Intermittent reinforcement
In a 2011 experiment, women were more drawn to men who might like them a lot or only an average amount than to men who were known to like them a lot. That puzzle is the everyday face of intermittent reinforcement: a reward that shows up on an unpredictable schedule grips attention harder than one you can count on. It matters to anyone who has checked a phone for one particular name, or kept opening an app that pays out in matches at random.
The term is also called variable reinforcement when the payout varies in size or timing, and the mechanism is the same: swipe culture is built on top of it, the way a casino floor is built on a slot machine. The point of a name for it is the one useful move it allows: noticing the schedule before it has finished doing its work on you.
Read people who answer in full sentencesFrom animal learning labs in the 1950s. B. F. Skinner and Charles Ferster catalogued how often and how predictably a reward arrives, in their 1957 book Schedules of Reinforcement, and showed that the schedule alone changes how an animal behaves. Psychologists later sorted them into fixed or variable, by number of responses or by time.
The schedule that matters for dating is the variable one. In a 2020 review of how online environments are designed, Kozyreva, Lewandowsky and Hertwig note that variable-ratio schedules produce the highest rates of response. Reward after an unpredictable number of tries is the pattern that keeps a pigeon pecking and a person scrolling.
Because certainty lets the mind close the file. The psychologists behind the 2011 study, Erin Whitchurch, Timothy Wilson and Daniel Gilbert, told 47 female undergraduates that four male students had seen their Facebook profiles and rated them. One group heard the men had liked them a lot. Another heard an average amount. A third heard only that it was one or the other. The uncertain group was the most attracted to the men, more than the group told the men liked them a lot.
The measured link was thought. Uncertain participants reported thinking about the men the most, and that frequency of thought accounted for part of the extra attraction. The authors' own reading is that people take an intrusive thought as a sign of feeling, so a man who keeps coming to mind must be someone you like. They add a condition that is easy to skip past. The effect should hold only when there is some initial attraction to begin with.
There is a neural footing too. In a 2003 study of monkeys, Fiorillo, Tobler and Schultz found that dopamine neurons showed a gradual rise in activity ahead of a possible reward, a signal that tracked uncertainty, which is highest when the odds are even. Nobody has put a camera inside a dating thread. The parallel is suggestive, not proven.
It looks like warmth you cannot forecast: attention that comes in bursts, with no pattern you could write down and no reason you were given. The common shapes are easy to recognize once you have a list.
- A reply within minutes on Monday, then nothing until Thursday night.
- Long messages on one day, one-word answers the next, with no event in between.
- A plan floated ("we should do that") and never confirmed or cancelled.
- A like or a match that arrives in bursts, then none for a week.
- The typing bubble that appears, disappears and does not return.
None of these needs bad intent. The schedule works on the person receiving it whether or not anyone chose it, which is why the table below compares how the two kinds of warmth feel from the inside.
| Steady warmth | Intermittent warmth | |
|---|---|---|
| What you can predict | Roughly when and how they will answer | Neither, so every gap is a question |
| Where attention goes | To the conversation itself | To the next possible reply |
| What a reply feels like | Nice, then filed away | A small win that restarts the waiting |
| What silence feels like | Ordinary | An open loop |
A line from the 2011 paper explains the right-hand column. When people are uncertain about an important outcome, the authors write, they "can hardly think about anything else".
An app that shows profiles one after another, with a match behind a small and unknowable share of them, is a variable-ratio machine. You cannot tell which swipe pays, so each one carries the same small charge. Winter and colleagues, opening a 2025 study of problematic dating-app use, say that the unpredictability and intermittent reinforcement of such apps may reinforce continuous use. Note the "may": that is a hypothesis in the paper's background, not a finding of their data.
Notifications raise the same lever. Kozyreva and colleagues pass on the Center for Humane Technology's point that the variable schedule of notifications can turn checking a phone into a habit.
A fair design question is whether a service has to run on this schedule at all. Anketta has no swipe. You read what someone wrote, and a match opens a 48-hour window in which the two of you write in full sentences you can reread. Silence ends the match instead of leaving it hanging, and it becomes permanent only if both people actually wrote to each other. That does not abolish uncertainty, nothing can, but it caps how long a gap is allowed to run.
Read a manuscript before you decide to wait on itNo. People are busy, avoidant, ambivalent or just bad at texting. None of that is strategy. Intermittent reinforcement describes what the pattern does to the person on the receiving end, not what is in the sender's head. That is the useful thing about it: you can recognize the effect without having to diagnose the other person.
Two cases do look like a strategy: a run of near-absence followed by a warm return, and the same cycle repeating with different excuses. A return after silence has its own entry under hoovering, and a steady drip of just-enough attention is breadcrumbing. Both are narrower than the schedule they borrow.
You cannot switch the pull off, but you can change what it feeds on, so that a reply stops being the only thing your evening is waiting for. Five moves, in the order that costs the least:
- Name it in one sentence: "I am checking because the next reply might be a good one."
- Move your checking to fixed times, so the reward stops being tied to your hand reaching for the phone.
- Ask a plain question that has a plain answer, such as when you will see each other. A person who answers it has stepped out of the schedule.
- Count patterns across a month, not replies across an evening. How long to wait to text back is a practical companion for the evening part.
- Spend your attention where the signal is steady: a written profile you can reread, a conversation with an end date, a person whose Tuesday looks like their Saturday.
If one person has taken up most of your mind for weeks, limerence is the neighboring term worth reading.
What is intermittent reinforcement in simple terms?
A reward that comes some of the time instead of every time, with no way to predict which time. The unpredictability keeps people trying: it makes the reward feel closer than a certain one would. Casinos, notifications and slow-burning texting patterns all lean on it.
Is intermittent reinforcement the same as variable reinforcement?
Close. People use them interchangeably. Intermittent is the broad family: rewards that do not come every time. Variable is the sub-type where the gap, in number of responses or in time, keeps changing, which is the form that produces the highest response rates in the research.
Does the 2011 study mean you should play hard to get?
No. The authors tested uncertainty about how much someone liked a person they had only seen online, and the effect held when there was some initial attraction. It describes a pull, not a technique, and it says nothing about whether the relationship that follows is a good one.
Why do I think about the person who replies less?
Uncertainty keeps the question open, and an open question stays in view. In the 2011 experiment the uncertain group thought about the men most, and that frequency partly explained their stronger attraction. Your mind is working on a problem it cannot yet solve.
Do dating apps use intermittent reinforcement on purpose?
The mechanics are there in the design: unpredictable matches, unpredictable notifications. Published work describes the schedule as a plausible driver of continued use, but intent is a different claim and nobody outside the companies can verify it. The effect on the user is the same either way.
How does a text-first app change the schedule?
It changes the unit being rewarded. When what you get is a paragraph someone wrote, not a face that appeared, the reward is the reading itself. It is the same each time you open it. On Anketta a 48-hour window also limits how long a silence can stretch.
How do I know if it is working on me?
Notice where your attention sits during the gaps. If you rehearse their next message more than you enjoy the last one, or you open the app with no reason except the chance of a reply, the schedule is running. The fix is not willpower but a fixed time to check.
A daisy runs out of petals by the twentieth pull; a half-answered message never does. That is the only reason it keeps being picked up.
Unsure about writing? Try reading first.