Relacoes semântico-cognitivas no uso da preposição "em" no português do Brasil

Abstract

The great interest in the semantics of prepositions by cognitive linguists has allowed for the emergence of a wealth of network models of polysemy based on prototype theory. Despite their acknowledged coherence, these models have given rise to criticism on both their diversity and, particularly, the presumed cognitive nature of these explanations (SANDRA & RICE, 1995 e CROFT, 1998), notably the degree of compatibility between the linguists view and that of thenon-informed native speaker. Inspired by the latter, in this dissertation we present a semantic description of the polysemy of the razilian-Portuguese preposition em, by using two empirical methods: introspective analysis of a 1.2-million-word corpus of journalistic texts and a psycholinguistic experiment involving untrained university undergraduates, all native speakers of Brazilian Portuguese. In order to explain the polysemy of this preposition, we adopted Ronald Langackers (1987) usage-based model of language and his concept of construal, Mark Johnsons (1987) proposal of embodied cognition, mainly, his ideas about image schemata, George Lakoff & Mark Johnsons (1980) theory of conceptual metaphor, Leonard Talmys (2000) closed-class semantics, and Claude Vandeloises (1991) study of functional relations as determinants of themeaning of spatial prepositions. The cognitive aspect of this theoretical approach was attained in two complementary stages: firstly, by following Langackers model of schematic network, we explored the cognitive processes involved in the interpretation of synchronic uses of thepreposition, and those responsible for the coherent links obtained between them. This model consists of a structure containing sanctioning and schema/instance relations between the uses.The analysis of 2813 samples from the corpus revealed two highly schematic senses for the preposition em, derived from the CONTAINER schema: location and specification. The first is elaborated in twenty-two categories and subcategories, representing 86.78% of the uses, eight ofthem belonging to space domain, three, to time domain, and the remaining eleven, to various other concrete and abstract domains. As a result of the functional effects of control and support, nine categories were found to instantiate the latter sense of specification, also in various concrete and abstract domains. Most of the variation obtained in spatial contexts could be explained by construal effects and the Container/contained functional relation. Other uses were interpreted asemerging from metaphorical and metonymic processes. In the second stage of the research, thirty-two subjects were involved in a sorting task, in which they should group forty-eight sentences representing categories obtained in the previous analysis, by using a similarity criterion. This task was followed by individual interviews, in which subjects explained thestrategies they used to complete the task. Statistical methods of cluster hierarchy and Tocher multidimensional analysis revealed a weak isomorphism between the network proposed by the researcher and the informants evaluation, i.e., they perceived its relational structure anddistinguished some of the categories predicted from the corpus analysis, producing seven sets of uses. The relational structure of the network could be seen, for instance, in the association of locative and non-locative uses in a large group of metaphorical uses, of spatial and non-spatialuses in a schematic location category, and of various types of specification (shape, color, material, abstract specification) in a single larger set. Besides, subjects formed large and small groups with differing levels of internal cohesion, which corroborates the typical granularity of these networks, even if obtained on a smaller scale, and the coherent relations between the semantic categories.