Don’t leave me
Someone I loved once used to say it out of nowhere. Don’t leave me.
We’d be talking about something unrelated (our day, a plan for the weekend, something we’d watched together) and the sentence would arrive, quiet and unprompted, like it had been waiting its turn. Please don’t leave me. Nothing in the conversation or in my behaviour would have suggested leaving was anywhere near the horizon. I was deeply in love. I’d told her so in most of the ways a person can tell another person that.
The first few times I heard it I thought I hadn’t been clear enough. So I’d get clearer. I’d reassure her directly. I’d be more present, more attentive, more affectionate, more explicit about the future I wanted with her. The words would land, for a little while (an hour, sometimes a day), and then the sentence would come back. Don’t leave me.
I spent a long time confused by this. Sometimes I was hurt by it, because the implication seemed to be that my love wasn’t landing, which felt like a verdict on me. Sometimes I wondered what I was getting wrong. Mostly I thought the solution was to love her more visibly. Louder words. Bigger gestures. More reassurance. It never worked for long.
It took me years to understand that I was not the problem and neither was she. We were speaking past each other on a plane I didn’t know existed.
Love Languages was not the problem
Most of my generation of dating-literate people grew up hearing about Love Languages. Gary Chapman’s 1992 book The Five Love Languages gave us a vocabulary for how different people prefer to be loved: Words of Affirmation, Quality Time, Acts of Service, Gifts, Physical Touch. Chapman is a Baptist pastor and marriage counsellor, and the idea grew out of his counselling work, not out of research. The claim is that each of us has one or two we most naturally give and most naturally receive, and that a lot of relational mismatch comes from one partner expressing love in a dialect the other doesn’t easily hear.
It’s a useful framework. I’ve found it genuinely helpful. (As a vocabulary, that is. A 2024 review by relationship scientists Impett, Park and Muise found that the small body of studies gives little support to the book’s core assumptions, including that each person has one preferred language and that couples are happier when they match.) And, in the relationship I’m describing, I had been fluent across several of those languages. Words. Quality time. Touch. The messages were being sent, clearly, in multiple dialects. They were getting through.
They just weren’t being believed.
That distinction is the whole point of this article, and it’s a distinction Chapman’s framework wasn’t designed to address. Love Languages tell you how people prefer to be loved. Attachment theory tells you whether they can receive it at all. One is about dialect. The other is about whether the radio is turned on.
You can speak every love language fluently to a partner whose attachment system is convinced they’re about to be abandoned, and the signal won’t land, not because you aren’t sending it, but because there’s a much older, much deeper pattern interpreting the incoming signal as unreliable. No amount of louder signalling solves that. What solves it is a different kind of literacy.
This article is about that second kind of literacy.
Where attachment theory comes from
The original work goes back to British psychiatrist and psychoanalyst John Bowlby in the 1950s and 60s, who was studying what happened to children who’d been separated from their primary caregivers. He observed that the early bond between an infant and its caregiver wasn’t just emotional decoration. It was a survival system, with predictable behavioural patterns when the bond worked and equally predictable patterns when it didn’t.
His collaborator, Mary Ainsworth, developed the famous Strange Situation procedure in the 1960s: a structured observation in which an infant is briefly separated from their caregiver and then reunited. From the infant’s responses (how they coped during separation, how they reconnected after), Ainsworth identified distinct styles of attachment: secure, anxious-resistant, avoidant. A fourth, disorganised, was added later by Mary Main and Judith Solomon.
For years this stayed largely in the world of child development. Then, in the late 1980s, two researchers, Cindy Hazan and Phillip Shaver, did something quietly radical: they asked whether Ainsworth’s three patterns might also describe how adults behave in romantic relationships. Their 1987 paper carried Bowlby and Ainsworth’s framework over to adult love, and it landed. (The four-style version that most dating articles use came four years later, from Bartholomew and Horowitz.) Decades of research since have broadly confirmed that adult attachment is a real, measurable thing. The link back to childhood is weaker than the popular version suggests: R. Chris Fraley’s 2002 meta-analysis found attachment security only moderately stable from infancy to early adulthood, so early patterns nudge adult romantic life more than they dictate it.
Chapman was writing in parallel, more or less, about something real but much closer to the surface: how we prefer to give and receive love. Bowlby and his intellectual descendants were writing about something underneath that: whether we can trust love to stay, whether we can let it in when it does. The two frameworks don’t replace each other. One sits on top of the other.
The four styles, plainly
Most popular attachment frameworks talk about four adult styles, following Bartholomew and Horowitz’s 1991 model. Researchers now place people on two dimensions, which they call attachment avoidance and attachment anxiety. In plain terms: how comfortable you are with closeness (intimacy, vulnerability, dependence) and how comfortable you are with distance (independence, your own space, a partner having theirs). Where you land on those two dimensions is what the style labels describe.
How common is each? The best-known figures come from the United States and use a three-category self-report. In a 1997 national sample, 59% of people chose the secure description, 25% the avoidant and 11% the anxious (Mickelson, Kessler and Shaver). There is no well-established figure for the fearful style, so treat any percentage you see quoted for it with caution.
Secure
Comfortable with closeness and comfortable with distance. Doesn’t panic when someone needs space; doesn’t withdraw when things get intimate. Gives reassurance freely. Receives reassurance freely. Conflict feels uncomfortable but solvable, not catastrophic.
In dating, secure attachment looks unremarkable, which is exactly why it’s hard to spot if you’ve never had it modelled. There’s no drama. There’s no agonising over messaging timing. There’s no power play. The person likes you, says so, makes plans, follows through on them. They’re not afraid of vulnerability and they’re not threatened by your independence. If something goes wrong they say so. If you say something has gone wrong they listen.
People with anxious or avoidant patterns can at first read secure attachment as “boring”, because it doesn’t produce the familiar ups and downs they associate with love. That is an observation from popular attachment writing more than a tested finding.
Anxious-Preoccupied
Very comfortable with closeness. Very uncomfortable with distance. The default fear is they will leave me, or stop loving me, or already have stopped and just haven’t told me. This is the style I’d seen up close without understanding it. Don’t leave me is, almost word-for-word, what the anxious-preoccupied system narrates when it spikes, and it spikes not in response to real signs of leaving, but in response to any moment of interior distance, often generated by the person’s own nervous system rather than anything the partner has actually done.
In dating, this looks like: hyper-attuned to small shifts in tone, constantly scanning for signs of waning interest, prone to the protest behaviours (extra texts, asking “is everything okay?” repeatedly, picking small fights to force engagement). Falls quickly. Idealises easily. Can be remarkable at intimacy when given safety; can also overwhelm a partner who needs more space.
Anxious attachment is often confused with “intensity” or “loving deeply”, which it can resemble, but at root it’s a regulation strategy. (Attachment researchers such as Mikulincer and Shaver write about both insecure patterns as strategies for regulating emotion.) The person isn’t actually trying to be that close; they’re trying to stop the alarm from going off in their nervous system about distance. When a well-meaning partner responds to the alarm by sending more love, the underlying pattern doesn’t update, because the pattern isn’t about the partner in the first place. That’s the bit I didn’t understand for years.
Dismissive-Avoidant
Comfortable with distance. Uncomfortable with closeness. Default move under pressure is to withdraw, both physically and emotionally. Independence is heavily prized, sometimes to the point of refusing real interdependence even when wanted.
In dating, this looks like: pulls back when things get too close, finds reasons the person isn’t quite right, often around the time things would otherwise get serious. Tends to see partners as “needy” or “too much” before fully understanding their own role in creating the dynamic. Can be self-sufficient and capable in admirable ways; can also struggle to stay present when a partner needs them emotionally.
The avoidant strategy isn’t a lack of feeling. It’s the same alarm system as the anxious strategy, just optimising for the opposite extreme: the threat is closeness, not distance, and the deactivation kicks in to dial things back to a comfortable temperature.
Fearful-Avoidant (sometimes loosely called disorganised)
The hardest one to describe because it’s both. Wants closeness and fears it. Tends to oscillate: pull people in, then push them away, then panic at the distance and pull them back in. The popular explanation is early experiences where the source of comfort was also the source of fear, leaving the system unable to settle into either strategy cleanly. That account comes from research on infants (“disorganised” is an infant classification, from a different measure at a different age), and how well it carries over to the adult style is less clear.
In dating, this often looks confusing, to the partner and to the person themselves. Intense connection followed by sudden withdrawal followed by hot pursuit. Difficulty trusting that closeness is safe. Often a history of dramatic, on-again-off-again relationships.
If this one sounds like you, professional support is worth considering, not because the style is “worse” but because the pattern can be harder to untangle on your own.
Why attachment matters more than Love Languages in dating
Most dating advice treats compatibility as a question of values, life goals, shared interests and, if it’s progressive, Love Languages. All of those things matter. But they aren’t the whole story of whether a relationship feels safe and steady or anxious and turbulent. Attachment patterns matter at least as much. In a 2012 review of 73 studies, higher attachment anxiety and higher avoidance both went with lower relationship quality (Li and Chan).
Two people can have every Love Language fluently covered and still build a relationship that grinds them both down. The classic painful pairing: anxious + avoidant. The anxious partner needs closeness reassurance; the avoidant partner experiences that need as pressure and pulls back; the pulling back triggers the anxious partner’s worst fear and they double down on seeking; the doubling-down feels like more pressure to the avoidant; and on it goes. Both partners are being themselves. Both partners feel the other one is making them miserable. Neither is “wrong”. It’s a structural mismatch, and without awareness of it, it can play out for years, through every love language they can think to speak to each other. Levine and Heller’s popular book Attached calls this the anxious-avoidant trap. (The research is less tidy than the story. In a study of 354 dating couples, relationships between avoidant men and anxious women turned out to be surprisingly stable over three years, even though the partners rated them poorly: Kirkpatrick and Davis, 1994.)
There’s a separate, gentler dynamic when one partner is securely attached and the other isn’t. Secure attachment has a stabilising gravity. In one long-running study that followed people from adolescence into their seventies, being in a relationship went with lower attachment anxiety and avoidance across adulthood (Chopik, Edelstein and Grimm, 2019). (This connects to what researchers call earned security; more on that in a moment.) But the secure partner has to be willing to stay, and the insecure partner has to be willing to do their work. Neither is automatic.
Love Languages sits inside this dynamic, not outside it. The secure partner’s acts of service, words of affirmation, and quality time land more fully because the underlying receiver is able to receive. The same gestures from the same person, with an anxious or avoidant partner who hasn’t yet done attachment work, will often land half-heard.
How attachment shows up on dating apps specifically
The apps amplify everything about your attachment style.
For anxiously attached people, the apps are a slot machine designed by sadists. Every match is a hit of reassurance; every unanswered message is a confirmation of the worst fear; the unpredictable timing of replies is the kind of variable reward schedule that keeps people checking. You can spend an evening checking the app every few minutes and feel worse with every check.
For avoidantly attached people, the apps are a perpetual exit ramp. There is always a next match, always a reason this conversation could be wound down, always plausible deniability for not getting closer. The apps make the strategy frictionless, and reward it, with novelty.
For fearful-avoidant people, the apps offer the worst of both: the pull of pursuit and the easy exit of withdrawal, both available at all times. The oscillation pattern can run faster on apps than it ever could in pre-app dating, because the loop closes in days, not months.
For securely attached people, the apps are merely irritating: a slow inefficient way to find someone you’d actually like to spend time with. They tend to use them sparingly and leave when something works.
If your app behaviour is starting to sound familiar, treat that as a prompt to look closer, not as a diagnosis. These four sketches are this article’s own reading of how the styles play out on apps, not research findings.
The good news: attachment style is stable, but not fixed
The most important thing the research has converged on, and the thing the Instagram-therapy summaries almost always miss, is this: attachment style is stable but not fixed. It is a deep pattern, slow to change, that orients you in default ways. It is not a sentence.
Earned security is a real term from attachment research: it describes adults who come across as secure despite a difficult start. And people do shift over time. In the long-running study mentioned above, attachment anxiety declined on average with age and avoidance fell steadily across the lifespan (Chopik, Edelstein and Grimm, 2019), and a 23-year study found that people who were insecure as infants and secure later went on to have success in their close relationships (Roisman and colleagues, 2002). What researchers can’t yet give you is a recipe. Fraley’s own summary is that the field does not have a strong understanding of what changes a person’s style, so stable relationships, therapy and self-awareness are reasonable bets, not guarantees. The patterns rewrite, slowly.
What probably isn’t enough on its own: hoping you’ll grow out of it. Reading articles about it (sorry). Identifying with the label and then using it as an excuse: “I’m just avoidant, that’s how I am”. The change happens in behaviour, repeatedly, in the moments where the old pattern wants to fire.
What to do with this
A few practical things.
Identify your own style as honestly as you can. Not the version of yourself you want to be. The version that shows up under stress, when a person you like goes quiet for two days. There are decent self-report questionnaires out there. The ECR-R (Fraley, Waller and Brennan, 2000) is one that researchers use to measure attachment anxiety and avoidance, and Levine and Heller’s Attached (2010) is a popular book on the subject. The Confidence Hub’s learning module on attachment is a good starting point for reflection, though it is not a validated test.
Watch for the pattern, not the person. When you find yourself mid-spiral (refreshing the app, drafting and redrafting a message, or alternatively scanning for reasons to fade out), name it. This is my anxious system firing, or this is my avoidant system kicking in. Just naming it can be enough to interrupt the autopilot.
Date toward security. Not in a calculating way, but if you have a choice between the person who generates familiar nervous-system chaos and the person who generates an unfamiliar, slightly boring-feeling steadiness, the second person is often the one your work-version-of-you would choose. Securely attached partners are not flashy. They are the most underrated dating outcome in the world.
Don’t make it your whole personality. Attachment style is a useful map, not the territory. People are more than their attachment patterns. The point of knowing your style is to change your behaviour, not to acquire a new identity to centre conversations around.
The honest caveat
A few things the research is much messier on than the popular versions admit:
- The four-style model is a useful simplification of a continuous reality. People don’t tidily fit into one box; we sit somewhere on the two axes, and our position can vary by partner and context.
- Cultural variation matters. Much of the best-known attachment research, including the figures above, comes from American samples. Large cross-cultural surveys exist (one covered 62 cultural regions), and they find differences between cultures.
- Self-report has known limits. People who score “secure” on a quiz they’re trying hard to look good on aren’t necessarily secure in life.
- Childhood attachment is one input among several. Adult relationships, therapy, trauma, life events: all of these can shift the pattern in either direction.
In other words: take it seriously, but lightly.
What I’d know now
If I could go back to that relationship with the vocabulary I have now, I don’t think I’d love her more clearly. I was already doing that, in several of the dialects Chapman describes, as honestly as I knew how. The problem was never the signal. The problem was that the part of her receiving the signal was running on an older script than I was, and no amount of louder speaking from my side could rewrite it.
What I’d do differently is stop treating her don’t leave me as a verdict on my communication and start treating it as weather from somewhere else entirely. I’d stop trying to reassure the fear away and start trying to help her feel, physically, that the moment she was in was safe: not safer than last moment but safe enough right now. I’d be less hurt by it, because I wouldn’t be reading it as a failure to land the love I was sending. I’d be less exhausted by it, because I wouldn’t be treating each spike as a crisis to solve. And I’d ask her, gently and over time, to do the work on her end, because the one thing that can’t happen, no matter how skilled the partner, is for attachment to rewrite itself in one direction only.
Love Languages was useful. It wasn’t wrong. It just wasn’t the part of the problem that was actually in play.
The map underneath the love languages is the one I wish I’d had.
The takeaway
You are not a victim of your attachment style. You are also not exempt from it. The patterns you keep finding yourself in are patterns, and they have a name, a cause, a structure, and, with attention, a path toward something more workable.
The single most useful thing you can do with this information is notice when the pattern is firing in real time. Once you can see it happening, you have a choice you didn’t have when it was running silently. That choice, made repeatedly, is the whole game.
Chapman’s five languages tell you how to express love. Bowlby, Ainsworth, Hazan and Shaver tell you whether the love is being received. Both matter.
One is about the dialect. The other is about whether the radio is turned on.
This article is part of “The Science Behind It” series, exploring the psychology and neuroscience that shapes how we date, connect, and build lasting relationships.