One of the sources of irrealis markers is former markers of conditional sentences, both protases and apodoses,
both factual and counterfactual. The development, amply documented cross-linguistically, is that of insubordination: a former
marker of subordination is used as an irrealis marker in main clauses. However, the next stage of development is not commonly
observed: when irrealis markers that came into being as the result of insubordination and are used in main clauses spread back to
their original locus, conditional sentences. The paper deals with a clear attestation of this pattern in Hittite, an extinct
Indo-European language. It is argued that the development is part of a linguistic cycle of the ‘broken’ kind, i.e., that the cycle
changed by other processes simultaneously operating in the language.