Do We Know Who We Want?

Men never know what they want, but they will go to any lengths to get it.

— Don Marquis

A week before a speed dating event, a participant writes that intelligence and a sense of humour matter most to him. The evening goes differently. He gives his highest rating to a person whose intelligence had no time to show itself in four minutes, and whose one joke was not particularly good. You could decide he lied. But he probably answered in good faith. The questionnaire simply asked him to do something people are worse at than they assume: predict their own reaction to a person they have not met yet.

The ideal as autobiography

A list of requirements looks more like an autobiography of taste than a map of future choice. Into it a person packs past connections that worked and past connections that failed, family scripts, social expectations and a desired image of himself. Calling intelligence important is pleasanter than admitting that a voice was decisive; stressing reliability is more respectable than admitting a pull towards the unpredictable. Different things end up on one list. Attitudes to children, where to live, monogamy and money set the terms of a joint project. Height, profession and style describe the presumed sources of first interest. By mixing them, the questionnaire gets a tidy answer to several different questions.

Some traits a person finds attractive in the abstract: height, profession, style, age. They can affect the first round of filtering. And then there is the reaction to a specific person. It appears only in interaction and owes no respect to the list drawn up in advance. When all of this is packed into the single question “what kind of partner do you want?”, the answer comes out tidier than reality.

The meeting as a test of the questionnaire

Before the event participants ranked the qualities of an ideal partner; after each short conversation they rated specific people. Individual lists predicted attraction poorly. On paper men rated looks higher and women rated income and prospects higher; at the tables the difference in how much these qualities mattered to men and to women turned out to be noticeably smaller. The stated differences, meanwhile, turn up regularly across many countries. They may reflect biological dispositions, available roles and the language a culture teaches people to use when explaining choice, all at once. What helps separate these sources is how much the effect varies between conditions.

Preferences about age and status differ between societies, but the link between sex differences in preferences and gender equality follows no simple law. In a sample of 45 countries, equality consistently predicted a smaller age gap between partners; for preferences about looks and financial prospects no equally stable link was found. This supports neither a rigid biological constant nor the simple formula that equality automatically erases differences. Reproductive risk, law, income and contraception change the constraints on choice, and culture supplies the language people use to explain the result.

A preference without a fixed weight

The questionnaire invites you to rate qualities as if each carried a permanent coefficient: looks, eight; reliability, nine; income, five. It then assumes you apply those weights to any set of traits you encounter. In real choice the weight shifts along with the range of options. Where educational level is similar, education almost stops distinguishing participants. A tenfold spread in income, by contrast, makes income visible. Intelligence combined with arrogance and unavailability can no longer be rated apart from the package it arrived in.

Qualities most people consider desirable do usually raise ratings. The weak point of the list is the unique part of choice: why the person who named humour as the main thing leaves with someone other than the funniest participant of the evening. A stated priority shows the direction of attention and an idea of what a good partner is; it reconstructs the exact reaction to a specific package of traits badly.

There is also a difference between importance and scarcity. Clean water is extremely important and costs almost nothing where there is plenty of it. A rare way of talking may not appear on any list of requirements, yet become decisive if someone has met it for the first time in years. People often name as important the traits they are used to thinking about, and underrate the ones whose presence or absence only becomes visible next to a specific person.

Speed dating tested a narrow ability: to reconstruct future, unique attraction from a ranking of qualities made in advance. It turned out to be weak. That does not turn ideals into empty noise. In a thirteen-year longitudinal study, initial preferences were linked to which traits participants later ascribed to their partners. But a closer match between the actual partner and the original ideal gave no stable prediction of relationship quality, of whether the relationship survived, or of who initiated the break-up; the conclusion depended on how the match was counted, and current ideals partly adjusted themselves to the person already chosen. A list sets the direction of a search better than it predicts a unique reaction to a specific package of traits.

Hard filter, soft taste

The mistake is easy to correct the wrong way: declare all requirements meaningless and “give everyone a chance”. That is a bad model too. A strong spark does not reconcile a refusal to have children with a life project built around parenthood. A plan to emigrate conflicts with attachment to a place, and different forms of union call for an agreement that attraction will not draft on the couple’s behalf. Lists are useful when they contain the constraints of a shared life. They are dangerous when they turn a presumed source of attraction into a mandatory pass.

You can see the difference in the question. “Could I live with someone who…?” is about compatibility. “Could I be attracted to someone who…?” is a forecast of feeling. In the second case confidence noticeably exceeds accuracy. A hard filter on height can remove the person you would have been attracted to. A filter on attitudes to children removes a likely future conflict. In the interface both look the same — two sliders — although psychologically they do different jobs.

The questionnaire lays qualities out one by one. The meeting presents them as a package. A sense of humour can be attractive in one person and tiring in another. Confidence without warmth is received differently than confidence with it. High status can signal competence, or a busy schedule, or power, or an incompatible way of living. A trait changes meaning in combination with others. Beyond that, part of attraction arises in the dynamic: how the person answers you, whether they notice your tone, whether they change the subject after an awkward moment, whether talking gets easier next to them. These properties cannot be assigned to one participant: they arise in the interaction.

A list of ideals often loses to an actual meeting. The person here is not necessarily inconsistent. A model of traits is trying to predict something that partly lives between people. It helps to split the list in two. The first part holds the conditions without which the shared project does not work: children, place, the format of the relationship, basic values, way of life. Those are worth discussing and taking into account. The second holds traits you assume ought to produce attraction. That part is better handled as a hypothesis than as a specification. It can guide the search, but it should not close it before the meeting.

Life first, portrait second

The marketing question “what is the person actually buying?” has a transferable part, once you remove the buyer and the product. People describe a future partner in detail and much more rarely describe the future life they are looking for a partner for. “Smart, tall, successful” is a portrait. “I want children, I don’t want to move every two years, I expect to share care for our parents and I am not giving up my profession” is the design of a project.

A portrait is easy to optimise for prestige, past attractions and generally accepted appeal. A project forces you to name the terms: what the two of you have to build together, what stays separate, which decisions are irreversible and what happens when you disagree. It does not tell you who will produce attraction, but it does a better job of exposing the conflicts attraction will later have to pay for.

So the question “who do I want?” is worth supplementing with another: “what repeating day am I trying to build?” The answer does not replace a meeting and does not turn a person into a job description. It only separates the requirements of a shared life from a decorative portrait of its presumed participant.

We usually know who would be hard to live with better than we know who we are capable of falling in love with. A questionnaire is useful where it stops pretending to be a heart.

Main sources

Eastwick, P. W., & Finkel, E. J. (2008). Sex differences in mate preferences revisited: Do people know what they initially desire in a romantic partner? Journal of Personality and Social Psychology, 94(2), 245–264; Eastwick, P. W., Luchies, L. B., Finkel, E. J., & Hunt, L. L. (2014). The predictive validity of ideal partner preferences: A review and meta-analysis. Psychological Bulletin, 140(3), 623–665.

Walter, K. V., Conroy-Beam, D., Buss, D. M., et al. (2020). Sex differences in mate preferences across 45 countries. Psychological Science, 31(4), 408–423; Eagly, A. H., & Wood, W. (1999). The origins of sex differences in human behavior. American Psychologist, 54(6), 408–423.

Driebe, J. C., Stern, J., Penke, L., & Gerlach, T. M. (2024). Probing the predictive validity of ideal partner preferences for future partner traits and relationship outcomes across 13 years. European Journal of Personality, 38(4), 707–723.