The project deals with Czech multiword expressions with light verbs which pose a serious challenge for both theoretical and computational linguistics. On the basis of the theoretical study of selected syntactic, semantic and morphological aspects of the given expressions, a formal model of their lexicographic representation will be proposed. This representation will be theoretically adequate and at the same time economical. Moreover, in designing such a model, the strong emphasis is put on systematicity and explicitness. This model is primarily designed for the on-line Valency Lexicon of Czech Verbs (VALLEX). The proposed representation will be applied in the description of a large amount of data provided by the lexicon. The lexicon enriched with the information on multiword expressions with light verbs will form a solid basis for both further theoretical research and natural language processing.