Categorization and vagueness: how are words stored in the mental lexicon?/Kategorizacija i neodredenost: kako su pojmovi pohranjeni u umnom rjecniku?

In this study event-related potentials method (ERP) was used to study the effect of vagueness in language. In difference to philosophy, where vagueness is a source of Sorites paradoxes, in linguistics vagueness is regarded as a design feature of human language. In this sense vagueness is modeled in contemporary cognitive science as a feature that allows for keeping the actual amount of exchanged information between the intelligent agents at minimum while preserving the effectiveness of communication. The theory of meaning that accounts for vagueness in language is Gardenfors' theory of conceptual spaces. According to this theory the concepts are positioned in a continuum along a vector that represents a feature, e. g. tallness. Gardenfors introduces a crisp function that corresponds to categorization. The crisp function divides the conceptual space into two parts (tall and non-tall regions in the given example). Categorization is therefore essential for the account of vagueness because it is the source of vague concepts (e.g. of being tall in a continuum of precise measures of tallness, say, in millimeters). In this experiment we used a well-known picture matching paradigm in which participants had to decide whether a word matches the given photograph or not (categorization task). The vagueness condition was added by adding hyponyms as words to the categorization task, thus creating four experimental conditions (non-vague match, vague match, non-vague mismatch and vague mismatch). The difference in N400 component of ERP was expected for all vague and mismatch conditions. However, two different components were obtained: both N400 for all mismatch conditions (including vague) and P600 for vague conditions. Therefore, both the experimental effect of vagueness was obtained and the results can be interpreted in terms of vagueness as facilitation of language communication. These results also indicate that retrieving words from mental lexicon is (at least) a two-phase process in which the word activated by the picture is (1.) compared with the stimulus word and then (2.) integrated into the context. The first phase is highly automatic and beyond speaker's control while the second depends on the context: vague words are correctly categorized, but require additional processing costs if there is no other reason for using them. Key words: vagueness (psycholinguistics), cognitive model of human communication, mental lexicon Za razliku od filozofije, u kojoj se neodred