Esta pesquisa apresenta análises sobre avaliação, produção e percepção linguística no português paulistano, por meio do exame de quatro variáveis sociolinguísticas: a realização de /e/ nasal como monotongo [e] ou ditongo [ej] (como em fazenda); a pronúncia de /r/ em coda silábica como tepe [R] ou retroflexo [õ] (como em porta); a concordância nominal de número (como em as casas/as casa); e a concordância verbal de primeira e de terceira pessoa do plural (como em nós fomos/nós foi, eles foram/eles foi). O objetivo central é analisar, em uma comunidade amplamente heterogênea de um ponto de vista sociodemográfico, as inter-relações entre a expressão de identidades sociais através de usos linguísticos e a possível influência dos significados sociais desses usos em processos de variação e mudança linguística. Para tanto, analisou-se qualitativa e quantitativamente uma amostra contemporânea do português paulistano, composta de 118 entrevistas sociolinguísticas com falantes nativos, à luz dos pressupostos teórico-metodológicos da Sociolinguística Variacionista (Labov, 2006 [1966], 2008 [1972]). Tais análises compreendem o encaixamento linguístico e social de cada variável, bem como seu encaixamento simultâneo na fala de cada indivíduo. Além disso, examinaram-se percepções sobre as variantes de (-r), com base na técnica de estímulos pareados (Lambert et al., 1960), a fim de melhor compreender os mecanismos subjacentes à associação de certos significados sociais ao emprego de diferentes formas linguísticas. Os resultados mostram que, embora as correlações entre as quatro variáveis sociolinguísticas e variáveis sociais sejam bastante semelhantes entre si (todas se correlacionam com o Sexo/Gênero, a Classe Social e o Nível de Escolaridade dos falantes), há diferentes tendências dentro da comunidade por exemplo, mudança em direção à variante ditongada [ej]; padrões divergentes quanto ao emprego de (-r) por parte de jovens de diferentes classes sociais; variação estável das concordâncias nominal e verbal em regiões periféricas e mudança em direção à variante padrão em regiões centrais. Para compreendê-los, o exame de seus significados sociais é fundamental. Argumenta-se que [ej] tem se difundido rápida e unidirecionalmente pelo fato de se constituir um marcador (Labov, 2008 [1972]) para paulistanos, que não revelam ter consciência da variável, tampouco apresentam um discurso metalinguístico sobre suas variantes. O forte favorecimento do retroflexo entre jovens de classes baixas foi desencadeado por uma reinterpretação de seu significado social como uma variante local e de prestígio, devido à presença maciça de migrantes do Norte/Nordeste, cuja variante fricativa é relativamente mais estigmatizada na comunidade. Ao mesmo tempo, ainda que o encaixamento social das concordâncias nominal e verbal seja bastante semelhante, a marca zero de concordância nominal (as casa) goza de maior vitalidade por indexicalizar significados como masculinidade, paulistanidade e morador da Mooca. Não obstante as diferentes tendências que se verificam na comunidade, os padrões de encaixamento das variáveis linguísticas se reproduzem sistematicamente na fala de cada indivíduo, o que permite caracterizar os paulistanos como uma única comunidade de fala (Labov, 2006 [1966]), que compartilha normas de produção e de avaliação linguística. De acordo com o teste de percepções, os moradores da cidade também são consistentes em suas reações subjetivas a variantes de (-r). Demonstra-se adicionalmente que a coesão dialetal é promovida não por amplas categorias sociais como Sexo/Gênero ou Faixas Etárias, mas pelo princípio mais fundamental de densidade de comunicação (Gumperz, 1971b,a). / This study examines linguistic evaluation, production, and perception in São Paulo Portuguese, through analyses of four sociolinguistic variables: the realization of nasal /e/ as a monophthong [e] or as a diphthong [ej] (as in fazenda farm); the realization of coda /r/ as a tap [R] or as a retroflex [õ] (as in porta door); nominal number agreement (as in as casas/as casa the houses); and first person plural and third person plural verb agreement (as in nós fomos/nós foi we went, eles foram/eles foi they went). The main goal is to investigate the inter-relation between the expression of social identities through language uses and the possible impact of social meanings on processes of language variation and change, in a highly diverse and heterogeneous community. Based on the theory and methods of Variationist Sociolinguistics (Labov, 2006 [1966], 2008 [1972]), each variables linguistic and social embedding, as well as their simultaneous embedding in individual speakers speech, were analyzed both qualitatively and quantitatively in 118 sociolinguistic interviews with native Paulistano speakers. In addition, perceptions on the variants of (-r) were examined through an experiment using the Matched Guise Technique (Lambert et al., 1960), aimed at describing the mechanisms underlying the association of certain social meanings with different language forms. The results show that, in spite of the similarity between the social embedding of the four variables (all of them are correlated with speakers sex/gender, social class, and level of education), there are different trends within the community for instance, change towards (e) diphthongization; divergent patterns regarding (-r) in the speech of younger speakers of different social classes; stable variation in nominal and verbal agreement in peripheral areas but change towards the prestige variant in central areas. The explanation for these patterns is related to the variants social meanings. It is argued that [ej] has spread rapidly and unidirectionally because it is a marker (Labov, 2008 [1972]) for Paulistanos, who are not aware of the variable and do not present an elaborate metalinguistic discourse on its variants. The fact that retroflex /r/ is strongly favored by working class youth may be attributed to a reinterpretation of its social meaning, due to the extensive presence of migrants from the Northern and Northeastern regions of the country, whose /r/ realization as a fricative is relatively more stigmatized in the community. At the same time, although nominal and verbal agreement are very similarly stratified, the nonstandard variant of the former (as casa the houses) exhibits greater vitality as it indexes masculinities and local identities with the city and with Mooca, one of its most traditional neighborhoods. Despite different trends by different social groups in the community, the embedding of the linguistic variables is systematically reproduced in each speakers speech, which allows for the characterization of São Paulo as a single speech community (Labov, 2006 [1966]) in that its native speakers share norms of use and evaluation of the variants. According to the perception test, the city inhabitants are also consistent in their subjective reactions to the variants of (-r). It is shown that such social cohesion is promoted not by census social categories such as sex/gender or age, but by the more fundamental principle of density of communication (Gumperz, 1971b,a).

MA (ELT) / Department of Engliish / Learning English as a second language by the South African learners of English First Additional Language (FAL) causes many challenges, such as committing errors in concord as there are differences between the learners’ mother tongue and the target language. Even at Grade 12 level, which is the exit point to institutions of higher learning or to the workplace, learners still display some deficiencies in the mastery of the English concord. This happens despite the fact that, in many South African schools, English is used as a medium of instruction and learnt as a First Additional Language (FAL) especially at high schools. Through the analysis of the written concord errors committed by the 72 of the 720 sampled Grade 12 English FAL learners in Vhembe District, the study answered to two questions which formed its cornerstone which are: what are the most common types of written concord/ subject-verb agreement errors which are committed by Grade 12 FAL learners and what are the causes thereof? The study employed both the quantitative and the qualitative methods to pursue the primary question. Learners responded to the questionnaires and the researcher also analysed their teacher-marked English FAL composition scripts with the focus on concord/subject-verb agreement usage. The findings were that concord/subject-verb agreement was a challenge to the majority of the participants. There was not even a single question which recorded a 100% correct entry. The question on collective nouns was found to be the hardest to the participants while comparatively, the singular indefinite pronoun question recorded the best results. The learners’ written compositions were also found to have been marred by concord/subject-verb agreement errors. In most instances, the learners had resorted to simple sentences avoiding the complex sentence construction as those would have required complicated application of concord/ subject-verb agreement usage. Based on the findings, the following recommendations were made: teaching of grammar should be intensified, and that teachers of English should be retrained even if it will be through the in-service programmes

La maitrise de l’écrit est importante et nous savons que celle-ci est loin d’être assurée par l’ensemble des scripteurs du français, au Québec comme ailleurs. À titre d’exemple, les performances des élèves du secondaire québécois en écriture (Ministère de l’Éducation et de l'Enseignement supérieur, 2019), et plus particulièrement en orthographe grammaticale, se conforment peu aux attentes sociales. Pour soutenir l’apprentissage des élèves dans ce domaine, dans une perspective de différenciation pédagogique, l’établissement d’un diagnostic initial permettant de cibler les forces et les faiblesses des apprenants s’avère une étape primordiale. Pour ce faire, les enseignants doivent disposer d’outils d’évaluation diagnostique adéquats qui permettent de brosser le portrait des difficultés rencontrées par leurs élèves. La présente recherche doctorale vise à répondre à ce besoin en développant un dispositif d’évaluation diagnostique qui cible, par l’analyse des erreurs commises et selon une approche cognitive, les difficultés rencontrées par les élèves de la fin du 1er cycle du secondaire avec l’accord du verbe et des participes passés. Les objectifs spécifiques de cette recherche consistent en la conception, la réalisation, la mise à l’essai et la finalisation de ce dispositif en suivant le cadre méthodologique général de Harvey et Loiselle (2009). La validation du dispositif et de la démarche employée pour l’élaborer se réalise de diverses façons : par l’emploi d’une approche « fondée sur des preuves » (Mislevy, Steinberg, et Almond, 2003) pour la conception du dispositif, par le recours à des experts de contenu pour valider les bases conceptuelles du dispositif, par un questionnaire aux enseignants (n= 4) et deux entrevues semi-dirigées menées avec eux à la suite des mises à l’essai empiriques (auprès de 360 élèves de 2e secondaire) pour vérifier la convivialité du dispositif et son adéquation aux besoins dégagés et, finalement, par l’analyse psychométrique des items des deux tests composant le dispositif créé. Des critères de scientificité sont aussi établis et respectés pour s’assurer que la recherche-développement réalisée comporte les caractéristiques attendues d’un tel type de recherche. L’objet résultant de cette recherche-développement, le dispositif nommé DIAgramm, est un outil d’évaluation diagnostique informatisé et accessible en ligne, composé de deux tests visant des niveaux de diagnostic différents (niveaux constat et recherche des causes, voir B. Rey, 2016) et générant des rapports de résultats complets et précis pour les enseignants. Le niveau constat de DIAgramm vise à cerner les configurations syntaxiques qui posent problème aux élèves alors que le niveau recherche des causes vise à identifier la cause des erreurs commises. Les résultats de cette recherche-développement nous laissent penser que DIAgramm répond tout à fait aux besoins éprouvés par les enseignants, tant en ce qui concerne la qualité des portraits diagnostiques brossés que la convivialité de l’instrument lui-même (administration facile, clarté des rapports de résultats fournis, etc.). / Proficiency in writing is important and we know that this skill is far from being mastered by all French writers, in Quebec as elsewhere. Indeed, the writing abilities of Quebec high school students, and more specifically in grammatical spelling, does not meet social expectations (Ministère de l'Éducation et de l’Enseignement supérieur, 2019). To support students' learning in this area, it is crucial to conduct an initial diagnosis of learners' strengths and weaknesses. To do this, teachers must have adequate diagnostic assessment tools that allow them to draw a picture of students’ writing difficulties. This doctoral research aims to address this need by developing a diagnostic assessment device that targets, through the analysis of errors made and according to a cognitive approach, the difficulties pertaining to the agreement of verbs and past participles encountered by students at the end of the first cycle of secondary school. The specific objectives of this research and development project are to design, implement, test, and finalize this device following Harvey and Loiselle’s (2009) methodological framework. The device and the method used to create it were validated in various ways: by adopting an evidence-based approach (Mislevy, Steinberg, et al., 2003) for the design of the device, by enlisting content experts to validate the conceptual foundations of the device, by surveying teachers and conducting two semi-structured interviews with them after the empirical trials (involving 360 secondary 2 students) to verify the device's usability and its relevance to the identified needs, and finally by psychometrically analyzing the items of the two tests comprising the created device. Scientific criteria were also established and adhered to in order to ensure that the carried out research and development has the expected characteristics for this type of research. The resulting product from this research and development is DIAgramm, an online, computerized diagnostic assessment tool consisting of two tests aiming at different levels of diagnosis (the observation (constat) level and the cause-finding (recherche des causes) level, see B. Rey, 2016). The diagnostic devide further generates complete and accurate results reports for the teachers. The observation level of DIAgramm aims to identify syntactic configurations that are problematic for students, while the cause-finding level aims to identify the cause of the errors made. The results of this research and development suggest that DIAgramm meets the needs of teachers, both in terms of the quality of the diagnostic portraits provided and the user-friendliness of the instrument itself (easy administration, clarity of the results reports provided, etc.).

Codirection / Bien que la communauté scientifique soit toujours à la recherche d'une caractéristique déterminante du trouble développemental du langage (TDL), les difficultés d'accord sujet-verbe, et par extension morphosyntaxiques, ont été identifiées comme un marqueur du TDL chez les enfants anglophones, autant chez les enfants du préscolaire que les plus vieux. Cependant, des études sur les enfants francophones d'âge préscolaire suggèrent que les déficits morphosyntaxiques ne seraient pas un marqueur fiable du TDL. Puisque que certains aspects de la morphosyntaxe en français ne sont acquis que vers l’âge de huit ans chez les enfants au développement typique, tels que l'accord en nombre des verbes sous-réguliers et irréguliers, ci-après SOUSIRR, les déficits morphosyntaxiques pourraient être un marqueur du TDL en français uniquement vers la (pré-)adolescence. Cette thèse a pour objectifs de déterminer si les (pré-)adolescents francophones au développement typique ont acquis l'accord en nombre des verbes SOUSIRR, si les (pré)adolescents francophones avec un TDL ont des déficits d'accord en nombre des verbes SOUSIRR, et à établir si la morphosyntaxe est un domaine de faiblesse par rapport à la lexico-sémantique dans cette population. L'accord en nombre des verbes SOUSIRR et les compétences morphosyntaxiques ont été évalués à l'aide de tâches ciblant les niveaux comportemental et neurocognitif en utilisant des tâches linguistiques et des potentiels évoqués (PÉ). De plus, nous avons développé des prédictions basées sur deux théories touchant les compétences morphosyntaxiques chez les (pré-)adolescents atteints de TDL : l'hypothèse du déficit procédural (Ullman & Pierpont, 2005 ; Ullman et al., 2020), et l'hypothèse du ralentissement généralisé (Kail, 1994). Cette thèse est composée de trois manuscrits pour publication. Le premier évalue les compétences des participants dans plusieurs domaines linguistiques, à l'aide de tâches comportementales typiquement utilisées en orthophonie et dans la recherche sur l’acquisition du langage. Les données révèlent des déficits lexico-sémantiques et morphosyntaxiques chez les participants avec un TDL, mais suggèrent qu'une tâche d'accord en nombre des verbes SOUSIRR était la meilleure pour discriminer les participants avec et sans TDL. Le deuxième article présente une étude innovante de PÉs utilisant uniquement des phrases grammaticales, présentées simultanément avec des images sémantiquement ou grammaticalement congruentes et incongruentes, afin d'évaluer le traitement morphosyntaxique et lexico-sémantique des phrases au niveau neurocognitif. Les résultats provenant de vingt-huit adultes francophones montrent qu'ils ont présenté les composantes PÉs attendues et comparables aux études utilisant des phrases agrammaticales. Ces données ont servi de référence pour établir si nos participants avec et sans TDL avaient un traitement linguistique mature. Le troisième article a testé cette nouvelle expérimentation avec nos participants (pré )adolescents. Les résultats suggèrent que, contrairement à la morphosyntaxe, la lexico-sémantique est une force relative chez les adolescents avec un TDL lors du traitement de l'information linguistique au niveau neurocognitif. Dans l'ensemble, cette thèse révèle que la morphosyntaxe est particulièrement altérée chez les adolescents francophones avec un TDL. Nous discutons les résultats en relation avec la pratique clinique orthophonique et soulignons l'importance d'examiner les processus neurocognitifs dans l'étude du TDL. / Although the scientific community is still searching for a defining characteristic of developmental language disorder (DLD), problems with subject-verb agreement, and by extension morphosyntax, have been identified as a hallmark of English-speaking preschoolers and older children with DLD. However, in studies of French-speaking preschoolers with DLD, morphosyntax has not been found to be a specific linguistic weakness. Since there is evidence that some aspects of morphosyntax in French are acquired by children with typical language (TL) development only later in childhood, such as subregular and irregular subject-verb number agreement, henceforth SUBIRR, morphosyntax has been argued to be a French marker for DLD only in older childhood and adolescence. The present thesis aimed to determine if French speaking (pre-)teenagers with TL have acquired SUBIRR number agreement, resolve whether French-speaking (pre-)teenagers with DLD are impaired on SUBIRR number agreement, and establish whether morphosyntax is an area of weakness as compared to lexico-semantics in this population. SUBIRR number agreement and morphosyntactic skills were evaluated with tasks targeting the behavioural and neurocognitive levels using linguistics tasks and event-related potentials (ERP). Furthermore, we contrasted two theories’ predictions on morphosyntactic skills in (pre-)teens with DLD : the procedural deficit hypothesis (Ullman & Pierpont, 2005; Ullman et al., 2020), and the generalized slowing hypothesis (Kail, 1994). This thesis is composed of three manuscripts for publication. The first evaluated our participants’ skills in multiple linguistic domains with behavioural tasks typical of clinical and research settings. Data reveal impairments in the DLD group in both lexico-semantic and morphosyntactic domains but suggest that a SUBIRR number agreement task was best at discriminating DLD from controls. The second article presents a novel ERP experimental design using only grammatical sentences, presented simultaneously with semantically and grammatically congruent or incongruent images, to assess morphosyntactic and lexico-semantic sentence processing at the neurocognitive level. Data from twenty-eight French-speaking adults show that they elicited the expected ERP components found in previous studies using ungrammatical sentences. These data served as a reference to establish whether our participants with and without TL process sentences in a mature way. The third article tested this novel ERP experiment with our (pre-)teen participants. We tested predictions of the procedural deficit hypothesis which states that children with DLD should have impaired morphosyntax due to an underlying procedural memory deficit, and the generalized slowing hypothesis, which proposes that all linguistic domains should be impaired due to an underlying processing deficit. This experimental design was run on teens with and without DLD. Although some processing delays were found in the DLD group, results on most conditions better fit the procedural deficit hypothesis. This study suggests that, in contrast with morphosyntax, lexico-semantics is a relative strength in teenagers with DLD when processing linguistic information at the neurocognitive level. Overall, this thesis reveals that morphosyntax, tested through SUBIRR number agreement, is especially impaired in French-speaking teens with DLD when compared to their TL peers. We discuss the findings in relation to clinical practice and highlight the importance of examining neurocognitive processes in language assessment.