Intermittent Reinforcement

The question that shames more survivors than any other: why can’t I stop wanting someone who treated me like that? The answer is not in your character. It is in a reinforcement schedule.

Of all the questions the aftermath produces, one carries the most shame, and it is asked in the dark, rarely out loud: what is wrong with me? The relationship was demonstrably harmful: the record says so, the pattern says so, some part of you has known for years. And yet the pull toward them is physical. The good memories arrive in high resolution at 2 a.m. The craving persists long after the reading is complete and the exit is made. Clarity, it turns out, does not switch off longing, and the gap between what you know and what you want feels like evidence of a broken will.

It is not. It is evidence of a schedule. The mechanism has a name: intermittent reinforcement, and it is one of the most robust findings in behavioural science. This piece reads it carefully, because of everything this newsletter has covered, this is the mechanism where understanding does the most direct work: the shame runs on the belief that the attachment is a verdict on you, and the science says, precisely, that it is not.

The Finding

The core discovery goes back to B.F. Skinner’s laboratory work in the mid-twentieth century, and it can be stated in one line: rewards delivered unpredictably produce stronger, more persistent behaviour than rewards delivered every time.

The experimental picture is famous. An animal presses a lever; food arrives. If food arrives on every press, continuous reinforcement, the animal learns quickly, and also unlearns quickly: stop the food, and the pressing stops soon after. But if food arrives unpredictably, sometimes on the third press, sometimes the thirtieth, on no schedule the animal can detect, something counterintuitive happens. The pressing becomes more intense, not less. And when the food stops entirely, the behaviour persists far, far longer before fading. In the vocabulary of the field, variable-ratio schedules produce the highest response rates and the greatest resistance to extinction. The unpredictability is not a weakness of the reward. It is an amplifier.

The everyday proof is the slot machine, which is a variable-ratio schedule with lights on it. Nobody stands for hours at a vending machine that pays out every time; the reliable machine generates no compulsion at all. The machine that might pay out this pull, or the next, or the fiftieth, holds people in place through hundreds of losses. The losing pulls do not weaken the behaviour. Inside a variable schedule, they are the behaviour’s fuel: each unrewarded attempt raises the felt probability that the next one pays.

Hold that structure in mind: reward, unpredictability, persistence beyond all reason — and read your relationship against it.

The Schedule in the Relationship

The dynamic this newsletter describes runs a variable-ratio schedule on attachment, and by now you can name the components yourself.

The reward is the good version: the warmth, the charm, the person from the beginning: attentive, present, apparently returned. The cycle piece of this literature calls its reappearances by various names; structurally it is the payout. The presses are everything you did in pursuit of it: the patience, the accommodation, the swallowed objections, the extra giving, the trying. And the schedule is the alternation the romantic operation piece described: warmth and withdrawal delivered on no pattern your behaviour could stabilise, because the pattern was never keyed to your behaviour at all. Some weeks the tenth accommodation produced the good version. Some months nothing did. Sometimes it arrived unearned, after the worst stretch, luminous.

Notice what this explains, item by item, about experiences that otherwise read as personal failure.

It explains why the bond intensified as the treatment worsened. Under continuous warmth, attachment is strong but ordinary, and, crucially, it can end when the warmth ends. Under variable warmth, the attachment system does what the lab animal does: it escalates. More pressing, more vigilance, more of your life organised around the possibility of the next payout. The worst stretches did not weaken the bond, and this always seemed like madness. It was not madness. It was the schedule, working exactly as documented, the losing pulls raising the stakes of the next attempt.

It explains why the good days hold such disproportionate weight in memory. Within a variable schedule, the payout is the event everything else is organised around; the neurochemistry of anticipation, dopamine is keyed to unpredicted reward far more than to reliable reward, stamps those moments in at an intensity a stable relationship never generates. The vividness of the good memories is not proof of what the relationship truly was. It is a property of the delivery schedule. Stable love, precisely because it is stable, is stored more quietly.

And it explains the aftermath’s cruellest feature: why the craving outlasts the clarity. Resistance to extinction is the defining signature of the variable schedule: the behaviour persists long after the rewards have stopped, because the schedule trained the system to treat long droughts as normal preludes to payout. The 2 a.m. longing months after exit is not your judgement relapsing. It is extinction running its documented course, slower than anyone deserves, on the timetable the schedule set, not the timetable your clarity sets. The aftermath piece called this load withdrawal, and meant it nearly literally.

What This Reading Is For and What It Is Not For

Two calibrations, because this mechanism is widely cited in this space and often mishandled in both directions.

First: this reading is a description of your side of the mechanism, what the alternation did to your attachment system, and it holds regardless of what was happening on the other side. It does not require believing the schedule was consciously designed, tuned like a laboratory apparatus; the framework’s standing position applies here too: the interior is not yours to certify, and the reading does not need it. What is observable is that the alternation ran for years, that it was never stabilised by any behaviour of yours, and that its effects on you are the textbook effects. The outputs, as always, are enough.

Second, and more important: the point of this reading is the retirement of the shame; not the licensing of the return. The mechanism explains the pull; it does not validate obeying it. Here the slot machine earns its place in the analogy one more time: understanding variable-ratio schedules does not make the machine safe to play. It makes the machine legible, and legibility changes your relationship to the pull without pretending the pull away. The craving, read correctly, gets reclassified: from evidence about them (they must have been special; this must have been love) and evidence about you (I am weak; I cannot be trusted) to evidence about a schedule, an artefact of delivery, carrying no information about the person’s worth or yours, fading on the slow curve that variable schedules are documented to fade on.

That reclassification is the practical work of this entire piece. The 2 a.m. wave arrives; some part of you now says extinction curve, right on time, and that part, as the DARVO piece put it in another context, is the part that is not drowning. The wave still comes. It is smaller for being named, and it passes without being obeyed. Enough repetitions of that and the schedule finally meets the one thing it cannot survive: a system that has stopped pressing the lever and knows why it wanted to.

You were not weak. You were on the strongest schedule behavioural science has ever documented, and you walked out of the casino anyway, with the craving still live, which is the only way anyone ever walks out. That is not the failure of your recovery. That is what the recovery of a well-built attachment system, exploited by an unpredictable one, actually looks like from the inside.

Once you can read the code, the code stops working, and this part of the code was written into the schedule, not into you.