In anomalous transport, a particle's mean displacement under a constant force may grow too slowly to be distinguishable from noise. The mean drift is there, but measuring it requires impractical observation times. The standard diagnostic for detecting applied force — measure the mean displacement — fails when the transport is sufficiently anomalous.
The paper shifts the diagnostic from mean to variance. In strongly heterogeneous media with subdiffusive transport (anomalous exponent α < 2), the displacement variance initially follows the unbiased scaling law, indistinguishable from zero force. But at a crossover time that depends on the force strength and the anomalous exponent, the variance transitions to a force-dominated regime with different scaling.
Note (2026-11-14; read against the paper's full text): the essay says the mean displacement fails as a diagnostic of a weak force in strongly anomalous transport. The paper treats the variance as the key signature rather than the mean drift, but it also reports that the mean has its own observation-time-induced crossover (from linear to nonlinear response, with a threshold force that scales as the observation time to the power minus alpha over one plus alpha), and that the variance in the driven continuous-time random walk shows the same time dependence while the quenched trap model has a different power. It says quenched disorder lowers the detection threshold compared with the continuous-time random walk and that the crossover reflects a competition between the finite observation time and the intrinsic relaxation time of the driven system; I did not find the essay's statement that spatial heterogeneity amplifies the force's effect on the variance. The models are a driven continuous-time random walk and a driven quenched trap model, studied analytically and by numerical simulation.
The crossover is observation-time-induced. At any fixed measurement window, there exists a detection threshold: forces weaker than this threshold are invisible. The threshold decreases with longer observation — weaker forces become detectable if you watch long enough. In quenched disorder systems, the threshold is even lower than in simpler continuous-time random walk models, because spatial heterogeneity amplifies the force's effect on the variance.
The structural insight: the force is not hidden — it's present from the start. But the observable that reveals it changes identity over time. At short times, the variance carries no force information. At long times, it carries all of it. The measurement doesn't change; the information content of the measurement does. The same number — displacement variance at time t — transitions from uninformative to diagnostic, and the transition time itself encodes the force strength.