The transient behavior of a quadratic model system perturbed by a multiplicative white noise has been investigated. The relaxation time of the system, as a function of the noise intensity D, has been determined by analog experiment and by digital simulation. The results obtained are mutually consistent, but contradict a recent theoretical prediction by H. K. Leung [Phys. Rev. A 37, 1341 (1988)] that there should be a critical slowing down of the system near the value of D for which a noise-induced transition occurs in the probability distribution. The discrepancy is resolved by deriving a new analytic result for the relaxation time, applicable to a range of systems described by separable stochastic differential equations.