WHAT WORD-GUESSING REVEALS ABOUT YOUR BRAIN: PATTERNS OF LEXICAL STORAGE AND PROCESSING

Abstract

This study explores the organization of the mental lexicon by analyzing lexical retrieval strategies used in the word-guessing game Alias, aiming to test competing theories of semantic representation: attributional (feature-based) and distributional (co-occurrence-based). We used a naturalistic approach (see Allen et al., 2024) with the Russian version of Alias, recording speaker-listener pairs collaboratively guessing words under time pressure. The stimuli comprised 96 words, balanced across grammatical categories (32 nouns, 32 verbs, 32 adjectives), concreteness (48 concrete, e.g., jacket, 48 abstract, e.g., inflation), and frequency. We recorded 180 explanations annotated for retrieval strategies: definitions (73 cases), synonyms (16), antonyms (8), hyponyms (7), co-hyponyms (3), and hypernyms (1) which align with attributional theories, and real-life examples (49) and collocations (23) which align with distributional theories. Overall, guessing success in Alias depended on both word properties (concreteness, frequency, part of speech) and retrieval strategies. Co-hyponymy, antonymy, and collocations were most effective, while synonymy was least reliable, especially for adjectives. High-frequency and concrete words were guessed more easily, but even low-frequency words could be retrieved if described using distinctive features or shared knowledge, supporting a hybrid model of the mental lexicon that combines attributional and distributional elements. Findings support hybrid models of semantic representation, with both attributional and distributional strategies shaping lexical production and comprehension. This study introduces a naturalistic method, supports hybrid semantic models, and shows how word-level traits influence retrieval. Limitations include sample size and language scope.