Someone warm on Tuesday and cold on Thursday is harder to leave than someone who was simply unkind throughout. That is a fact about a pattern rather than a fact about how right they are for you, and mistaking the first for the second is what keeps people in place for years.
The pattern, in plain terms
They are warm sometimes and absent sometimes, and you cannot find the rule. So you keep testing. You send the message, adjust the tone, wait longer, wait less, and every so often the warmth comes back.
That occasional return is the whole mechanism. When a response comes every time, its absence tells you something has changed and you stop. When it comes sometimes, unpredictably, no single silence carries any information at all. Nothing ever tells you the run is over, so the testing continues, and it continues long after you would have stopped with someone consistent.
The bit that surprises people is that inconsistency is stronger than consistency here. Steady warmth is easier to walk away from than warmth you have to earn, which is why the relationship you cannot leave is often not the best one you have had.
Where the idea comes from, and where it does not
The mechanism is long established in learning research generally, in settings that have nothing to do with romance. It has never been tested on adult romantic attachment in the way the internet implies when it explains your relationship to you with a diagram.
The same goes for the vocabulary. There is no diagnosis in this territory. Love addiction and relationship addiction are recognised nowhere, in the DSM-5 or the ICD-11, and "trauma bonding" is popular vocabulary rather than a clinical category. A 2025 validation paper puts the general position plainly: these patterns are ["not yet considered official diagnosis"](https://pmc.ncbi.nlm.nih.gov/articles/PMC12044147/) in the DSM-5-TR or the ICD-11, and researchers "disagree on whether it should be defined narrowly or broadly".
So treat the schedule as a lens rather than as a law. It fits what people describe. It has not been demonstrated here.
What the evidence in this area actually is
Thin, and you should know how thin before you build anything on it.
A 2023 systematic review in *Health Psychology Research* screened 102 articles and ["included five"](https://pubmed.ncbi.nlm.nih.gov/37139461/). A 2026 review of 102 studies in *Archives of Sexual Behavior* found ["pronounced conceptual heterogeneity in the literature"](https://pubmed.ncbi.nlm.nih.gov/42029817/), meaning researchers are not measuring the same thing as each other.
The one solid anchor is attachment. A systematic review and meta-analysis of 15 studies covering 3,628 people found anxious attachment correlated with what the literature calls love addiction at [r = 0.39, a large effect, and avoidant attachment at r = -0.09](https://pmc.ncbi.nlm.nih.gov/articles/PMC12284683/). The authors flag that the instruments differed between studies and most samples were non-clinical.
Put together: an alarm system calibrated to fire early, meeting a person whose signals arrive unpredictably. That is a plausible account of why this particular person is harder to leave than the others. It is not a proof.
Why "we have a connection" feels like evidence
Intensity produced by inconsistency reads as significance, and that is the trap.
The uncertainty generates constant attention, and constant attention feels like depth. The rare warmth arrives against a background of scarcity, which makes it feel more valuable than the same warmth would feel from somebody who offered it every day. And the relief when they come back is genuinely enormous, because the alarm was genuinely loud.
None of that is fake. All of it is a poor guide to compatibility. The plainest test is the one in [is it love, or is it a compulsion](/guides/love_relationship/is-it-love-or-is-it-a-compulsion): what happens when things go calm and they become fully available.
What the frame actually changes
Three things, and they are practical rather than philosophical.
**It stops you reading intensity as destiny.** "I've never felt like this about anyone" becomes a statement about the schedule you are on rather than a verdict about the person. That single reframe does more work than most advice in this area.
**It tells you what to remove.** Not their good qualities, which are usually real. The unpredictable signal. The notification, the story view, the second account, the mutual friend who reports back. Each fresh piece of information restarts the loop, which is why closing channels outperforms resolve. [No contact and how to actually do it](/guides/love_relationship/no-contact-how-to-actually-do-it) is the full sweep.
**It predicts that it gets worse before it stops.** When a pattern that used to work stops working, the model says the testing intensifies first. In practice that means the messages get more frequent, or they suddenly reappear after you go quiet. Knowing that in advance is the difference between reading a spike as a relapse and reading it as an expected stage. That is what the model predicts. It has not been measured here.
Where the frame gets misused
**As an excuse.** "It's intermittent reinforcement" is a true sentence that can be deployed to explain why none of this is your responsibility, and it usually appears in the hour before somebody sends a message they had decided not to send.
**As a diagnosis of them.** The schedule does not tell you that they are cruel or calculating. Plenty of unpredictable people are unpredictable because their own lives are, or because they are avoidant, or because they are as confused as you are. You can act on the pattern without needing them to be a villain.
**As a reason to stay.** "He's not cruel, it's just a schedule" is the same sentence running the other way. The pattern is not a defence.
And the important exception. If the unpredictability includes fear, if the warm phases are apologies and the cold phases are frightening, this is not a reward schedule to be understood. It is abuse, and it needs different help.
- **National Domestic Abuse Helpline** (run by Refuge): 0808 2000 247, 24 hours a day, every day.
- **Men's Advice Line**, for male victims including trans and non-binary people: 0808 801 0327, Monday to Friday, 10am to 5pm.
- **Galop**, for LGBT+ people: 0800 999 5428, Monday and Tuesday 9.15am to 8pm, Wednesday to Friday 9.15am to 4.30pm.
- **Samaritans**: 116 123, any time.
- In an emergency, 999.
Common questions
**What is intermittent reinforcement in a relationship?** Warmth that arrives unpredictably rather than reliably. The idea comes from learning research in other settings, and it has not been tested empirically on adult romantic relationships, so treat it as a useful lens rather than a finding.
**Why is it harder to leave someone who treats me badly sometimes than someone who treats me badly always?** Because consistency gives you information and inconsistency does not. With someone consistently unkind, the pattern is legible and the decision is available. With someone unpredictable, no single bad week ever settles anything.
**Is this a trauma bond?** That term has no diagnostic status and no agreed definition. If the underlying experience includes fear, use the helplines above rather than the vocabulary.
**Does understanding it make it stop?** No, on its own. It tells you what to change, which is the unpredictable signal rather than your feelings about them. [The urge routine](/guides/love_relationship/getting-through-the-urge-to-text-them) is the version for tonight.
**Will they come back when I go quiet?** The model predicts that contact attempts often increase when the old pattern stops working, and plenty of people describe exactly that. Decide in advance what you will do with a message, because you will not decide it well while reading one.
Sources
- [Love addiction and attachment styles: a systematic review and meta-analysis](https://pmc.ncbi.nlm.nih.gov/articles/PMC12284683/). 15 studies, 3,628 participants, anxious attachment r = 0.39, avoidant r = -0.09.
- [Love Addiction, Current Diagnostic and Therapeutic Paradigms in Clinical Psychology (Health Psychol Res, 2023)](https://pubmed.ncbi.nlm.nih.gov/37139461/). 102 articles screened, five included.
- [Review of 102 studies, Archives of Sexual Behavior, 2026](https://pubmed.ncbi.nlm.nih.gov/42029817/). Pronounced conceptual heterogeneity in the literature.
- [Addicted to love? Validity evidence for the Love Addiction Inventory, Brazilian version (2025)](https://pmc.ncbi.nlm.nih.gov/articles/PMC12044147/). Not an official diagnosis in DSM-5-TR or ICD-11.
