guises of Canadian bilingual speakers and studied the perception of these productions by other participants who used only one school and primary social language: either English or French. The speech productions in both languages were judged in relation to the following speakers´ qualities: the height, good looks, leadership capacity, sense of humor, intelligence, religiousness, self-confidence, dependability, entertainingness, kindness, ambition, sociability, character, and likability. The listeners did not know that the speech guises were produced by bilingual speakers, and that each speaker was listened to twice, once in each language. In the experiments by Lambert et al. (1960), the listeners who used English as their school and primary social language evaluated seven out of 14 traits more favorably for the English than for the French guises; the listeners who used mainly French evaluated more favorably the English guises for ten traits. In this experiment, the bilingual speakers did not have or were trained not to show a foreign accent. However, the idea that the same speaker could be perceived in different ways when he/she speaks different languages and the selection of judged traits greatly influenced the experimental design of the perception studies of the foreign accent.
Numerous studies demonstrated that non-native accented speech in English largely influence the perception of the foreign speakers' personal qualities and affect listeners´ judgement of attitudes. Carlson and McHenry (2006) reported that the employability of the job applicants is affected not by the perception of their ethnicity, but by the degree of perceived accentedness of the applicants' speech. Lev-Ari and Keysar (2010) focused on the influence of accent on people´s credibility: people judged trivial statements such as "Ants don't sleep" as less true when spoken by a non-native speaker. Russo, Islam and Koyuncu (2017) examined the impact of non-native accent on career advancement and career satisfaction. They concluded that speaking with a non-native accent may lead speakers to feel excluded and devalued at work, and even to assume an avoidance attitude towards work. Fuse, Navichkova and Alloggio (2018) asked monolingual and bilingual English speakers to rate four non-native accents (Spanish, Chinese, Russian, and Indian) on perceived intelligibility and perceived personal qualities (i.e., professionalism, intelligence, resourcefulness, empathy, and patience) necessary for speech-language pathologists. Greater perception of intelligibility was related to a positive perception of personal qualities such as professionalism, intelligence, and resourcefulness.

Stigmatization of a foreign accent in English may affect millions of people who speak
English as L2 and live in English-speaking countries (LEV-ARI; KEYSAR, 2010).
However, the perception of a foreign accent in other languages is very poorly explored. In

METHODS AND RESULTS
The experiment involved two kinds of tasks: 1) a production task performed by 12 native Russian speakers and six Brazilians: the speakers produced a semi-spontaneous speech in Brazilian Portuguese; 2) a perception task performed by 129 Brazilians and 23 native Russian speakers fluent in Brazilian Portuguese: the listeners judged the personal characteristics of the speakers based on their voice recordings.

Subjects of the production task (speakers)
The production task was performed by 12 native Russian speakers (six males and six females) and six Brazilians (three males and three females), all currently living in São Paulo. Their semi-spontaneous speech in Portuguese about their home city was recorded in a sound-isolated booth. These recordings were part of a more extended protocol that we applied for the collection of data for our database of Russian native speakers living in São Paulo (SMIRNOVA HENRIQUES et al., in press). This database includes only Russian speaking immigrants who moved to Brazil as adults and stayed as permanent residents.
These 12 native Russian speakers were selected among 40 volunteers to guarantee the same number of males and females and equivalent representation of different degrees of Russian accented speech. The degree of Russian accent was evaluated by an experienced phonetician, native speaker of Brazilian Portuguese, using the scale from 1 (a very weak accent) to 5 (a very strong accent). The degree of the foreign accent in the speech of native Brazilians was assumed to have a value of 0. All the Brazilian participants grew in São Paulo metropolitan area and did not present other regional accents.

Stimuli
The recordings of the subjects´ semi-spontaneous speech in Portuguese about their home cities were edited to obtain small fragments from 6 to 14 seconds (Table 1).
We did not establish the same fixed duration of the final recordings in order to keep the spontaneous aspects of the speech. All the words that have an affective value (as "bad", "good", "boring") were cut off in order not to influence listeners' judgements (PITTAM, 1994). The speech of the native Russian speakers was additionally edited to eliminate ungrammatical constructions and the wrong use of articles. The recordings obtained from 12 native Russian speakers and from six Brazilians were organized in a random order (see the subject random code in Table 1). Afterwards, the audio stimuli were embedded in videos with the same blue background and incorporated into a Google form. * the degree of the Russian accent was defined following the scale from 0 (native Brazilians) and 1 (a very weak accent) to 5 (a very strong accent). ** most participants have at least an undergraduate degree, so the area of education always means that they completed (or are attending, if mentioned) an undergraduate course.

Subjects of the perception task (listeners)
The total number of the listeners who made judgments in the perception test was 184. Most participants were recruited through social networks, personal contacts and also through a mailing list of Brazilians who study the Russian language in São Paulo. The participants were asked to take part in an online speech perception test to contribute to the investigation about how voice quality affects the perception of personal characteristics. They were informed that they would hear 18 voices, answer 16 questions about each voice, and also answer some sociodemographic questions about themselves. They were also informed that the whole protocol would take approximately 30 minutes to be answered. However, the participants were not instructed about the participation of foreigners in the test and the importance of the foreign accent for our study. All the personal questions about contact with foreigners and foreign languages were posed after the perception evaluation. For the data analysis, the participants were organized into four groups following the criteria of nationality, familiarity with foreigners and foreign languages, and place of residence (Table 2). All the nationalities other than Russian and Brazilian, and all the participants living in the states of Brazil other than São Paulo were eliminated from the present analysis. We kept only four Russian speakers that had already lived in São Paulo but currently were in Russia. After the filtering, the total number of the participants that suited the requirements described above was 152. All the data were collected in January and February, 2019.

Perceptual test procedures
The listeners accessed the Google form containing the perception test at home or in other place of their convenience. They were asked to use headphones. First, they listened to 18 audio stimuli in Portuguese (see Stimuli section) and, one by one, evaluated each voice based on 16 criteria. The current work is focused on the essential criteria of voice pleasantness. However, the participants also evaluated the voices for expressivity and comprehensibility; and the respective speakers for humility, confidence, happiness, After the perception test, in the same questionnaire the participants were asked to give some personal information: gender; age; nationality; education; profession; their place of birth; the current city of residence; other cities where he/she had already lived in Brazil or abroad; the spoken/studied languages; whether they had contact with foreigners and, if so, the nationality of the foreigners; how good their day had been. The participants were also asked to continue the phrase in Portuguese "Foreigners in Brazil…" and to evaluate, following their perception and memory, without effectively going back to the recordings, how many voices from these 18 stimuli belonged to foreigners. No connections with Russians were mentioned to avoid that the participants shared information. The real number of foreigners' recordings was also not revealed until the end of the data collection.

Analysis of the correlation between the Russian accent degree in Brazilian Portuguese speech and the perception of voice pleasantness
All the answers provided by the listeners who made judgments in the perception test were extracted from the Google form and transferred to an Excel spreadsheet. The listeners were separated into groups through Excel data filtering ( Table 2). The scale of voice pleasantness was reduced to three categories: pleasant (includes both "very pleasant" and "pleasant", initially suggested); not applicable; unpleasant (includes both "unpleasant" and "very unpleasant"). The number of the positive evaluations ("pleasant" category) was counted for each audio stimuli (Table 1) in each group of the participants ( Table   2). The data in each group were normally distributed (   profiles. However, we have recently described that some of the native Russian speakers from our databank replace native intonation patterns in Russian by those used in Brazilian Portuguese (KACHKOVSKAIA et al., in press). So, we also will be able to evaluate if the evaluation of speakers´ characteristics by listeners native in Russian is influenced by the attrition profile in the Russian speech of the speakers.

The perception tests on a foreign-accented speech open a new discussion space
in the fields of sociolinguistics, second language acquisition, and bilingualism studies, especially when we expand the analysis beyond the relatively well studied Englishspeaking countries. The knowledge about the effects that "sounding foreign" produces on listeners has not only linguistic importance (MOYER, 2013), but could also foster the development of the language policies, urgently needed in Brazil.

ACKNOWLEDGMENTS
Dr. Smirnova Henriques is supported by a postdoctoral fellowship PNPD/CAPES