The strong motion data from the Izmit, Turkey and Chi-Chi, Taiwan earthquak
es have pointed out uncertainties in current strong motion attenuation curv
es for large earthquakes. Although the near-source strong motion data from
these two well-recorded large earthquakes were well below the estimated val
ues by the ground motion models, they were consistent with constraints esti
mated from precarious rock methodology. This discrepancy could be a result
of inadequate ground motion data for large earthquakes and possible flaws w
ith a number of statistical parameter assumptions that were necessary for e
xtrapolation from existing database, which is dominated by small earthquake
data. This review article discusses several important issues that have the
potential to cause major impact on seismic hazard analysis. They are: (i)
partitioning of uncertainties into aleatory and epistemic contributions, (i
i) quantification of precarious rock observations and use of the data to co
nstrain and improve ground motion models, (iii) continuing to deploy strong
motion instruments near major faults since only more strong motion data wi
ll definitively resolve the issues of what is normal behaviour, and (iv) un
derstanding through modelling and observations, the physical phenomena that
affect strong motion, including the effect of total fault offset, surface
rupture and type of faulting.