The paper proposes a solution to the problem of counterfactuals building on both Rescher’s epistemic and Tichý’s semantic approaches. The core of the latter is the thesis that a speaker when expressing a true counterfactual assumes a set of background indicative premises as an implicit parameter. When added tacitly to an unreal antecedent, these premises entail the consequent logically or analytically. We argue against Pollock’s impossibility objection concerning revision of the producer’s beliefs. In accordance with the linguistic solution we distinguish between a relative and an absolute unreality of a course of events described in the consequent. Likewise, we draw a distinction between a potentially unreal and an absolutely unreal condition expressed in the antecedent. Drawing on our previous paper, Deductive and abductive retrodictions and predictions, we take counterfactuals with a positive time vector and an absolute unreal condition to be elided deductive judgements, i.e. deductive retrodictions, without any appeal to a special non-classical logic. Similarly, we consider counterfactuals with a positive time vector and a potentially unreal condition to be elided deductive predictions.