10 Toxic Relationships to Avoid, Backed by Science

"Toxic" is overused online, often stretched to cover any uncomfortable dynamic. The ten patterns below are different: each has been measured in longitudinal studies against wellbeing outcomes, and each carries a specific mechanism worth knowing. Understanding them does not mean labelling people. It means seeing clearly which dynamics erode happiness over time — and doing so with enough precision to decide what, if anything, to do about them.

A grounding fact before the list: the quality of close relationships is the single most robust predictor of both happiness and physical health in the research literature. The Harvard Study of Adult Development, tracking hundreds of participants for more than eighty years, consistently finds that relationship quality in midlife predicts health and happiness in old age more reliably than cholesterol, wealth, or any other variable measured. That is the stake. The ten patterns below are, in different ways, attacks on that predictor.

1. The chronic critic

John Gottman's four decades of research at the University of Washington identifies criticism — as distinct from complaint — as one of the four strongest predictors of relationship dissolution. The distinction matters: a complaint is about a specific behaviour ("you didn't call when you said you would"). Criticism attacks character ("you're always thoughtless and selfish"). The mechanism is cumulative: character attacks shift the perceived ratio of positive to negative interactions below the 5-to-1 threshold that characterises stable, satisfying relationships in Gottman's data. Below that threshold, negative interactions outweigh positive ones in perceived weight, and the relationship's emotional balance sheet tips into deficit.

Chronic criticism also produces a physiological stress response in the recipient that, over months and years, is associated with elevated cortisol, immune suppression, and increased risk of depression. The chronic critic rarely intends harm; they have usually normalised a communication style they learned early. That does not make the pattern less damaging to the person receiving it.

2. The stonewaller

Stonewalling — emotional withdrawal during conflict: shutting down, going silent, physically leaving — is the second of Gottman's "Four Horsemen," the four interaction patterns most predictive of relationship breakdown. The mechanism is specific: when one partner stonewalls, the other partner's physiological stress response escalates because the emotional feedback loop is severed. There is no repair signal, no response to send a message to, no way to gauge where the relationship stands. Over months, this pattern is strongly associated with depression and reduced life satisfaction in the partner who is stonewalled, and — counterintuitively — in the partner who stonewalles, who typically does so to regulate their own overwhelm but in the process prevents the repair that would actually relieve it.

3. The scoreboard-keeper

Tracking favours owed and grievances filed — researchers call this "exchange orientation" — correlates inversely with relationship satisfaction in both friendships and romantic partnerships. The mechanism is attentional: a person focused on what is owed to them cannot simultaneously focus on what they are contributing and receiving freely. Intimacy requires at least partial amnesia about who did what last; the moment every interaction is evaluated for reciprocity, the relationship shifts from connection to transaction, and the emotional quality changes accordingly. Studies on married couples find exchange-oriented partners report lower satisfaction and higher resentment even when the objective balance of contribution is equitable.

4. The person who can never be wrong

Defensive rigidity — the inability to accept influence, to acknowledge error, to update a position in response to evidence — is the third of Gottman's Four Horsemen and predicts divorce with high accuracy in his data. In friendships, it predicts gradual drift. The mechanism operates at two levels: practically, nothing ever gets resolved, because resolution requires at least one party to shift; and relationally, the message communicated is "I am not a person you can reach." Over time, this message extinguishes the other person's attempts at real communication and replaces them with managed distance or escalation.

The diagnostic test: a relationship where one person has never been wrong, never apologised sincerely, and never updated their account of a conflict based on what the other person says is one where defensive rigidity is operating. This is not a character flaw to argue with. It is a pattern to recognise clearly and decide about.

5. The drainer

One-way emotional support relationships — where one person consistently listens, empathises, and helps, and the other person never inquires, deflects when the subject is turned, or is consistently unavailable — show up in happiness research as a chronic low-level stressor. A modest amount of asymmetry is normal in any relationship across any given period; life circumstances create temporary imbalances. A pattern of it across years is different. Research finds this pattern correlates with measurable increases in cortisol, emotional exhaustion, and reduced relationship satisfaction for the over-giving partner — even when no other toxic pattern is present. The cost is the invisible price paid for a connection that operates in only one direction.

6. The gaslighter

Chronic reality-denial — "that didn't happen," "you're imagining it," "you're too sensitive," "you always misremember things" — is associated with clinically significant anxiety disorders and long-term erosion of self-trust in the people on the receiving end. Research on psychological abuse documents that gaslighting functions by targeting the victim's perceptual reliability: once a person begins to doubt whether their own memory and interpretation of events is accurate, their capacity to accurately assess their situation — including the relationship itself — is impaired. This one is serious. The science backs taking it seriously, and the appropriate response is not more deliberate self-reflection about whether the doubts might be valid.

7. The contempt-holder

Contempt — eye-rolling, mockery, sneering, sarcasm deployed with real edge — is Gottman's single strongest predictor of relationship dissolution, more predictive than any of the other horsemen. The mechanism is clear: contempt communicates that one partner regards the other as beneath them, as inadequate or ridiculous. It converts the other person from a subject of genuine regard into an object of disdain. Gottman's data finds that contempt predicts not only divorce but deteriorating physical health in the partner on the receiving end, including reduced immune function. No other single communication pattern in the research literature carries this combination of consequences.

8. The isolator

Relationships that gradually cut you off from other relationships — monitoring calls, resenting outside friendships, creating tension around contact with family, framing your existing social network as a threat — produce some of the worst wellbeing outcomes in the relationship literature. The mechanism operates through two pathways. First, your social network is independently protective for physical and mental health; an isolating partner removes that protection. Second, a person without external social contact loses the calibrating effect of other people's perspectives — making it harder to accurately assess the quality of the primary relationship. Isolation is therefore both directly harmful and self-reinforcing: it worsens wellbeing while simultaneously reducing the conditions under which a person can recognise what is happening.

The U.S. Surgeon General's 2023 advisory on loneliness cites social isolation as a mortality risk comparable to smoking fifteen cigarettes a day. A partner who systematically reduces your social connections is, among other things, a health risk.

9. The intermittent reinforcer

Relationships characterised by alternating warmth and withdrawal — periods of genuine closeness followed by cold distance, criticism, or contempt — are psychologically similar to a variable-ratio reinforcement schedule: the same structure that makes slot machines addictive. Research on these dynamics finds they are harder to leave than consistently hostile relationships, partly because the periods of warmth feel more intense by contrast, and partly because the person experiencing them is trained, neurologically, to work hard for unpredictable rewards. The elevated anxiety scores associated with this pattern are consistent with a stress system that is chronically activated by unpredictability rather than by any single severe event.

10. The quietly unequal partner

Sustained imbalance in effort, care, and emotional labour — the partner who consistently plans, anticipates, manages, and absorbs while the other partner is available but passive — does not feel toxic day-to-day. There is no specific incident to point to. The cost accumulates over years rather than episodes. Research on married couples finds this pattern is one of the more robust correlates of lower life satisfaction for the over-contributing partner, even in the absence of any other harmful dynamic. The mechanism is partly exhaustion and partly a kind of invisible social injustice: the effort is real, the invisibility is real, and the accumulated effect on the person carrying it is real even when neither partner explicitly recognises what is happening.

What to do with this information

Some of these patterns are fixable with a direct, honest conversation, especially when both people can hear it as information about a dynamic rather than an attack on character. Some are not — particularly those involving contempt, gaslighting, and isolation, which tend to be deeply embedded and resistant to change through conversation alone. The science is most useful here for seeing what is happening with enough clarity to decide deliberately, rather than continuing in something whose costs are felt but not named.

If any of these patterns are causing significant distress, working with a therapist or couples counsellor provides access to structured, evidence-based tools (including Gottman Method couples therapy and Emotionally Focused Therapy, both with strong research bases) that go beyond what self-directed recognition alone can accomplish. Recognising a pattern is the first step. What you do with that recognition is a human decision, not a scientific one.

For a complementary picture of what healthy relationship investment looks like — the deliberate practices that build the kind of relationship quality the research identifies as protective — and for broader steps toward building a happier life beyond any single relationship, the evidence points in consistent directions.

For more on what happiness research reliably tells us about social connection and why it occupies such a central position in the evidence, the picture is both clearer and more urgent than most casual summaries capture.