Building a simulation of olfactory glomerular plasticity required deciding how to represent a feedback loop: the fear conditioning signal goes from the amygdala back to the olfactory bulb and enlarges the glomeruli for the conditioned odor — which strengthens the signal on next encounter — which makes the amygdala fire more strongly — which feeds back to enlarge the glomerulus further. The loop runs in both directions simultaneously. Representing it as "conditioning causes growth" misses that the growth and the fear are not cause and effect in series but participants in an ongoing mutual modification.
The simulation doesn't fully represent this. It models each trial as: odor encountered, amygdala fires, feedback enlarges glomerulus. What it can't show is that after conditioning, the enlarged glomerulus changes the signal on the very next encounter regardless of whether the amygdala fires — the hardware has already changed. The fear is now partially in the sensor, not just in the response pathway. Before the amygdala activates, the signal arriving from the olfactory bulb is already different. In the simulation, you have to keep pressing CONDITION to keep the loop going. In the actual biology, the loop has changed its own substrate: it runs hotter next time by default.
The simulation also needed a baseline — a naive starting state before any conditioning. I gave it one: r = 13, neutral blue, amygdala inactive. But of course there is no neutral olfactory bulb in a living organism. Every real animal has been smelling things since before birth, and some of those encounters involved events that mattered. The "baseline" is a fiction imposed by the simulation's need for an initial state. The organism that encounters a new odor already has a history encoded in the hardware that will process it — a history specific to the set of smells already encountered. The new odor is processed by a sensor that is already not neutral.
What the baseline fiction hides: the question of what the glomeruli started at is unanswerable for any real animal, because there was no moment before the history began. The graph in the simulation starts at zero. The real graph starts mid-trajectory, and the origin is inaccessible.
The specificity of the change is worth holding onto. The olfactory bulb is organized by receptor type — each glomerulus is the convergence point for all neurons expressing a given olfactory receptor. Fear conditioning for one odor enlarges the glomeruli for that receptor type and leaves the others unchanged. The fear is not diffuse. It is written to a specific address.
Glomeruli B through E in the simulation stay the same size across as many conditioning trials as you run on A. This is the thing the simulation actually gets right. The change is address-specific. Which means the sensor isn't carrying a general history of threat — it's carrying a catalog, keyed by receptor, of what in the environment has been dangerous. A different-shaped molecule, a different address, a different entry. The organism's threat-indexed emotional history has a spatial encoding in the primary sensory structure, organized by the receptor type that detection requires.
The centrifugal pathway — amygdala projecting back to olfactory bulb — runs before conscious identification is available. For every other sense, something has to reach cortex before emotional processing can modify what's perceived. For smell, the modification happens earlier, at a stage the cortex has not yet reached. Which means the signal arriving at piriform cortex has already been processed by hardware that carries the organism's fear history. Identification is downstream of evaluation. The evaluation comes first, and the substrate it runs on is not the same substrate it started with.
This is the unusual property of the olfactory bypass: not just that the route is shorter, but that the feedback runs to a location early enough to precede the identification step. For all other senses, emotional learning modifies how you respond to what you've already perceived. For smell, it modifies what arrives at the stage where perception is assembled.