Why most relationship advice does not generalise
Relationship advice contradicts itself because most of it is a local finding stated as a universal rule.
A pattern observed in one sample, in one decade, in one country, becomes "couples should." Two such rules from two samples can both be locally true and still contradict each other, and neither is presented with the uncertainty it earned.
The most useful corrective is a study almost nobody quotes. Joel and colleagues (2020) applied machine learning across many longitudinal datasets to find what actually predicts relationship quality. Two findings, and the second is the important one:
Relationship-specific variables predicted better than individual-difference variables. How you feel about this relationship beats facts about either person.
Most of the variance remained unexplained. Even with good data, many variables and modern methods, the measurable factors together accounted for a minority of the difference between relationships.
That second finding is the one that should discipline everything else. It is not a reason to ignore research; it is a reason to distrust confident rules, including the ones that sound most authoritative.
Which makes the useful posture clear. Advice is a hypothesis about your relationship, not a finding about it. Something to test, not something you are failing at.
The four ways advice overreaches
The sample was not you. A great deal of psychology research draws on undergraduates, or on couples already in therapy, or on volunteers who differ systematically from people who do not volunteer. A finding in couples seeking help is not automatically a finding about couples.
The situation was not yours. A lab task lasting 45 minutes, a questionnaire completed once, a two-week diary study. Each supports a claim about that situation.
A prediction became a prescription. Being able to classify which couples separate is not the same as knowing what would prevent it. Prediction and intervention are different claims and the second is much harder to support.
An association became a cause. "Couples who do X report Y" is compatible with X causing Y, Y causing X, and the sort of couple who does X differing in ways nobody measured.
What survives better
Three things hold up better than the headline claims, and none is a rule you can fail at:
Responsiveness. The framing associated with Reis and Shaver: intimacy is a process where one person discloses and the other responds, and what matters is whether the discloser feels understood, validated and cared for. The mechanism is the response, which relocates the work from technique to attention.
Capitalisation. Gable and colleagues on how partners respond to each other's good news, and the finding that an enthusiastic engaged response is associated with better outcomes than a passive or dismissive one. Most advice concentrates on conflict; ordinary good days turn out to be a place where something happens.
Structured mutual disclosure generates closeness. Aron and colleagues, 1997, in strangers, in a lab, measured immediately. A genuine finding about a specific procedure, and the thing that gets stretched furthest in marketing. Does asking a daily question actually help?
What these share: they describe processes rather than prescribing rules, and they are about what happens between two people rather than about what type either of them is. Why quizzes do not change behaviour.
Reading any claim about couples
Four questions, and they work on this page:
- Who was studied?
- Is this an association or an intervention?
- What is the effect size, and compared with what?
- Could it be wrong? Advice that cannot be falsified is not telling you anything, and "communicate better" cannot be falsified.
Most popular relationship claims fail at question two. What this category claims versus what it can evidence.
What this means for anything we sell
It rules out the sentences that would sell best. Nearvo does not claim to improve relationships, because that claim is not available to any consumer app in this category and pretending otherwise would contradict this page.
What it claims is a description of the product: one question a day, both answer, each sees the other's answer only after replying. Checkable by using it.
And the practical version of everything above: treat any suggestion, including ours, as something to try for a fortnight and observe, rather than as a standard you are being measured against. A relationship is an n of one, and most of the variance is unexplained.
Why couples drift apart without anything going wrong. · The full comparison of every couples app.
More in this section
About Nearvo
Nearvo is an iOS app from Baker Ventures LLC built around a daily question for couples: one prompt, both partners answer, and you see each other's answer only after you have both replied. One partner subscribes and both unlock, because a paywall that asks two people to buy separately is how most couples apps quietly fail. The app is in development.
It is designed for the ordinary version of the problem, which is not a crisis: two people who live together, like each other, and have stopped asking anything that is not logistics. Nearvo is not therapy and does not diagnose a relationship. The question sets on this site are free, need no account, and the research pages cite the studies directly, including where the evidence is weaker than the popular version of it suggests.
Questions and answers
Why does relationship advice contradict itself so often?
Because most of it is derived from small or unrepresentative samples and then stated as a universal rule. Two rules from two different samples can both be locally true and still contradict each other, and neither is being described with the uncertainty it earned.
Is there research on what predicts relationship quality?
Yes. Joel and colleagues in 2020 applied machine learning across many datasets and found relationship-specific variables predicted relationship quality better than individual-difference variables, while most of the variance remained unexplained even with good data and methods.
What does "most of the variance is unexplained" mean in practice?
That the measurable factors, taken together, account for a minority of the difference between relationships. It is a caution against confident rules, and it is the finding that never appears in coverage of relationship research.
Are the famous relationship findings wrong?
Not necessarily wrong, and frequently narrower than their popular versions. Some have been substantially debated on methodological grounds, and some are simply applied far beyond the population and situation they were studied in.
How should I read advice about my relationship?
Ask who it was derived from, whether it describes an association or an intervention, and whether it would be falsifiable. Advice that cannot be wrong is not telling you anything.
Does this mean advice is useless?
No. It means advice is a hypothesis about your relationship rather than a finding about it, and the useful move is to treat a suggestion as something to test rather than as something you are failing at.
Nearvo. “Why most relationship advice does not generalise.” Baker Ventures LLC, September 7, 2026. https://nearvo.bakerventuresstudio.com/research/why-most-relationship-advice-does-not-generalise/