The "random frequency modulation" model (introduced, for the sake of explaining exchange narrowing of spectral lines, by Anderson and consolidated by Kubo) is generalized with an eye toward providing an analytical basis for some recent computer simulations by Grunwald and Steel on environmental isomerism. The principal problem is to consider a multistate model and calculate the power spectrum of the signal : mu(t) = mu(0) exp[iota integral(0)(t)v(t’)dt’] = mu(0) exp[iota x(t)] (-infinity less than or equal to v less than or equal to +infinity), where mu(0) = 1 and v is a random function of time. The generalized model is specified by probability distributions, g(i)(v), and a transition matrix, K = (k(ij)), which are defined as follows : When the system is in state i, 1 less than or equal to i less than or equal to s, it generates a signal mu(t) with a constant angular frequency, so that x = vt; the probability that the system jumps to state j during a time interval Delta t is k(ij)Delta t + o(Delta t); after the jump, it continues to generate the signal but with frequency v’, which is a random variable with density g(j)(v) dv; the possibility that transitions can occur to the same state is admitted by taking k(ii) > 0. The situations examined by Anderson and Kubo correspond to two special cases : (a) a single-state model, i.e. s = 1, and (b) a multistate model with a "line spectrum", i.e. g(i)(v) = delta(v - v(i)). Divergence from previous treatments, which deal directly with mu (which lies on the unit circle), stems from the recognition that since x lies between +/-infinity, the task of determining f(x,v;t), the time-dependent probability density in the two-dimensional (x,v)-space, is completely equivalent to finding the probability density in the phase space of a point particle which is constrained to move (along a straight line) with a velocity that suffers random changes, Computation of the power spectrum may thus be viewed as a problem in linear transport theory; indeed, the integrodifferential equation which governs the evolution off(x,v;t) for the one-state model (case a) turns out to be formally identical with the single relaxation time approximation to the linear Boltzmann equation. The new approach, which recovers previously published results after minimal manipulation of the underlying transport equation, is brought to bear on the issue, raised by Grunwald and Steel, of the distinguishability of environmental isomers; the conclusions reached by them are corroborated and extended.