The English voice system has two passive auxiliaries: the canonical be-passive, and the more recent get-passive. Accounts of the get-passive in the linguistic literature draw from descriptive, historical, corpus linguistic, and variationist perspectives. Much existing work on the get-passive from the former three traditions notes semantic dissimilarities from the be-passive, suggesting that these two forms are not interchangeable and therefore do not constitute a typical sociolinguistic variable. Nonetheless, variationist work has treated the be- and get-passives as alternants expressing the same function. This latter work has focused on social factors alone, setting aside purported linguistic differences. This thesis provides a variationist account of the be- and get-passives, considering not only social factors, but also operationalizing as linguistic factors previously noted semantic characteristics, demonstrating which factors constrain variation and providing a holistic picture of the get-passive in vernacular English. The speakers in this study span a birth range of 1865 to 1996, providing a longitudinal scope from which to view the grammaticalization of the feature. Following the principle of accountability (Labov, 1972), instances of be- and get-passives were extracted from 108 speakers born and raised in Victoria, British Columbia, Canada (N=1716). Distributional and inferential results show a substantial increase in rates of get-passive over the last 130 years, indicating an active and ongoing change in progress. Social and linguistic factors alike are shown to play meaningful roles in variant selection, revealing a (largely) longitudinally stable variable grammar. The longitudinal scope of the study illuminates grammaticalization pathways into the 20th century and reinforces attested semantic links between the contemporary get-passive and its proposed lexical source(s). / Graduate / 2023-08-17

The aim of this study is to investigate the process of accent levelling in the Regional French of Alsace and its relationship with the social variables of age, gender, social class, urban or rural origin of speakers and feelings of regional attachment. Accent levelling, which can be defined as the process of speakers abandoning local phonological forms in favour of supralocal variants, has been the focus of much recent sociolinguistic research on British English, French and other languages. Since knowledge of Alsatian (a Germanic language spoken in Alsace) is decreasing, it is possible that the resulting lack of interference between Alsatian and French is leading to levelling of the traditional accent features of Alsatian Regional French. In order to provide data for this research project, sociolinguistic interviews were conducted and written questionnaires used in Strasbourg and in the village of Helsheim (a fictional name used for reasons of confidentiality) with 56 informants. The data obtained were then subjected to quantitative analysis with regard to the linguistic variables of aspirate h (which can be realised as a supralocal zero variant or as a regional [h] variant) and the devoicing of canonically voiced plosives and fricatives (for example, sage pronounced [saʃ]). The results of the data analysis revealed that the regional variants of both linguistic variables are used more frequently by older than younger, working class than middle class, rural than urban speakers and that level of regional attachment correlates with use of the linguistic variables, as predicted in the research hypotheses. However, the relationship between levelling and gender proved to be more unexpected, with no clear pattern emerging for the (h) variable and a complex one involving the acquisition of supralocal patterns of sociolinguistic variation as well as the supralocal phonological variant in the case of consonant devoicing.

À is the prototypical preposition used before city names in French, yet there are reports that, since the mid-20th century, sur also appears in this context in vernacular varieties of European French. To date, research on the choice between à and sur has focused on semantic and pragmatic differences between the two, and has relied on made-up examples, small participant usage surveys, or empirical datasets that were not systematically analyzed. Moreover, the influence of social factors has received only scant attention. This thesis addresses these shortcomings by providing a quantitative, variationist and longitudinal account of à/sur alternations. It asks the following question: Which factors (linguistic and social) can account for prepositional variation before city names in Hexagonal French? The data was drawn from two publicly available corpora of spoken Hexagonal French, representing three locales: the Parisian city-centre, a group of suburban cities surrounding Paris, and the midsize provincial city of Orléans. The speakers (N = 151) were born between 1878 and 1994, providing a mixture of real- and apparent-time perspectives on variation. Following variationist methods, the analysis considers all contexts where à/sur variation is possible (N = 2542) and seeks to elucidate the variable grammar. Results indicate important differences between the three areas under study, both in terms of social patterns and in terms of linguistic constraints. In Paris, the use of sur is restricted to a few speakers and a few linguistic contexts, and is overall very infrequent. In the suburbs and Orléans, sur is more widely attested across speakers and contexts, but it remains a minority variant which seems to be on the decline, especially in the suburbs. Further, in Orléans, the variable grammar is less linguistically constrained, suggesting a trajectory of geographic diffusion. Overall, the quantitative findings support some of the semantic and pragmatic hypotheses proposed in earlier work while shedding light on how geographic, social and linguistic factors combine to explain a phenomenon of variation that has never been studied using variationist methods before. / Graduate / 2022-07-19

CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior / Nesta pesquisa, analisamos a variÃvel concordÃncia nominal de nÃmero entre os constituintes do sintagma nominal (SN) no falar da regiÃo do Cariri cearense à luz dos pressupostos teÃrico-metodolÃgicos da Teoria da VariaÃÃo e MudanÃa LinguÃstica (LABOV, WEINREICH, HERZOG, 2006; LABOV, 1978, l994; 2001; 2003). Nosso objetivo geral foi investigar em que medida os fatores linguÃsticos posiÃÃo dos elementos no sintagma, classe e posiÃÃo em relaÃÃo ao nÃcleo e à posiÃÃo nuclear; classe gramatical do sintagma nominal e processos morfofonolÃgicos de formaÃÃo de plural; tonicidade das sÃlabas dos itens lexicais singulares e marcas precedentes de plural no Ãmbito do sintagma nominal, e fatores extralingÃÃsticos, como sexo; anos de escolaridade e faixa etÃria condicionam o uso de concordÃncia nominal no falar dos municÃpios que formam a regiÃo CRAJUBAR â Crato, Juazeiro do Norte e Barbalha â na regiÃo do Cariri cearense. Para anÃlise quantitativa, coletamos 3304 dados de entrevistas realizadas com 24 informantes nesses trÃs municÃpios. As entrevistas estÃo presentes no corpus O PortuguÃs falado no Cearà do banco de dados do PROFALA . Os dados foram submetidos à anÃlise estatÃstica no programa GoldVarb X. No tocante aos nove grupos de fatores considerados inicialmente, seis grupos foram selecionados como significativos: posiÃÃo dos elementos no sintagma; classe e posiÃÃo em relaÃÃo ao nÃcleo e posiÃÃo nuclear; processos morfofonolÃgicos de formaÃÃo de plural; marcas precedentes de plural no Ãmbito do sintagma nominal; e os fatores extralingÃÃsticos: sexo e faixa etÃria.Quanto aos resultados obtidos, identificamos um alto Ãndice de marcaÃÃo, totalizando um percentual de 76,9 %. As mulheres apresentaram uma probabilidade de marcaÃÃo maior que a dos homens, e os informantes na faixa etÃria de 15 a 25 anos e a partir de 50 anos apresentaram Ãndices maiores que os dos informantes com faixa etÃria de 26 a 49 anos. / This work analyzes the nominal agreement of number between the constituents of the nominal phrase (NP) in Portuguese spoken in region of Cariri, in the state of CearÃ, Brazil, in light of the theoretical-methodological assumptions of the Theory of Variation and Linguistic Change (LABOV, WEINREICH, HERZOG, 2006; LABOV 1978, 1994, 2001). Its general objective is to investigate to what extent the linguistic factors position of the elements in the phrase, class and position in relation to nucleus and nuclear position; grammatical class of the nominal phrase and morpho-phonological processes of formation of plural; tonicity of the syllables of the singular lexical items and precedent marks of plural within the scope of the nominal phrase, and extralinguistic factors sex; years of schooling and age range condition the use of nominal agreement in the speech of the towns that make up the CRAJUBAR region â Crato, Juazeiro do Norte, and Barbalha â in the region of Cariri, in the south of CearÃ, Brazil. For quantitative analysis, it was collected 3304 data from interviews with 24 informants in these three towns. The interviews are present in the corpus PROFALA (The Portuguese spoken in CearÃ). The data were submitted to statistical analysis in the GoldVarb X program. For the nine groups of factors initially considered, six groups were selected as significant: position of the elements in the nominal phrase, class and position in relation to the nucleus and nuclear position; morpho-phonological processes of plural formation; precedent marks of plural in the scope of the nominal phrase and the extralinguistic factors; sex; and age group. Regarding the results obtained, it was identified a high marking index totaling a percentage of 76.9%. Women had a marking probability higher than men, and informants between the ages of 15 and 25 years and 50 years old presented rates higher than the informants with ages ranging from 26 to 49 years.

This paper investigates the possible effect of age on quotative variation in spoken British English with focus on the innovative quotative constructions be like and go and the standard construction say. The study is corpus-based and uses the Spoken British National Corpus 2014 as its material. Using the search tools provided in the corpus, datasets were restricted to include material from female speakers only and for each age-bracket in isolation. The results of the study were analysed in apparent time and through real time comparisons with previous studies. Similarly to previous studies, it was found that be like constructions are still favoured by young speakers, but the results also indicate that be like is used at higher frequencies among middle-aged speakers than previous studies have demonstrated. This indicates that be like is indeed an example of language change and not just an age-graded feature. The second innovative quotative investigated was go. The frequency distribution demonstrated by go was very different to that of be like. The results indicate that the ratio of go in comparison to be like (and say) have decreased drastically in the past twenty years when the results of the present study were compared to previous studies. This may indicate that the presence of two or more quotative variants within a speaker community may lead to the reduction in use of one of these variants due to "linguistic competition". The results of this study strengthen previous arguments that the presence of be like may lead to a decrease in the use of quotative go. The standard form say is still the most common variant for most age-brackets, apart from adolescent and young-adult speakers. However, in comparison to earlier studies the ratios of say have decreased for middle-aged speakers and younger. This may be due to an increased choice of quotative variants which are available to the speaker.

En el marco del variacionismo, en este trabajo se realiza una aproximación sociolingüística al empleo de de repente como introductor de la complicación de la narración en una muestra de 300 narraciones de experiencia personal (Labov y Waletzky 1967, y Labov 1972) de hablantes santiaguinos. Primero, se establece la frecuencia de empleo de dicha partícula. Luego, se correlacionan los factores sociodemográficos de los hablantes con el empleo de de repente y, finalmente, se describe el funcionamiento de esta partícula en las narraciones que conforman la muestra. La hipótesis de trabajo fue que las características sociodemográficas de los hablantes de español en Santiago de Chile están correlacionadas con el empleo de de repente como introductor de complicación de la narración. El análisis de los datos dio cuenta de que de repente tiene una frecuencia de 100 casos, de los cuales 95 (95%) cumplen la función de introducir la acción complicada de los relatos. Asimismo, demostramos que se trata de una partícula que suele emplearse más por informantes mujeres de edad joven (20-34 años). El análisis, a su vez, permitió dar cuenta de que una colocación preponderante es la que se genera entre la forma de repente + presente histórico. / Situated within the variationist sociolinguistic framework, in this paper we analyze the use of de repente as a marker that introduces complication or climax in a sample of 300 personal experience narratives (Labov y Waletzky 1967 y Labov 1972) of Santiago speakers. First, we establish the frequency of use of this particle. Then, we correlate the speakers’ sociodemographic factors with the use of this marker, and finally, we describe the functionality of this particle within the narratives that make up the sample. Our hypothesis was that the sociodemographic characteristics of Santiago Spanish speakers would be correlated with the use of de repente as a phrase that introduces complication or climax within the narrative. The data analysis shows that de repente was used a total of 100 times, of which 95 (95%) function in this manner, or as introducing the complicated action of the story. Additionally, we demonstrate that this particle is used most frequently by female informants from the younger age group (20-34 years old). The analysis, in turn, enabled us to determine that a main collocation is that of de repente + historical present.

Essa dissertação trata da pronúncia variável de /ndo/, com /d/ apagado ou pronunciado. Baseada na teoria e metodologias da Sociolinguística Variacionista (LABOV, 2006[1966], 2008[1972]), focaliza tal variável aqui referida como (NDO) na variedade paulistana do português brasileiro, através das 60 entrevistas do Projeto SP2010 (MENDES & OUSHIRO, 2012), estratificadas por Sexo/Gênero, Faixa Etária e Escolaridade. A variável (NDO) tem sido estudada desde o final dos anos 70 sob diferentes perspectivas. Alguns pesquisadores se atêm à pronúncia variável do sufixo de gerúndio (por exemplo, falano ou falando em diferentes variedades do português: Rio de Janeiro (MOLLICA, 1989), Belo Horizonte (CRISTÓFARO SILVA, 1996), João Pessoa (MARTINS, 1999, 2001), Minas Gerais (MARTINS, 2006), São José do Rio Preto (FERREIRA, 2010), Dourados e Ponta Porã MS (MARTINS e BUENO, 2011), Taboco MS (VIEIRA, 2011), Fortaleza (NASCIMENTO ET AL., 2013), Maceió (ALMEIDA e OLIVEIRA, 2017)). Leva-se em consideração o apagamento de /d/ como uma variável cujo padrão de realização pode ter especificidades na fala paulistana uma variedade que, ainda que mais detidamente estudada nos últimos anos, ainda não foi analisada no que toca a (NDO), nem no gerúndio, especificamente, nem em outros contextos morfológicos ou itens lexicais, tal como na palavra quando. O objetivo central é, então, investigar quais as variáveis linguísticas e sociais que se correlacionam à realização de (NDO). Para tanto, analisou-se qualitativa e quantitativamente a amostra, onde foram encontrados casos de apagamento de /d/ na pronúncia da Conjunção quando (93/1557) e em verbos no gerúndio (1001/4143). Segundo as análises estatísticas multivariadas, verificou-se que não há correlação significativa entre (NDO) e nenhuma das variáveis linguísticas, o apagamento ocorre com maior frequência na fala dos homens, sendo mais propício na fala daqueles com menor grau de instrução, assim como ocorre para aqueles falantes da Classe Social menos favorecida. Com relação à Faixa Etária, existe um favorecimento na fala dos mais jovens, principalmente naqueles da Faixa Etária intermediária, já para a terceira faixa existe um desfavorecimento, esses resultados levam a indícios de que o apagamento de /d/ em (NDO) seja uma variável estável na capital paulista. / This masters thesis deals with /ndo/ variable pronunciation, with /d/ deletion or pronunciation. Based on Variationist Sociolinguistics theory and methodologies (LABOV, 2006[1966], 2008[1972]), we focus on this variable here called as (NDO) on São Paulos speech of Brazilian Portuguese, through 60 interviews from SP2010 project (MENDES & OUSHIRO, 2012), stratified by Sex/Gender, Age and Level of Education. The variable (NDO) has been studied since the 1970s from different perspectives. Some researchers focused only on the pronunciation of the suffix of gerund (for example, falano or falando in different portuguese varieties as Rio de Janeiro (MOLLICA, 1989), Belo Horizonte (CRISTÓFARO SILVA, 1996), João Pessoa (MARTINS, 1999, 2001), Minas Gerais (MARTINS, 2006), São José do Rio Preto (FERREIRA, 2010), Dourados and Ponta Porã MS (MARTINS e BUENO, 2011), Taboco MS (VIEIRA, 2011), Fortaleza (NASCIMENTO ET AL., 2013), Maceió (ALMEIDA e OLIVEIRA, 2017)). Taking into account the deletion of /d/ as a variable in which the performance pattern may have specificities in Paulistano speech - a variety that, although studied in recent years, has not yet been analyzed regarding (NDO), nor in the gerund, specifically, nor in other morphological contexts or lexical items, as in the word quando. The main goal is to investigate the linguistic and social variables that are correlated to the performance of (NDO). For that, the sample was analyzed qualitatively and quantitatively, in which cases of deletion of /d/ were found in the pronunciation of the conjunction \"quando\" (93/1557) and in verbs in the gerund (1001/4143). According to the multivariate statistical analysis, it was verified that there is no significant correlation between (NDO) and the linguistic variables, deletion occurs more frequently in men\'s speech, being more propitious in the speech of those with lower Education Degree, as it occurs for those speakers of the less favored Social Class. Regarding the Age Range, there is a favoring of young people\'s speech, at all levels, the results lead to an indication that the /d/ deletion in (NDO) is a variable stability in São Paulo city.

Essa dissertação trata da pronúncia variável de /ndo/, com /d/ apagado ou pronunciado. Baseada na teoria e metodologias da Sociolinguística Variacionista (LABOV, 2006[1966], 2008[1972]), focaliza tal variável aqui referida como (NDO) na variedade paulistana do português brasileiro, através das 60 entrevistas do Projeto SP2010 (MENDES & OUSHIRO, 2012), estratificadas por Sexo/Gênero, Faixa Etária e Escolaridade. A variável (NDO) tem sido estudada desde o final dos anos 70 sob diferentes perspectivas. Alguns pesquisadores se atêm à pronúncia variável do sufixo de gerúndio (por exemplo, falano ou falando em diferentes variedades do português: Rio de Janeiro (MOLLICA, 1989), Belo Horizonte (CRISTÓFARO SILVA, 1996), João Pessoa (MARTINS, 1999, 2001), Minas Gerais (MARTINS, 2006), São José do Rio Preto (FERREIRA, 2010), Dourados e Ponta Porã MS (MARTINS e BUENO, 2011), Taboco MS (VIEIRA, 2011), Fortaleza (NASCIMENTO ET AL., 2013), Maceió (ALMEIDA e OLIVEIRA, 2017)). Leva-se em consideração o apagamento de /d/ como uma variável cujo padrão de realização pode ter especificidades na fala paulistana uma variedade que, ainda que mais detidamente estudada nos últimos anos, ainda não foi analisada no que toca a (NDO), nem no gerúndio, especificamente, nem em outros contextos morfológicos ou itens lexicais, tal como na palavra quando. O objetivo central é, então, investigar quais as variáveis linguísticas e sociais que se correlacionam à realização de (NDO). Para tanto, analisou-se qualitativa e quantitativamente a amostra, onde foram encontrados casos de apagamento de /d/ na pronúncia da Conjunção quando (93/1557) e em verbos no gerúndio (1001/4143). Segundo as análises estatísticas multivariadas, verificou-se que não há correlação significativa entre (NDO) e nenhuma das variáveis linguísticas, o apagamento ocorre com maior frequência na fala dos homens, sendo mais propício na fala daqueles com menor grau de instrução, assim como ocorre para aqueles falantes da Classe Social menos favorecida. Com relação à Faixa Etária, existe um favorecimento na fala dos mais jovens, principalmente naqueles da Faixa Etária intermediária, já para a terceira faixa existe um desfavorecimento, esses resultados levam a indícios de que o apagamento de /d/ em (NDO) seja uma variável estável na capital paulista. / This masters thesis deals with /ndo/ variable pronunciation, with /d/ deletion or pronunciation. Based on Variationist Sociolinguistics theory and methodologies (LABOV, 2006[1966], 2008[1972]), we focus on this variable here called as (NDO) on São Paulos speech of Brazilian Portuguese, through 60 interviews from SP2010 project (MENDES & OUSHIRO, 2012), stratified by Sex/Gender, Age and Level of Education. The variable (NDO) has been studied since the 1970s from different perspectives. Some researchers focused only on the pronunciation of the suffix of gerund (for example, falano or falando in different portuguese varieties as Rio de Janeiro (MOLLICA, 1989), Belo Horizonte (CRISTÓFARO SILVA, 1996), João Pessoa (MARTINS, 1999, 2001), Minas Gerais (MARTINS, 2006), São José do Rio Preto (FERREIRA, 2010), Dourados and Ponta Porã MS (MARTINS e BUENO, 2011), Taboco MS (VIEIRA, 2011), Fortaleza (NASCIMENTO ET AL., 2013), Maceió (ALMEIDA e OLIVEIRA, 2017)). Taking into account the deletion of /d/ as a variable in which the performance pattern may have specificities in Paulistano speech - a variety that, although studied in recent years, has not yet been analyzed regarding (NDO), nor in the gerund, specifically, nor in other morphological contexts or lexical items, as in the word quando. The main goal is to investigate the linguistic and social variables that are correlated to the performance of (NDO). For that, the sample was analyzed qualitatively and quantitatively, in which cases of deletion of /d/ were found in the pronunciation of the conjunction \"quando\" (93/1557) and in verbs in the gerund (1001/4143). According to the multivariate statistical analysis, it was verified that there is no significant correlation between (NDO) and the linguistic variables, deletion occurs more frequently in men\'s speech, being more propitious in the speech of those with lower Education Degree, as it occurs for those speakers of the less favored Social Class. Regarding the Age Range, there is a favoring of young people\'s speech, at all levels, the results lead to an indication that the /d/ deletion in (NDO) is a variable stability in São Paulo city.

Este trabalho descreve o emprego variável de estruturas de negação na variedade paulistana do português. São três as formas variantes: não pré-verbal (NEG1, p.ex. Não gosto de chocolate), não pré e pós-verbal (NEG2, p. ex. Não gosto de chocolate não) e não pós-verbal apenas (NEG3, p. ex. Gosto de chocolate não). A partir da discussão de restrições discursivo-pragmáticas, define-se o envelope de variação (os contextos em que as formas são semanticamente equivalentes). NEG1 e NEG2 são ambas alternativas quando a proposição que está sendo negada foi direta ou indiretamente ativada no discurso anterior. Quando a informação proposicional é nova no discurso, apenas NEG1 é possível. Já NEG3 parece improdutiva na variedade paulistana: representa menos de 1% em mais de cinco mil ocorrências de sentenças negativas. Os dados analisados foram extraídos de 48 entrevistas sociolinguísticas com paulistanos estratificados em Sexo/gênero, Faixa etária e Escolaridade. Elas foram coletadas pelo Grupo de Estudos e Pesquisa em Sociolinguística da Universidade de São Paulo (GESOL-USP), entre os anos de 2003 e 2010. Além da amostra geral (com todas essas entrevistas), são analisadas duas subamostras menores (24 entrevistas cada), organizadas com base na região da cidade em que vivem os paulistanos (mais central ou mais periférica) e na sua geração (filhos de pais paulistanos ou filhos de migrantes de outros estados brasileiros). Os resultados indicam que a variante NEG2, bastante infrequente (cerca de 6% no conjunto de 48 entrevistas), é favorecida por informantes de escolaridade mais baixa e de 1ª geração na cidade de São Paulo. Algumas análises sugerem ainda que informantes que vivem em bairros mais periféricos favorecem NEG2. Não há indícios, contudo, de correlação entre a variável e Sexo/gênero; tampouco há indício de mudança em curso. No âmbito linguístico, a ativação direta de proposições é o fator que mais fortemente favorece o emprego da variante NEG2, seguido da ausência de marcadores discursivos ou de outros termos negativos na sentença. / This masters thesis describes the variable use of negation structures in São Paulo Portuguese. There are three variants: pre-verbal não (NEG1, e.g. Não gosto de chocolate), pre and post-verbal não (NEG2, e.g. Não gosto de chocolate não) and post-verbal não only (NEG3, e.g. Gosto de chocolate não). All of these sentences mean I dont like chocolate. By discussing discourse-pragmatic restrictions on the use of such structures, the envelope of variation is defined as follows: NEG1 and NEG2 can both be employed when the proposition that is being negated was activated earlier (directly or indirectly) in discourse. When the propositional information is discourse-new, only NEG1 is possible. As for NEG3, it seems very unproductive in São Paulo Portuguese; it occurs in less than 1% of 5,000 tokens of negative sentences. The data was extracted from 48 sociolinguistic interviews with Paulistanos (those born and raised in the city of São Paulo) stratified by sex/gender, age group and level of education. From the general corpus of 48 interviews, two subsamples of 24 interviews each have also been analyzed. One of them is based on the region of the city where the informants live (central or peripheral), whereas the other is based on their generation in the city (Paulistano parents or migrants from other states in Brazil). Results show that NEG2 (an infrequent variant about 6% of the tokens extracted from the 48 interviews) is favored by those with lower level of education and by first generation Paulistanos (that is, by those whose parents were not born and raised in the city). Some analyses suggest that NEG2 is also favored by those groups that live in more peripheral areas of São Paulo. There is no indication of correlation between the variable and sex/gender, nor is there any indication of change in progress. Linguistically, the factor that most favors NEG2 is the direct activation of propositional information in discourse, followed by the absence of discourse markers or other negative elements in the sentence.

Este trabalho descreve o emprego variável de estruturas de negação na variedade paulistana do português. São três as formas variantes: não pré-verbal (NEG1, p.ex. Não gosto de chocolate), não pré e pós-verbal (NEG2, p. ex. Não gosto de chocolate não) e não pós-verbal apenas (NEG3, p. ex. Gosto de chocolate não). A partir da discussão de restrições discursivo-pragmáticas, define-se o envelope de variação (os contextos em que as formas são semanticamente equivalentes). NEG1 e NEG2 são ambas alternativas quando a proposição que está sendo negada foi direta ou indiretamente ativada no discurso anterior. Quando a informação proposicional é nova no discurso, apenas NEG1 é possível. Já NEG3 parece improdutiva na variedade paulistana: representa menos de 1% em mais de cinco mil ocorrências de sentenças negativas. Os dados analisados foram extraídos de 48 entrevistas sociolinguísticas com paulistanos estratificados em Sexo/gênero, Faixa etária e Escolaridade. Elas foram coletadas pelo Grupo de Estudos e Pesquisa em Sociolinguística da Universidade de São Paulo (GESOL-USP), entre os anos de 2003 e 2010. Além da amostra geral (com todas essas entrevistas), são analisadas duas subamostras menores (24 entrevistas cada), organizadas com base na região da cidade em que vivem os paulistanos (mais central ou mais periférica) e na sua geração (filhos de pais paulistanos ou filhos de migrantes de outros estados brasileiros). Os resultados indicam que a variante NEG2, bastante infrequente (cerca de 6% no conjunto de 48 entrevistas), é favorecida por informantes de escolaridade mais baixa e de 1ª geração na cidade de São Paulo. Algumas análises sugerem ainda que informantes que vivem em bairros mais periféricos favorecem NEG2. Não há indícios, contudo, de correlação entre a variável e Sexo/gênero; tampouco há indício de mudança em curso. No âmbito linguístico, a ativação direta de proposições é o fator que mais fortemente favorece o emprego da variante NEG2, seguido da ausência de marcadores discursivos ou de outros termos negativos na sentença. / This masters thesis describes the variable use of negation structures in São Paulo Portuguese. There are three variants: pre-verbal não (NEG1, e.g. Não gosto de chocolate), pre and post-verbal não (NEG2, e.g. Não gosto de chocolate não) and post-verbal não only (NEG3, e.g. Gosto de chocolate não). All of these sentences mean I dont like chocolate. By discussing discourse-pragmatic restrictions on the use of such structures, the envelope of variation is defined as follows: NEG1 and NEG2 can both be employed when the proposition that is being negated was activated earlier (directly or indirectly) in discourse. When the propositional information is discourse-new, only NEG1 is possible. As for NEG3, it seems very unproductive in São Paulo Portuguese; it occurs in less than 1% of 5,000 tokens of negative sentences. The data was extracted from 48 sociolinguistic interviews with Paulistanos (those born and raised in the city of São Paulo) stratified by sex/gender, age group and level of education. From the general corpus of 48 interviews, two subsamples of 24 interviews each have also been analyzed. One of them is based on the region of the city where the informants live (central or peripheral), whereas the other is based on their generation in the city (Paulistano parents or migrants from other states in Brazil). Results show that NEG2 (an infrequent variant about 6% of the tokens extracted from the 48 interviews) is favored by those with lower level of education and by first generation Paulistanos (that is, by those whose parents were not born and raised in the city). Some analyses suggest that NEG2 is also favored by those groups that live in more peripheral areas of São Paulo. There is no indication of correlation between the variable and sex/gender, nor is there any indication of change in progress. Linguistically, the factor that most favors NEG2 is the direct activation of propositional information in discourse, followed by the absence of discourse markers or other negative elements in the sentence.