In this paper we consider two points of views to the problem of coherent integration of distributed data. First we give a pure model-theoretic analysis of the possible ways to `repair' a database. We do so by characterizing the possibilities to `recover' consistent data from an inconsistent database in terms of those models of the database that exhibit as minimal inconsistent information as reasonably possible. Then we introduce an abductive application to restore the consistency of a given database. This application is based on an abductive solver (A-system) that implements an SLDNFA-resolution procedure, and computes a list of data-facts that should be inserted to the database or retracted from it in order to keep the database consistent. The two approaches for coherent data integration are related by soundness and completeness results.