On Chemistry

The worst week of a marathon training block isn’t the 20-miler. It’s taper week — the two weeks before the race where the volume drops on purpose, the legs feel fresh, and I feel worse than I did grinding through peak mileage. More irritable, more anxious, weirdly flat. For a long time I filed that under personality: I don’t do well with less to do. It isn’t personality. It’s a hormone with fewer places left to discharge, and it has a name and a half-life. That’s the thesis of this piece: most of what reads as “just how I am” is a small number of chemical systems doing exactly what they evolved to do, in a body that is currently running a founder’s workload, a coaching practice, and marathon training on top of it. Knowing the mechanism doesn’t excuse the behavior. It does tell you which lever to pull.

Dopamine has a reputation as the “pleasure chemical.” It isn’t. Wolfram Schultz’s recordings from single dopamine neurons found something more specific and more useful: the cells fire on reward prediction error— the gap between what was expected and what actually happened, not the reward itself^[1]. A fully predicted reward that arrives on schedule gets no response at all. An unexpected one gets a burst. A predicted one that fails to show gets a dip below baseline, timed to the exact moment it was supposed to land.

Kent Berridge and Terry Robinson pushed this further and split dopamine’s job in two: wanting(pursuit, effort, the drive to go get something) is dopamine-dependent; liking (the actual pleasure once you have it) runs mostly on a separate opioid system^[2]. Dopamine-depleted rats still show every sign of enjoying sucrose once it’s in front of them. They just won’t work for it anymore. Wanting and liking can come apart completely.

That’s the mechanism behind a pattern I know well: the milestone stops feeling like enough the moment it’s hit. Not because I don’t value it — the liking system is fine. It’s that the wanting system doesn’t pay out on arrival. It paid out on the gap, and the gap is gone, so it’s already scanning for the next one. I wrote about the founder version of this appetite in On Ambition without knowing the neural name for it. The name is reward prediction error, and it explains why the fix was never going to be a bigger milestone.

Oxytocin doesn’t create warmth out of nothing. It lowers the price of trusting someone you don’t yet have proof about. In the clearest demonstration of this, 128 men played a one-shot trust game with real money after a nasal spray of either oxytocin or placebo^[3]. Paul Zak and colleagues have since run versions of this with generosity, not just trust, as the outcome, and found the same shape^[4]: give people more oxytocin, and the leap of extending trust to a stranger gets measurably smaller.

In pair-bonding species, the same system (oxytocin plus its close relative vasopressin) is what turns a mating encounter into a durable attachment — the difference between prairie voles, who pair for life, and their close cousin the montane vole, who doesn’t, tracks almost entirely to oxytocin and vasopressin receptor density in the brain’s reward circuitry^[5].

I used to think of warmth and being a hard-nosed coach as two different settings I had to switch between. The actual mechanism argues they run on the same circuit: a coaching debrief after a brutal workout with one of my runners works because the earned trust makes the next hard ask land, not despite it. I wrote about that trade directly in On Mentorship. This is the chemistry underneath the “warm demander” framing — not a personality trick, a system that makes vulnerability cheaper once trust is banked.

Walter Cannon named it “fight or flight” in 1915: under threat, the adrenal medulla dumps adrenaline, the sympathetic nervous system takes over, and the body redirects blood, glucose, and attention toward whatever is about to happen^[6]. In the brain, the matching system runs on norepinephrine out of a small structure called the locus coeruleus, which acts like a gain knob on the rest of the brain — turn it up and focus narrows and sharpens; turn it up too far and that same focus becomes unable to disengage from the threat long enough to think^[7].

The popular version of this — performance rises with arousal, then falls, one universal curve — flattens the actual 1908 finding, which is that the peak shifts left as the task gets harder or less rehearsed^[8]; a century of citing it as one fixed law has been flagged directly as a misunderstanding of the original data^[9]. A well-drilled task tolerates a lot more adrenaline before it breaks down than a novel one does.

Race pace on a course I’ve run in training a dozen times is the rehearsed condition — the adrenaline sharpens it. A cold-open pitch to investors is the novel condition, and the same spike that helps the run works against the pitch. Same chemical, same gland, completely different optimal dose, because the task is different.

Cortisol runs on a slower clock than adrenaline — minutes and hours, not seconds — through the hypothalamic-pituitary-adrenal axis. In a healthy rhythm, it spikes 50% or more in the first 30 to 45 minutes after waking (the “cortisol awakening response”), then declines steadily across the day^[10]. Bruce McEwen’s concept of “allostatic load” is the cost of that system getting activated over and over without enough recovery between activations — not damage from any one spike, but wear from a system that never fully resets^[11].

Robert Sapolsky’s framing of this is the one that stuck with me: a zebra that outruns a lion has its stress response shut all the way off within minutes of reaching safety. A person who survives a bad client call keeps replaying it at midnight, and the replaying keeps the same physiological system switched on long after the actual threat is gone^[12]. Rumination is the human-specific failure mode of the same system: built for animals that shut it off and reach a species that mostly doesn’t.

Taper week is the honest version of this for me. Sports-science reviews of tapering document real mood disturbance during the reduced-volume weeks before a race, tracked through standard mood inventories^[13]. The mechanism — a hard training session is normally where a day’s worth of cortisol gets used up, and removing the session removes the outlet without removing the hormone — is my own read of why, not something the review measured directly.

The runner’s high has a real mechanism, and it took until 2008 to get direct evidence for it in a human brain rather than a blood draw — endorphins don’t cross the blood-brain barrier easily, so measuring them in blood was never proof of what was happening in the skull. Ten trained runners were scanned with PET imaging before and after a 2-hour, roughly 21.5km run. Self-reported euphoria rose from 37.6 to 73.3 out of 100, and the increase tracked with reduced opioid-receptor binding — more opioid released, so fewer receptors were free for the tracer to find — specifically in the brain’s emotion-processing regions^[14].

A separate line of work suggests endorphins aren’t doing this alone. Anandamide, an endocannabinoid, spikes measurably after sustained running in both humans and dogs — species built for endurance running — but not in ferrets, and not after walking^[15]. That’s a strange, specific finding: the reward system isn’t rewarding exertion in general. It’s rewarding the particular kind of sustained aerobic effort that cursorial animals evolved to do.

None of that is available on mile two. It shows up late, after the system decides the effort is real and sustained — closer to mile sixteen than mile two, on my own watch. It’s the thing I tell Ian and Max before their first long run past 18 miles: the part that makes you want to quit is scheduled to happen before the part that pays out, not after it.

GABA is the brain’s main brake — the inhibitory counterweight to glutamate’s excitation, and the receptor benzodiazepines act on. A single bout of vigorous exercise measurably raises both GABA and glutamate in the visual cortex within the same session^[16], which is a more mechanistic answer than I expected to “a hard workout clears my head” — it isn’t a metaphor. The brake system gets a real, short-term boost.

Serotonin doesn’t spike and dip the way dopamine does. It’s tonic rather than phasic — a slower-moving regulator of mood and behavioral restraint that the phasic systems operate on top of^[17], associated more with harm avoidance and behavioral inhibition than with pursuit or reward^[18]. It responds to diet, sunlight, and exercise on the order of days, not seconds — which is a large part of why fixing a bad mood by chasing a dopamine spike (a new project, a new purchase) doesn’t actually work, and going outside sometimes does.

The honest caveat: none of this is destiny. A year of aerobic exercise physically grew brain tissue in adults in their sixties and seventies — a measurable increase in hippocampal volume, not a metaphor^[19]. If a system this “fixed” is still building new volume on a one-year timescale, treating any of these seven systems as a life sentence is a category error. Twin studies put the heritability of major personality traits around 40 to 50 percent^[20], which means roughly half of the variance is something other than fixed wiring — environment, practice, circumstance, and the parts of a life nobody can fully model.

The other honest caveat is about who gets to use any of this. Everything in this piece assumes a person with room to structure a taper, book a therapy session, and choose when to train — a level of control over one’s own schedule and stress exposure that is itself a resource, not a given. Knowing the mechanism behind a bad week is a genuine advantage. It is not available to everyone on the same terms, and I’d rather say that outright than write around it.

What I do think holds: the discomfort of taper week has a name now, and it isn’t a personal failing. It’s cortisol with nowhere to go. The founder-brain restlessness the moment a milestone lands has a name too — reward prediction error, already scanning for the next gap. None of that makes the feelings optional. It does mean I stopped arguing with them and started scheduling around them.