The population registers of the Isle de Coudres in Quebec have been us
ed to reconstitute families and their ancestors, and to classify them
by the parameters needed to produce a distribution of infant deaths an
d to compute coefficients of kinship and consanguinity. The quality of
the information obtained is shown by computing the variations between
different genealogies, and by studying religious dispensations. Group
s consisting, respectively, of related and unrelated individuals were
constructed and matched, taking into account the different risks assoc
iated with infant mortality. Globally, no significant difference betwe
en the mortality of the groups during the first year of life was obser
ved. The apparent similarity, however, conceals basic differences betw
een the 'related' group and the 'unrelated' control group. In the 'rel
ated' group, infant mortality rates were significantly higher for fema
les than for males, whereas the opposite holds for the 'unrelated' gro
up. In the 'related' group, the proportion of children who died during
their first week of life was higher (although the difference was not
statistically significant), and first children faced greater risks of
dying during their first year of life. The masculinity ratio at birth
was significantly lower than in the control group, or in the populatio
n as a whole.