Changes of belief represent a central aspect in a person's life and mental being All of a human's senses, such as vision, touch, hearing, tasting, etc., trigger menta processes of learning which may involve accepting new knowledge, discarding one, strengthening or weakening the degrees of some beliefs, or correcting othe. beliefs. The knowledge (most recently) acquired may even contradict the set of beliefs which would then need to be handled appropriately so that the contradiction is removed. The modelling of such processes of changes of beliefs has been a major concerr in various areas such as databases and artiicial intelligence. In this thesis, a logic programming approach is adopted. The main problem tackled is how to update knowledge bases represented as logic programs. More speciically, given a logic program P and a formula W, the problem at stake is to design an insertion (resp., deletion) procedure so that W becomes (resp., is no longer) a logical consequence of the completion of P (while satisfying some integrity constraint theory). Procedures to update a normal logic program P with an atom A are irst presented. The insertion procedure constructs nite tree for P Uand completes some derivation of this tree into a refutation by asserting relevant facts. Similarly, the deletion procedure proceeds by constructing a inite tree, but prunes all non-failed derivations by appropriate retractions of clauses and assertions of facts. Properties of these procedures are studied, especially in the presence of negation. These procedures are then generalised so as to update arbitrary logic programs with (arbitrary) formulas. To this end, the procedures are first made mutually recursive. These are then used by the generalised update procedures. In the latter case, a program is transformed into normal form; this normal form is updated; and the selected transactions on the normal form are then transformed back into transactions on the arbitrary program using the compaction algorithm. Theorems