Uncertainty arising from our relationship with the real world can often lea
d to imprecise probabilities, whose bounds are captured by Random Set Theor
y. Reasons leading to this kind of uncertainty in rock engineering are inve
stigated, as well as the limitations of the probabilistic approach. Procedu
res are given to handle the usual information gained either in the field or
from expert's opinions, taking into account not only dissonance, but also
non-specificity. Examples of applications are presented with special regard
to event tree analysis, rock mass classifications, and reliability-based d
esign of tunnels by means of the empirical recommendations of rock mass cla
ssifications. (C) 2000 Elsevier Science Ltd. All rights reserved.