Yao, Jones, Larkin, and Mugler discover that different growth limitations produce qualitatively distinguishable expansion patterns in microbial colonies. Growth arrest still gives accelerating expansion. Mechanical pressure from neighboring cells gives constant speed. But nutrient limitation triggers a phase transition: depending on supply and biomass conversion, spreading can either accelerate or decelerate.
Note (2026-11-10): I checked the abstract of the Yao, Jones, Larkin and Mugler paper. It is a theoretical study with a continuum active fluid model and simulations, not an observation of colonies. It finds accelerating spreading for memoryless growth arrest (at a reduced rate), constant-speed spreading for pressure-limited growth, and a phase transition for nutrient-limited growth, and concludes that an observed slowdown is likely, not certainly, due to nutrient depletion. The essay's rules of the form 'if it is slowing, the cause is nutrient depletion', the attribution of acceleration to collective rearrangement, and the practical claim for fermentation are not in the abstract.
The macroscopic behavior encodes the microscopic mechanism.
If a colony is slowing down, the cause is nutrient depletion. If it's expanding at constant speed, the cause is mechanical crowding. If it's accelerating despite individual cells ceasing to divide, the cause is collective rearrangement. You don't need to measure individual cells. You just need to watch the edge.
This inverts the standard diagnostic strategy. In microbiology, you typically measure cell-level properties (growth rate, nutrient uptake, motility) and try to predict population-level behavior. This paper shows the reverse: population-level dynamics are sufficient to infer cell-level constraints. The colony is its own diagnostic.
For fermentation, this is immediately practical. The spreading behavior of a culture tells you what's limiting it without requiring intracellular assays. But the deeper point is structural: the mapping from microscopic mechanism to macroscopic pattern is not many-to-one (many mechanisms producing the same expansion). It's distinguishable. Each constraint has its own signature. The ambiguity that everyone assumed was there — isn't.