Made available in DSpace on 2015-04-14T13:38:45Z (GMT). No. of bitstreams: 1 445010.pdf: 2062999 bytes, checksum: d6e1e040ed45bb8943d38936857e2ae1 (MD5) Previous issue date: 2012-12-20 / This study brings together theoretical assumptions from Corpus Linguistics and metaphor theories and aims at verifying variation in metaphor production by Brazilian English learners following a corpus-based approach. The corpus used for the investigation was the Brazilian English Leaner Corpus (BELC) (PACHECO, 2010), which is composed of four proficiency levels: (i) Beginner, (ii) Pre-Intermediate, (iii) Intermediate, and (iv) Advanced; and three types of task: (i) task 1 (descriptive text with personal information in 1st person), (ii) task 2 (descriptive text with personal information in 3rd person), and (iii) task 3 (narrative text about a trip). The corpus was manually annotated based on the procedures established by Cameron (2003) and by the Pragglejaz Group (2007). The metaphor frequency was later calculated using the software WordSmith tools (SCOTT, 2012), more specifically, the Concord tool. The research was organized in the following stages: (i) manual annotation of metaphor occurrences; (ii) extraction of metaphor frequency in the whole corpus and in its subcorpora; (iii) comparison of frequencies of use of linguistic metaphors in the three proficiency levels subcorpora; (iv) comparison of frequencies of use of linguistic metaphors in the four textual types subcorpora; (v) comparison of frequencies of use of linguistic metaphors in the twelve individual subcorpora. The significance levels of the frequency variations were calculated with the statistical test Log Likelihood. The results show that metaphor production increases at each proficiency level and varies according to the different text types: texts containing personal information in 3rd person tend to present higher frequency of metaphorical items when compared to personal narratives in 1st person, what corroborates findings from previous studies (BERBER SARDINHA, 2012). / Este trabalho re?ne referenciais te?ricos da Lingu?stica de Corpus e de correntes te?ricas da met?fora e tem o objetivo de verificar a varia??o na produ??o de met?foras por aprendizes brasileiros de ingl?s, atrav?s de uma abordagem baseada em corpus. O corpus utilizado na investiga??o foi o Brazilian English Leaner Corpus (BELC) (PACHECO, 2010), o qual ? composto por quatro n?veis de profici?ncia: (i) Beginner, (ii) Pre-Intermediate, (iii) Intermediate e (iv) Advanced; e tr?s tarefas: (i) tarefa 1 texto descritivo com informa??es pessoais em 1? pessoa, (ii) tarefa 2 texto descritivo com informa??es pessoais em 3? pessoa e (iii) tarefa 3 texto narrativo sobre uma viagem. O corpus foi anotado manualmente, com base nos procedimentos de Cameron (2003) e do Grupo Pragglejaz (2007). A frequ?ncia de met?foras foi extra?da com a ferramenta Concord do WordSmith Tools (SCOTT, 2012). A pesquisa foi organizada nas seguintes fases: (i) anota??o manual de met?foras no corpus; (ii) extra??o da frequ?ncia de met?foras no BELC e em seus subcorpora; (iii) compara??o das frequ?ncias de uso de met?foras lingu?sticas nos quatro subcorpora de n?veis de profici?ncia; (iv) compara??o das frequ?ncias de uso de met?foras lingu?sticas nos tr?s subcorpora de tipos textuais; (v) compara??o das frequ?ncias de uso de met?foras lingu?sticas nos doze subcorpora individuais. Os n?veis de signific?ncia das varia??es de frequ?ncia foram calculados com o teste estat?stico Log Likelihood. Os resultados mostram que a produ??o de met?foras aumenta em cada n?vel de profici?ncia e varia de acordo com tipos textuais diferentes: textos com informa??es pessoais em 3? pessoa tendem a apresentar frequ?ncia mais alta de itens metaf?ricos em compara??o com narrativas pessoais em 1? pessoa, o que corrobora resultados de estudos anteriores (BERBER SARDINHA, 2012).