Lead isotope data from geologically mixed suites of Archaean gneisses
often yield secondary isochrons that have statistically better fits th
an might be expected from the geological constraints on the range of s
amples used. Accepted as true isochrons, these Pb-Pb lines are commonl
y used both to determine the age of the rock suite and as evidence tha
t all the rocks that yield data points within error of the isochron we
re formed at the same time from the same source. Using within geologic
ally reasonable limits random data for initial lead and secondary lead
growth, we demonstrate that heterogeneous sample sets may yield stati
stically valid secondary isochrons. This implies that the fit to a com
mon secondary isochron does not necessarily mean that the samples used
have the same age or an isotopically similar source. Numerical modell
ing of artificial data sets suggests that high correlation between mu(
1) and mu(2) values is the single most important factor that enhances
the statistical fit of the samples about a lead line and smoothes the
initial scatter. In contrast, initial scatter will be preserved in dat
a sets with a random relation between mu(1) and mu(2) values. The mu(1
) value estimated from the lead line will be biased towards the sample
with the lowest mu(2) value.