The paper describes work on a linguistic database of grammatical sentences of the Croatian language. In databases, sentences are shown as constituent structure trees, and words are shown with features and their values. The main project which is concerned with language databases is the Penn Treebank project of the Linguistic Data Consortium (A. Bias et al., 1995). The database is a result of theoretical research in the field of computational linguistics and its application in the Croatian language. The database of grammatical sentences of the Croatian language can also be accessed through the Internet.