Differential mortality in Norway has been studied on the basis of a sa
mple of data derived from the linkage of the 1960, 1970, and 1980 cens
uses to vital registration records. Based on the hypothesis that the d
eterminants of survival act in interaction, two models are proposed. T
he first is based on states defined at each observation period by the
conjunction of attributes characterizing each individual. The second m
odel considers the chronological order of the states. Logistic regress
ion applied to the latter shows that the most favourable male and fema
le life trajectories are those for married people belonging to rather
privileged categories. There are however some differences by sex, as f
avourable trajectories concern both economically inactive females and
employees.