A model-based experimental method for tracking parameter drift in nonlinear dynamical systems is described. Local linear tracking models are constructed using data sampled over a fast time scale. These models are used to analyze data from systems with parameters which evolve over a slow time scale according to a "hidden" rate law. The method is applied to a numerical study of a nonlinear electrical circuit with a variable resistance as the drifting parameter. The mean-square tracking model prediction error is shown to follow successfully both ramped and sinusoidal parameter variations, suggesting that, at least in the cases studied, the method provides an invertible mapping between the parameter space and the observable space. Thus it should be possible to extract rate information about hidden drift, a requirement for true prediction.