We develop a theory of contracting systems, where behavioural contracts may be violated by dishonest participants after they have been agreed upon — unlike in traditional approaches based on behavioural types. We consider the contracts of [10], and we embed them in a calculus that allows distributed participants to advertise contracts, reach agreements, query the fulfilment of contracts, and realise them (or choose not to). Our contract theory makes explicit who is culpable at each step of a computation. A participant is honest in a given context S when she is not culpable in each possible interaction with S. Our main result is a sufficient criterion for classifying a participant as honest in all possible contexts.