The dopaminergic system

The catecholamine cell groups were numbered A1 upward in the 1960s, and the dopaminergic ones are A8 to A16 ^[2]. Three of them sit in the midbrain and do almost everything this page is about: A8 in the retrorubral field, A9 in the substantia nigra pars compacta, and A10 in the ventral tegmental area. A three-dimensional reconstruction in the rat counted roughly 1,300 A8 cells, 10,500 A9 cells and 10,200 A10 cells on one side of the brain ^[3]. The remaining groups sit in the hypothalamus, the olfactory bulb and the retina, and one of them, A12 in the arcuate nucleus, has a job that has nothing to do with behaviour at all.

In a human the numbers are larger but still modest. Unbiased stereology puts the average total at about 550,000 pigmented neurons in the substantia nigra of a healthy adult, alongside about 260,000 non-pigmented ones; in Parkinson's disease the pigmented population is 66 percent smaller ^[4]. A single one of those neurons carries an axon that branches enormously, so the ratio of released transmitter to releasing cells is extreme. That is the anatomical reason dopamine behaves like a broadcast rather than a wire.

How dopamine is cleared varies by region, and this detail explains more than it looks like it should. In the striatum the dopamine transporter (DAT) pulls it back in quickly. In the prefrontal cortex the dopamine varicosities show only sparse DAT staining, so clearance falls instead to the enzyme COMT and to the noradrenaline transporter ^[11]. Two consequences follow: cortical dopamine lingers longer and spreads further than striatal dopamine, and a drug that blocks DAT does much more to the striatum than to the cortex. It is also why a common variant in the COMT gene has measurable cognitive consequences while having very little to say about the striatum ^[13].

Dopamine does genuinely different things depending on where it is released ^[2]. The four named pathways below are a simplification of a messier anatomy, but they are the right simplification, because the clinical and pharmacological consequences really do sort along these lines.

The practical point is that no drug picks one. Raising dopamine broadly sharpens focus and affects movement, mood and hormones at once, and a side effect is usually just the same mechanism arriving in a pathway nobody was aiming at. Antipsychotic-induced prolactin elevation is the cleanest example: the intended target is mesolimbic, the prolactin comes from the tuberoinfundibular pathway, and the drug cannot tell them apart ^[19].

The single most influential result in this field is that midbrain dopamine neurons fire to surprise, not to reward. A reward that was fully predicted produces no burst; an unexpected one produces a large burst; an expected one that fails to arrive produces a dip below baseline at exactly the moment it was due. That pattern is a reward prediction error, the same quantity a reinforcement-learning algorithm needs in order to update its estimates, and it is why dopamine is described as a teaching signal ^[1].

A second line of work asks what the signal is for behaviourally, and the answer is closer to wanting than to liking. Dopamine depletion leaves hedonic taste reactions intact while stripping out the motivation to pursue; the proposal that came out of that work is incentive salience, meaning dopamine sets how much a cue pulls at you, separately from how good the thing feels when you get it ^[7]. The clinical version is that pleasure and motivation come apart, which is exactly what apathy in Parkinson's disease and anergia in depression look like.

Effort is the sharpest formulation. Depleting accumbens dopamine does not reduce appetite or the primary value of food; it reallocates behaviour away from options that require work and toward options that do not. Lever-pressing schedules with trivial requirements survive depletion almost untouched, while high-ratio schedules collapse ^[9][8]. Read that way, dopamine is less about wanting things and more about being willing to pay for them.

Two complications keep this from being tidy. The first is that dopamine neurons have two firing modes, a slow irregular tonic background that sets the ambient extracellular level and fast phasic bursts that carry the prediction error, and drugs do not affect them equally ^[6]. The second is that the neurons are not one population. Recording in monkeys during a task with both liquid rewards and airpuffs found that only a subset behaved the way the value hypothesis predicts; a larger number were excited by both the reward-predicting and the aversive-predicting cue, and those neurons sat more dorsolaterally in the substantia nigra ^[10]. Whatever the whole system is computing, it is not one number.

There are six ways in, and they differ enormously in how physiological they are. Blocking reuptake amplifies whatever the neuron was already doing; forcing release overrides it. That distinction matters more than potency.

Reuptake inhibitors block DAT so released dopamine lingers, and the effect size is measurable rather than theoretical: oral methylphenidate at around 0.8 mg/kg reduced striatal D2 receptor availability by about 20 percent in healthy volunteers, which is the imaging signature of a genuine rise in extracellular dopamine ^[14]. Modafinil belongs on the same list, whatever its marketing said. Positron emission tomography in monkeys showed it occupying 35 to 54 percent of striatal DAT depending on dose ^[16], and in humans at ordinary therapeutic doses it blocked DAT and raised extracellular dopamine ^[15]. See eugeroics for why that does not make it interchangeable with amphetamine.

Releasers such as amphetamine reverse the transporter and empty vesicles directly. That produces a much larger and much less physiological signal than blocking reuptake, and it is the mechanism most associated with tolerance and with the imaging changes seen in long-term stimulant users ^[18]. Direct agonists skip the neuron entirely; their effects depend on which receptor family they favour, which is a question for dopamine receptors rather than for this page.

The gentler levers are gentler mostly because they do less. Precursors supply raw material, and the honest summary is narrow: tyrosine reliably helps in short-term stressful or cognitively demanding situations and does very little otherwise, because synthesis is not usually the limiting step in a well-fed brain ^[17]. Enzyme inhibition slows degradation rather than adding anything, and COMT inhibition matters mainly in the cortex, where there is little DAT to do the clearing ^[11]. Several compounds in this catalogue sit in a fourth category with a plausible story and thin human evidence, including phenylpiracetam, whose only clear target in a broad receptor screen was a weak micromolar interaction with DAT, and bromantane, whose synthesis-upregulating account rests almost entirely on Russian-language work.

Force dopamine hard and often, and the system adapts. Receptors down-regulate, release blunts, and the result is tolerance plus, on the far side of the drug, less baseline motivation than there was to begin with. This is not folklore: pooled in vivo imaging comparing people who use cocaine, amphetamine or methamphetamine against healthy controls finds consistent striatal dopaminergic differences, including reduced dopamine release and reduced receptor availability ^[18]. That is the substrate of stimulant burnout and of the flat, uninterested day after a heavy session. See tolerance and dependence.

The deeper problem is that dopamine has no monotonic dose response in cortex. Prefrontal function follows an inverted U: there is an optimum, and both too little and too much degrade performance. Because individuals start at different points on that curve, the same drug at the same dose improves one person's working memory and worsens another's, and the direction is predictable from baseline rather than random ^[12]. A striking amount of the disagreement about whether stimulants help healthy people is this curve being sampled at different places.

Genetics gives a concrete handle on it. The common Val108/158 Met variant in COMT changes enzyme activity roughly four-fold, and therefore changes how fast prefrontal dopamine is cleared. In a study of 175 patients with schizophrenia, 219 unaffected siblings and 55 controls, genotype tracked in allele-dosage fashion with perseverative errors on the Wisconsin Card Sorting Test, explaining about 4 percent of the variance ^[13]. Four percent is small in absolute terms, and it is a real, measured, mechanistic reason two healthy people respond differently to the same dopaminergic drug.

The practical conclusion is unglamorous. The cleaner approach is to support synthesis, sleep and recovery rather than to force large releases the brain will immediately compensate for, and to treat any protocol that promises escalating dopamine as promising escalating tolerance. Note also that the popular "dopamine detox" framing has no basis in this biology; abstaining from a stimulus does not measurably reset receptor populations on the timescale those articles describe.

Parkinson's disease is the loss of the nigrostriatal pathway, and its numbers are worth knowing because they explain why the disease is diagnosed late. Pigmented neurons in the caudal substantia nigra fall away with normal ageing at a linear 4.7 percent per decade; in Parkinson's disease the loss is exponential, with roughly 45 percent gone in the first decade ^[5]. The regional pattern is the opposite of ageing: the lateral ventral tier, relatively spared by age, is the tier hit hardest by the disease, losing an average of 91 percent. By the time motor symptoms appear a large fraction of the population is already gone, which is why replacing dopamine with L-DOPA treats the symptom and changes nothing about the trajectory.

Psychosis is the other end. The current formulation of the dopamine hypothesis of schizophrenia places the abnormality at presynaptic striatal dopamine synthesis and release, treating that as the final common pathway through which genetic risk, obstetric complications, stress and drug use converge ^[20]. Its most useful implication is uncomfortable for drug development: every existing antipsychotic acts downstream of the actual abnormality, blocking receptors rather than correcting the presynaptic problem, which is a plausible reason they control positive symptoms and do so little for the rest.

Prolactin is the quiet third failure mode. Dopamine reaching the anterior pituitary through the hypophysial portal blood tonically suppresses prolactin release, so anything that blocks D2 receptors releases that brake and raises prolactin, while dopamine agonists are the treatment for hyperprolactinaemia ^[19]. It is the clearest case in the whole system of a therapeutic effect and a side effect being the identical mechanism in two different places.

Several important things are genuinely unresolved. Whether phasic bursts or tonic level matter more for a given behaviour is still argued, and drugs move the two differently ^[6]. Why a substantial subset of dopamine neurons responds to aversive events as well as rewarding ones has no settled interpretation ^[10]. Whether any dietary or supplemental intervention meaningfully changes dopaminergic tone in a healthy, well-nourished adult is not established; the tyrosine literature is the best case and it is narrow ^[17]. And the question that matters most to readers of this site, whether raising dopamine improves cognition in people who are already well, has an answer that depends on where on the inverted U they started, which no consumer product can measure ^[12].