This paper gives an overview of linking methods of personal data compu
ted in different files as registers and administrative or medical reco
rds. Epidemiological and demographic examples are studied, then the su
ccessive steps of record linkage are described : definition and choice
of identifiants, searching and comparing pairs of records to determin
e whether they should be linked. A special mention is made to probabil
ity matching and methods optimizing the matching procedures. The discu
ssion envisages the advantages of record linkage : better use of healt
h statistics, new types of epidemiologic studies. The ethical problems
and the need for evaluation are also discussed.