We introduce a new method, allowing one to describe slowly
time-dependent Langevin equations through the behaviour of
individual paths. This approach yields considerably more
information than the computation of the probability density. In
particular, scaling laws can be obtained easily. The main idea
is to show that for sufficiently small noise intensity and slow
time dependence, the vast majority of paths remain in small
space-time sets, typically in the neighbourhood of potential
wells. The size of these sets often has a power-law dependence
on the small parameters, with universal exponents. The overall
probability of exceptional paths is exponentially small, with an
exponent also showing power-law behaviour. The results cover
time spans up to the maximal Kramers time of the system. We
apply our method to three phenomena characteristic for bistable
systems: stochastic resonance, dynamical hysteresis and
bifurcation delay, where it yields precise bounds on transition
probabilities, and the distribution of hysteresis areas and
first-exit times. We also discuss the effect of coloured noise.