Vann McGee publis ed a paper where he claimed that modus ponens is not a generally valid rule of inference for the natural language conditionals. Although he spoke only about the indicative conditionals, the counterexample of his that is considered in my thesis can be easily applied to subjunctive (counterfactual) conditionals also. After explaining what is paradoxical about the counterexample, and commenting upon McGee's solution, I consider various remarks to McGee's point that might be or already are made by some philosophers. I try to show that none of them helps us to remove the problem. I argue that in fact the 'counterexample' points mainly to the need of distinguishing the truth-functional and intensional disjunctions.