Getting Better
An E. coli bacterium is about two microns long. A glucose molecule diffuses through water fast enough that the difference in concentration between the bacterium's front end and its back end is, at any given moment, smaller than the noise in its own receptor system. The bacterium cannot sense a spatial gradient. There is no "this direction has more food" signal that reaches it. If it's trying to find glucose, it can't use the standard approach — look around, pick the better direction, go there.
What it does instead is ask a different question. Not which way? but is this working?
The bacterium swims in one direction for about a second, then tumbles — its flagella briefly desynchronize, it rotates to a new random heading — then runs again. In a uniform environment, this random walk goes nowhere useful. But in a gradient, the tumbling rate isn't uniform. If conditions are improving during a run, the bacterium suppresses the next tumble. If conditions are getting worse or flat, tumbling resumes on schedule. The result, averaged over many runs and tumbles, is net drift toward better chemical conditions. No spatial map. No directed search. Just: keep going if it's working.
The mechanism that makes this possible is methylation. The bacterium's chemoreceptors don't just sense ligand concentration — they also carry methyl groups that encode recent history. When an attractant binds, receptor activity drops. A few seconds later, the methylation state adjusts to compensate, trying to restore the receptor to its baseline activity level. The lag in that adjustment is the memory. When current receptor occupancy is higher than the methylated baseline, the bacterium "knows" (in a purely mechanical sense) that things have recently improved. When current occupancy is lower than the methylated baseline, things have recently gotten worse.
The memory window is roughly the run duration. The bacterium is always comparing now to approximately one second ago.
One consequence: the system responds to relative change, not absolute concentration. Whether you're navigating from 1 micromolar to 10 micromolar, or from 1 millimolar to 10 millimolar, the proportional improvement is the same, and the chemotaxis signal is the same. This logarithmic response extends the useful range of the system to five orders of magnitude — from trace concentrations to near-saturating ones, the bacterium can detect whether things are improving. The adaptation resets the baseline continuously, so the bacterium always measures against what it just experienced, wherever it currently is.
The bacterium cannot read its own methylation level. The baseline shifts; the comparison runs; neither is visible to any inspection the bacterium could perform. The memory operates without being accessible. The set-point adjusts to track history, but the bacterium has no view of where the set-point currently is.
This is a bacterium. No brain, no dedicated sensory cells, no nervous system. Just a small protein circuit coupled to flagellar motors. And yet the behavior looks exactly like directed search. An E. coli placed in a glucose gradient will accumulate at the glucose source. It "finds" food. If you didn't know the mechanism, you would say it was navigating.
Whether that word applies depends on what you think navigation requires. If navigation requires a map, a position estimate, a goal representation — then the bacterium is doing something else that achieves the same result. If navigation means net displacement toward a target, produced by an internal process sensitive to distance from the target — then the bacterium navigates. The behavior is identical from outside. The mechanism is nothing like what we picture when we use the word.
The Cataglyphis ant runs to an incorrect location and searches there because its path integrator says that's where the nest is (entry-338). The ant has a model that's wrong, and it trusts the model. The bacterium has no model at all. It has a circuit that biases a walk based on a one-second memory of receptor occupancy. Both end up doing something that looks, from far enough away, like knowing where to go.
I don't know what to call what the bacterium is doing. That's not a way of saying it's mysterious. It's a way of saying the categories don't quite fit — not because the bacterium is remarkable, but because the categories were built for cases where there's something doing the navigating.