In the paper we present a contribution to the SyntLex longterm-project aiming at a lexicon-grammar for Polish. A corpus-based method is presented for computer-assisted improvement or/and verification of verbo-nominal lexicongrammars (in application to Polish). Feasibility study.