The problem of the meaning of a reasonable natural language expression is solved. First, traditional ”denotational” approach is criticized. The meaning of a sentence is not its truth value, similarly the meaning of, eg, ”The president of U.S.A.” is not Bill Clinton, etc. Frege met this problem when analyzing the so called propositional attitudes in which ”denotational” approach has lead to the paradox of analysis. His well-known solution consists in splitting the meaning into sense and reference. But this is rejected in the paper as well, for its radical contextualism. In the first attempt, meaning is defined as an intension — mapping from possible worlds and time points — (empirical expressions) or extension (analytical expressions), respectively. The problem of the propositional attitudes is seemingly solved. The proposition that Morning Star = Evening Star is diffe_rent than that of Morning Star = Morning Star. But, alas, in the case of analytical expressions we get the paradox of omniscience. A fine-grained solution is, therefore, proposed: the meaning is a closed construction denoted by the respective expression. Yet this solution is still not precise enough. Eventually the meaning is a concept which is the equivalence class of quasi-identical constructions indiscernible from the conceptual point of view. Finally, homonymous, synonymous and (analytically and empirically) equi_valent expressions are precisely defined. Concluding we state that only synonymous expressions, having exactly the same meaning, ie. representing one and the same concept, can be mutually substituted in propositional attitudes without lea_ding to paradoxes.