Validade empírica das redes de polissemia para o significado preposicional

This paper evaluates the empirical validity of a polysemy network proposed for the preposition em in Brazilian Portuguese, by comparing the linguist's and the non-specialist language user's views of categorization. The network was based on the Schematic Network cognitive model of R. Langacker (2008, 1987), with scheme/instance and semantic extension relationships, and involved 48 sentences taken from the internet, which represented 24 usage patterns of em as defined by the linguist. Thirty-two undergraduate students with no training in linguistics, all native speakers of Brazilian Portuguese, took part as subjects in a psycholinguistic experiment in which they sorted these sentences according to similarities perceived in the meaning of em. After completing this task, the participants reported their categorization strategies in individual interviews. A dissimilarity matrix based on the participants' classification was subjected to cluster analyses by the Ward (1963) and Tocher (cited in RAO, 1952) methods. The latter revealed seven categories created by the participants, four of which exhibited stronger semantic motivation than the other three. Participants differed in the number of groups they formed, and each participant created groups of different sizes. As well as coherent connections among categories, the results showed an important level of isomorphism between the originally proposed network and the participants' evaluation, corroborating the relational structure of the network and a significant part of its granularity. Overall, the study also confirmed the greater predominance of space over the other domains and of time among non-spatial domains.