Toxic friends: what the research actually shows, and the type nobody warns you about

Search “signs of a toxic friend” and you’ll get some version of the same list everywhere: they criticize you, they compete with you, they only call when they need something. All true patterns. But twenty years of research on friendship and physical health points to a different, less obvious category doing more damage to your body than any of that: not the friend who’s clearly unkind, but the one you can’t decide how you feel about.

Where the usual “toxic friend” advice gets it wrong

Most toxic-friend content assumes toxicity is obvious: a friend who is consistently negative, draining, or unkind. That’s a real category, and it’s worth naming. But it quietly skips over the friendships that don’t sort neatly into “good” or “bad,” the ones with a friend who is genuinely warm sometimes and genuinely hard other times. Researchers call this an ambivalent tie, and it turns out to be its own distinct category, not a milder version of a toxic one.

What the research actually found

Psychologist Bert Uchino and colleagues were among the first to test this directly. A 2001 study in the Journal of Behavioral Medicine sorted 133 young and older adults’ social ties into positive, negative, and ambivalent categories, then measured cardiovascular stress responses. The number of ambivalent ties in a person’s network predicted depression and heart-rate stress reactivity on its own, even after accounting for how many purely positive or purely negative ties that person also had. Ambivalent relationships weren’t just diluted positive ones. They behaved like their own thing.

Julianne Holt-Lunstad and Uchino followed up directly on friendship specifically. In a 2007 study in Annals of Behavioral Medicine, 107 adults and a same-sex friend were asked to discuss a positive and a negative life event while researchers tracked blood pressure, heart rate, and vagal tone (a marker of how well the nervous system recovers from stress). The sharpest systolic blood pressure spike came when people discussed a negative event with a friend they felt ambivalent about, compared with a friend they rated as supportive. People who had brought an ambivalent friend also showed higher resting heart rate and lower vagal tone than those who had brought a supportive one. Their conclusion, in their own words, was that people “may not be able to fully relax in the presence of ambivalent friends and may not benefit from support during stress.”

By 2019, Holt-Lunstad and Uchino had formalized this into a broader theoretical framework in Perspectives on Psychological Science, proposing that ambivalent relationships carry their own distinct pathway to poorer health, separate from the well-established harm of purely negative ties and the well-established benefit of purely positive ones.

Why the confusing friend is harder to spot, and harder to leave

A purely negative friendship is, in a strange way, the easy case. Someone is consistently unkind, and most people eventually act on that. An ambivalent friend doesn’t give you that clean signal. They’ve made you laugh until you couldn’t breathe, shown up when it counted, and genuinely understood something about you that few other people do, which is exactly why walking away feels disproportionate, even when the pattern of feeling drained afterward keeps repeating. Your body, per Holt-Lunstad and Uchino’s own framing, doesn’t get the relief of sorting them into “safe” or “threat.” It stays on alert in both directions at once, which is a more exhausting place to sit than either extreme.

What this actually looks like

Reframed through ambivalence rather than the usual checklist, the pattern tends to show up as: leaving a hangout feeling both closer to someone and slightly worse, unable to explain why. Genuinely looking forward to seeing them and quietly dreading it, at the same time. Defending them to other people more often than the friendship actually seems to warrant, as if the good moments need constant restating to stay believable. And noticing, if you're honest, that you feel calmer for a day or two after time apart from them, then miss them anyway.

What actually helps

Name the specific pattern, not the vague feeling. “I feel worse for the rest of the day after we hang out, even on the good visits” is something you can track. “This friendship feels off” isn’t.

You don’t have to choose between keeping them and cutting them off completely. For a tie you can’t or don’t want to fully step away from, a coworker, a roommate, someone tied to your friend group, Holt-Lunstad’s own recommendation is to limit contact and lower your expectations of what the relationship provides, rather than either forcing it into being purely positive or ending it outright.

Let the real warmth be real, and the real cost be real, at the same time. Ambivalence isn’t a puzzle to resolve into either love or dislike. Both things being true is the actual, researched shape of the relationship, and naming it that way tends to be more useful than talking yourself into a cleaner story.

If this pattern shows up across more than one friendship, that’s a bigger, more worthwhile question than any single relationship, and NexSpark’s Magnetism track was built for working through exactly that kind of pattern.

Common questions

What makes a friend toxic?
Most popular lists define it as obvious negativity: criticism, competitiveness, one-sided support, drama. That’s real, but friendship-and-health research points to a different, less obvious category doing more damage: ambivalent friends, the ones who are genuinely warm sometimes and draining other times, so you can never fully write them off or relax into them.

Are the obviously mean friends actually the most harmful ones?
The cardiovascular research suggests they may not be. Uchino’s 2001 study found the number of ambivalent ties in a person’s network was a distinct predictor of stress reactivity, separate from purely negative ties, and a 2007 study by Julianne Holt-Lunstad and Bert Uchino found people showed their sharpest blood-pressure spike while discussing a hard topic with an ambivalent friend rather than a supportive one. A clearly bad friendship is easier to write off; a mixed one keeps your body on alert because it can’t settle on treating the person as safe or as a threat.

What is an ambivalent friendship?
Researchers use it to describe a relationship that carries real positive and real negative qualities at the same time, rather than sitting clearly on one side. It's not the same as a purely negative friendship: an ambivalent friend has genuinely made you laugh, shown up for you, or understood you, which is exactly why they’re harder to identify as a problem and harder to walk away from than someone who is simply unkind.

How do you deal with a toxic or ambivalent friend?
Name the specific pattern rather than the vague feeling (what actually happens, not just that it feels bad), reduce contact frequency rather than forcing an all-or-nothing cutoff when the tie has real value, and be honest with yourself about whether you're staying out of genuine connection or out of history and guilt. Holt-Lunstad's own recommendation, for ties you can't fully step away from, like a coworker or roommate, is to limit contact and lower your expectations of the relationship rather than trying to force it into being purely positive.

Sources

• Bert N. Uchino, Julianne Holt-Lunstad, and colleagues, Heterogeneity in the social networks of young and older adults: prediction of mental health and cardiovascular reactivity during acute stress, Journal of Behavioral Medicine 24(4), 361–382, 2001.
• Julianne Holt-Lunstad, Bert N. Uchino, Timothy W. Smith, and Angela Hicks, On the importance of relationship quality: the impact of ambivalence in friendships on cardiovascular functioning, Annals of Behavioral Medicine 33(3), 278–290, 2007.
• Julianne Holt-Lunstad & Bert N. Uchino, Social Ambivalence and Disease (SAD): A Theoretical Model Aimed at Understanding the Health Implications of Ambivalent Relationships, Perspectives on Psychological Science, 2019.

Figures are rounded as published. See each source for full methodology. Journalists and writers are welcome to cite this page, just link back so others can find the sources too.