Coercion matters to almost everybody, and almost everybody thinks that it is often or always bad. Yet few agree on what counts as coercion. The theorists of coercion propose “analyses” of the concept of coercion, and are promptly refuted by rival theorists. In a recent essay with the implausible title “The Last Word on Coercive Offers …(?)” Daniel Lyons criticizes numerous complex and competing analyses of coercion proposed in recent philosophical discussions. (His topic isn't only “offers” in a narrow sense that contrasts with “threats,” but the whole gamut of proposals and demands that might be judged coercive.) He hopes to have the last word because, he claims, his own proposal of a few years earlier has stood up to trial by counterexample not perfectly but still rather well.

The debates that have led to this type of claim strike me as a reductio ad absurdum of certain procedures often used in ethical discussion. These procedures assume that “our” intuitions about possible examples and counterexamples can be treated as data, by which we may test and either refute or confirm proposed analyses of the necessary and sufficient conditions for the application of moral concepts. However, trial by counterexample cannot get going unless there is agreement on the classification of cases. It breaks down when examples and counterexamples cannot be reliably distinguished. In discussions of coercion there is no agreement about cases: Consider the long-running dispute between liberals and socialists about whether the wage bargain under capitalism coerces.