Twenty statements, four minutes, no email. Scored on satisfaction — and on the three other measures the research uses, because how good a relationship feels turns out not to be the same question as whether anybody stays in it.
Answer about one specific relationship, as it has been over the last few months rather than at its best or its worst. Some statements are worded the other way round on purpose, so read each one before answering.
Scored in your browser. Nothing you answer is sent anywhere or stored.
The four readings this test can return are Satisfied and committed, Committed without being satisfied, Satisfied but not committed, Low on both, each reported alongside separate scores for satisfaction, the quality of your alternatives, how much you have invested, and commitment.
Nearly every relationship satisfaction quiz online measures one thing and reports it as a verdict: a percentage, a colour, a grade. The measurement itself is usually fine — global satisfaction is easy to measure and hard to get wrong, which is why Hendrick’s seven-item Relationship Assessment Scale from 1988 is still in service and why Funk and Rogge could build a more precise version of the same idea two decades later. The problem is what happens next. A satisfaction score answers “how is it going”. Almost nobody arrives at a page like this one asking that. They arrive asking what it means, and usually what to do.
The framework that gets closest to the second question is Caryl Rusbult’s investment model. Its claim is that the intention to stay in a relationship — commitment — rests on three separable things: how satisfying the relationship is, how good the alternatives to it look, and how much has been put into it that could not be taken out. Rusbult, Martz and Agnew published the scale for it in 1998 across three studies, and found the four constructs loading on four independent factors rather than collapsing into one, with the earlier measures predicting later relationship status — persisted versus ended.
Then Le and Agnew pooled 52 studies, 60 independent samples and 11,582 participants. All three bases correlated with commitment: satisfaction most strongly at r = .68 [.67, .69], quality of alternatives at r = −.48 [−.49, −.46], investment at r = .46 [.45, .48]. Together they accounted for close to two-thirds of the variance in commitment, and commitment in turn correlated r = .47 [.43, .50] with actually staying versus leaving.
Read those numbers the way they deserve to be read. Satisfaction is the biggest single piece — the quizzes are not measuring a triviality. It is also nowhere near the whole thing, and the two pieces it leaves out are the two that explain every case where somebody’s score and somebody’s behaviour point in opposite directions. That is the gap this test is trying to fill, and it is why it asks twenty questions instead of seven.
Ask that question on a search engine and you get a list of adjectives — fear of change, low self-esteem, sunk cost. The investment model gives a more specific and more testable answer: because two of the three things holding a relationship together were never about satisfaction in the first place. Poor alternatives and heavy investment keep commitment high while satisfaction falls, and the result is a person who can list everything that is wrong and has no intention of leaving. That is not a contradiction and it is not weakness. It is what the model predicts.
The strongest evidence for it comes from the place where it matters most. Rusbult and Martz analysed intake interviews at a shelter for women who had been abused by their partners. In their words, commitment was greater among the women who “had poorer-quality economic alternatives, were more heavily invested in their relationships (e.g., were married), and experienced lesser dissatisfaction”, and commitment was “strongly linked to stay/leave behavior, significantly distinguishing between women who returned to their partners immediately on leaving the shelter and women who did not”.
Two things follow, and both are worth saying plainly. The first is that “why did they not just leave” is a question with a documented answer, and the answer is largely about alternatives and investment rather than about character. The second is a warning about this very test: the same scores that describe an ordinary long relationship — a lot invested, few appealing alternatives, low satisfaction, high commitment — also describe a dangerous one, and a questionnaire cannot tell the two apart. Only you can, and the distinguishing question is not how satisfied you are. It is whether you are afraid.
It cannot predict. The largest attempt anyone has made pooled 43 longitudinal datasets and 11,196 couples and used machine learning on everything the field measures. Self-reported relationship-specific variables — perceived partner commitment, appreciation, sexual satisfaction, conflict — explained up to 45% of relationship quality at baseline and up to 18% at follow-up. Models for change in satisfaction never accounted for more than about 5% of the variance, from any combination of variables. Nothing sold to you as a forecast is beating that, and most of what is sold to you as a forecast is not even trying.
It cannot see the other person. Everything here is your report about your side. The strongest predictor in that PNAS analysis was how committed a person believed their partner to be — a belief, held by one person, about another — which is a reminder of how much of what we call “the relationship” is actually one person’s reading of it. The obvious next step after this test is not another test. It is asking.
And it has no clinical threshold. Some validated instruments do define one, established on clinical samples; an educational adaptation cannot inherit that, so this page reports bands it calls conventions and says so. Any free quiz that tells you your relationship is “in the distressed range” without naming the instrument and the sample that range came from is making the number up, and a made-up threshold on a subject like this one is not a harmless flourish.
If satisfaction came out high, the useful follow-up is not celebratory. The ordinary maintenance that keeps it there is unglamorous and well documented, and the habits of couples who stay in love goes through what the observational work actually supports. If it came out low, the first question is what kind of low: a relationship that is going badly and a relationship that has gone quiet are different problems, and what predicts that a relationship has real potential is the piece that separates them.
A large share of low satisfaction scores are downstream of how disagreements go rather than of the relationship as a whole. The conflict style test scores pressing and retreating separately and is the natural second reading here; what happens in the hour afterwards is covered in what the repair research supports doing after a fight. If the dissatisfaction is really about not knowing where you stand, that is a different thing again, and the attachment style test plus the difference between love and anxious attachment are aimed squarely at it.
Two traps specific to this subject. The first is the milestone: waiting for a month, a date or a number to tell you what the relationship is, which is the mechanism picked apart in the 3-6-9 rule and the 72-hour rule. The second is the questionnaire itself — including this one — used as a substitute for the conversation; what questions can and cannot reveal about compatibility is about exactly that limit. If you are on the other side of it and trying to work out whether you have already left, the signs you have moved on is the closer piece. And if the score came out low mostly on the closeness items, the two pieces that speak to that directly are what the research says about how people open up to each other and the habits of emotionally mature people in love. More of this under Mindset and Dating.
With short self-report scales. The best known is Hendrick's Relationship Assessment Scale, seven items published in 1988 and still in wide use; the Couples Satisfaction Index, built by Funk and Rogge in 2007 using item response theory, is the more precise modern option and comes in 32-, 16- and 4-item versions. Relationship satisfaction is one of the most heavily measured constructs in the field — a reliability generalization meta-analysis pooled hundreds of studies of these scales alone. What none of them do is tell one person what to do next. They produce a number for a sample, and the interpretation is the part you have to do.
No. It is a 20-item educational adaptation that follows the conceptual structure of the Investment Model Scale, which is validated, and it is written in our own words rather than reproducing the original items. It gives you a reading, not a measurement. Anywhere you see a free online quiz claiming to be a published scale, check whether it names the paper — and then check whether the items match.
Because satisfaction on its own does not explain who stays. In Rusbult's investment model, commitment rests on three things: how satisfied you are, how good your alternatives look, and how much you have tied up in the relationship. A meta-analysis of 52 studies and 11,582 participants found all three associated with commitment — satisfaction most strongly at r = .68, alternatives at r = -.48, investment at r = .46 — and together they accounted for close to two-thirds of the variance in commitment. A test that measures only the first one is measuring the largest piece and calling it the whole thing.
No test can answer that, and this one is not built to. It gives you a description of where things stand from your side, on one day. The largest study of prediction in this field pooled 43 longitudinal datasets and 11,196 couples and found that self-report measures explain a fair amount of current relationship quality and almost nothing about how it changes — never more than about 5% of the variance. What a low score is genuinely useful for is turning something vague into something you can say in a sentence, which is a better starting point for a conversation, with your partner or with a therapist, than a feeling you cannot name.
Yes, and the shape is well documented. A meta-analysis of 165 samples and 165,039 people found that relationship satisfaction declines over roughly the first ten years of a relationship, reaches its low point around year ten, and rises again through year twenty. That is an average trajectory and not a prophecy, but it does mean a middling score at year eight and a middling score at year one are not the same information, and any test that hands you a verdict without asking how long you have been together is ignoring the largest known moderator of the number it just gave you.
Because commitment is not satisfaction. The investment model's answer is that poor alternatives and heavy investment hold people in place independently of how good the relationship feels, and the evidence for that is strongest exactly where it matters most: Rusbult and Martz analysed intake interviews at a shelter for women who had been abused and found commitment was higher among those with poorer economic alternatives and more invested in the relationship, and that commitment distinguished the women who returned to their partners on leaving. Naming the mechanism is not the same as excusing it. It is what makes the answer to "why am I still here" something other than a personal failing.
No. Every question is scored in your browser and nothing is transmitted or saved. There is no account, no email field, and no results URL that could be shared by accident. Close the tab and the answers are gone. On a test that asks whether you are thinking about ending your relationship, that is not a footnote.
How we choose and read sources is set out in our editorial policy. If something here is wrong, tell us and it goes on the corrections page.