ENIn the article, the existing methods of morphological annotation for historical corpora are analysed. A new method of an unsupervised dictionary-free morphological tagging is proposed which is based on applying syntactical dependency constraints to a set of possible morphological interpretations of word finals. The procedure starts with a draft set of orthographical, morphological and syntactic rules that are adjusted and refined as the analysed text is processed. The method is specifically tailored to the highly-inflectionate languages of ‘classical’ Indo-European type. The ambiguity of annotation is further reduced by applying a set of language-neutral constraints, such as the well-known principle of projectivity or the minimization of possible word stems. The application of the method to tagging M. Daukša's Cathechism of 1595 is described.