An original mathematically sound methodology for modeling and analyzing natural language reasoning is presented in an accessible form. The possibility of combining in a logical model not only methods of logical inference, but also methods of checking the compatibility of initial assumptions, generating hypotheses and obtaining inductive conclusions has been established.