Do distinct subpopulations signify modes of behaviour in a noisy single cell?

Population-level distributions of fluorescence or molecule counts are often taken to reflect the behaviours of individual cells within that population. In this conceptual article, I argue that counting subpopulations can be a misleading proxy for identifying the number of behavioural modes accessible to individual cells within a system. I show that definitions of behavioural modes based on deterministic modelling can fail when fluctuations in a system's state-or noise-become significant. In such cases, peaks in the probability distribution-emerging from stochastic descriptions-are often interpreted as substitutes for deterministically defined stable modes. However, I demonstrate that this interpretation can break down: it is possible to construct counterexamples in which two subpopulations arise from a system that supports only a single mode of behaviour, driven by non-equilibrium transient dynamics. Better understanding the role of noise in transient biological randomness may allow for the discovery of novel mechanisms of regulation that are not apparent in steady-state behaviours.