The tunneling-time problem is shown to be analogous to the interpretat
ion of the two-slit interference experiment. A measurement assuming an
arbitrarily small interaction between a particle and a clock is shown
to contradict the uncertainty principle and leads to complex times. A
real non-negative traversal time is obtained in a measurement which s
elects Feynman paths that spend in the barrier a known amount of time;
this, however, strongly perturbs the tunneling. The Larmor clock demo
nstrates both types of behavior.