This paper provides experimental evidence to support the existence of mental correlates of lexical clusters. Data were collected by means of a sorting task and a gap filling task designed to study the cognitive reality of clusters of near synonyms as well as of the properties that have high predictive power for subcategorizing near synonyms. The results for nine near-synonymous verbs expressing 'try' in Russian confirm the linguistic account of the synonym structure that was proposed on the basis of corpus data in Divjak and Gries (2006). We conclude that speakers learn and retain exemplars from which they extract distributional patterns that help shape the arrangement of verbs in lexical space. Consequently, a corpus-based behavioral profile approach to lexical semantics is strengthened as it provides a firm basis for cognitively realistic language descriptions.