Following certain general remarks concerning the concept of 'theory' and possible steps in formulating a theory of language, in particular, about its genesis (such an empirical theory is claimed to be impossible, on logical grounds), the author deals with what he looks upon as a central part of any viable theory of language, viz. a theory of the verb say (in Polish 'powiedziec'), and formulates a set of preliminaries for such a theory. The main point of the preliminaries is the distinction 'say that -' vs. 'say:-' which is claimed to be a distinction of two separate units of language (basically, in agreement with Austin). Both these units are claimed to be definable (contrary to what has been claimed by Wierzbicka). The author confines himself to the former one. He formulates several properties to be assigned to it. In the Appendix (Aneks) the proposed definition / paralocution of 'say that' (powiedziec, ze) called (DIXIT) is presented and supplied with explanations and only partial commentaries (proofs of the properties of the verb, insofar as they are possible, are left out).