This article develops a process-ontological account of sexual partner choice, arguing that its distinctive feature lies not in mere unpredictability or incomplete information but in a structural condition termed interrecursive indeterminacy. Unlike most domains of human choice, the object of sexual-romantic evaluation is itself another evaluator engaged in a parallel process of valuation, such that each party's regard for the other becomes an input that can transform the very evaluation being made. Drawing on Helmuth Plessner's philosophical anthropology, particularly his concepts of excentric positionality, the "unfixed animal," and the dignity of reserve, the article distinguishes erotic valuation, partner formation, and institutional stabilization as three recursively distinct processes that conventional "choice" models wrongly compress into a single decisional event. The argument proceeds through several interconnected claims. First, empirical research on mate preferences, assortative mating, and speed-dating demonstrates that stated ideals predict actual attraction only weakly, and that population-level regularities cannot recover the generative history of individual cases. Second, the five mediations through which humans value can diverge in ways that no single-currency model of preference can represent, explaining the familiar experience of someone being "wrong on paper but right in person." Third, dating platforms and contemporary romantic ideology do not eliminate determination but relocate it: from kinship institutions into individual self-recursivity and symbolic sedimentation, creating an unprecedented expansion of the symbolic field alongside an unprecedented removal of institutional buffers that once absorbed recursive load. My thesis is that sexual partner choice reveals the doubled condition of Plessner's unfixed animal: two beings, neither fully determined, each capable of being changed by the other's regard, nonetheless construct between them one of the most durable structures either will inhabit, and remain, even inside that structure, never quite finished with each other. The relationship is not discovered but generated through recursive encounters whose outcome no amount of prior information could specify in advance, because the object itself has no determinate identity independent of the process that generates it.
Introduction: Three Kinds of Undetermined
This article explores sexual partner choice in humans. The German proverb “die Liebe fällt, wohin sie will“ means that love falls wherever it wants. This this is the key argument: choosing a sexual partner is an utterly undetermined event. It is ontologically uncertain because it is one of the most intensely interrecursive processes in the human mesocosm.
A newborn's future is unpredictable in almost every respect. No one at the moment of a child's birth can specify what work that child will do, which city will hold their final address, which illnesses will eventually claim them, or which of the thousands of people they will meet will still be present at their funeral. Sexual partner choice is often introduced, in both popular and scholarly writing, through some version of this same observation: that the identity of a future spouse or lover is, at birth, entirely unspecified. It is worth being honest, at the outset, about how little that observation actually establishes. Occupation, city, illness, and friendship are equally unspecified at birth, and equally unpredictable in their particulars, without anyone supposing that career choice or the acquisition of friends poses a distinctive anthropological puzzle. If sexual partner choice is going to earn a claim to special theoretical attention, the claim needs to rest on more than the generic unpredictability every human future shares.
Three distinguishable senses of "undetermined" are usually compressed into one whenever this topic is discussed, and separating them is the first task of this essay.
The first is ontogenetic openness: the plain fact that the concrete identity of a future partner remains unspecified unusually late in development, later and more completely than almost any other consequential relationship a person will form. A child typically has some early contact with the social world of future colleagues, future neighborhoods, and future professional milieux, mediated by school, class position, and family network. The identity of a future spouse is frequently untouched by any of this. The person may not yet have been born, may grow up on a different continent, and may belong to a social world the family has no contact with at all. This is a real and interesting fact, but by itself it is only a fact about timing: a claim that determination, for this particular relationship, arrives unusually late rather than never.
The second, stronger sense is selection underdetermination. Even granting extensive information about a person's traits, stated preferences, and the institutional constraints acting on them (family expectations, religious rules, the local dating market, platform search filters), this information does not uniquely specify who, among an eligible field, will become significant. The empirical literature reviewed below documents this underdetermination with real precision: ideal preferences predict actual attraction only weakly and conditionally, institutional filters narrow a field without collapsing it to a single answer, and even detailed compatibility algorithms leave enormous variance unexplained. This is stronger than ontogenetic openness, because it holds even when a great deal is already known. But it remains, in an important sense, an epistemic claim: it says that available information falls short of a full explanation, which leaves open the possibility that some further variable (genetic, developmental, situational) could in principle close the gap.
The third sense is the strongest, and it is the one this essay will argue actually deserves to be called the distinctive feature of sexual partner choice: interrecursive indeterminacy. On this account, the underdetermination is not simply a matter of missing information about an otherwise fixed outcome. The process by which one person, A, evaluates another, B, can change B; B's resulting evaluation of A can change A; and each person's apprehension of how the other is valuing them becomes, in turn, an input into their own subsequent valuation. The relevant future state (whether this becomes a couple, and what kind) is not merely unknown. It is partly generated by the very process attempting to determine it. No increase in information about either party's fixed prior attributes could, even in principle, settle the question in advance, because the object being predicted does not yet exist independently of the interaction that will produce it.
The first two senses prepare the ground for this claim; they do not establish it. What establishes it is a specific structural feature of sexual and romantic evaluation with no straightforward analogue in most other domains of choice: the object being evaluated is itself another evaluator, engaged in the same kind of process, aimed reflexively back at the first person. Helmuth Plessner's philosophical anthropology supplies this structure's sharpest available name. Drawing on Nietzsche's description of the human being as das noch nicht festgestellte Tier (the animal not yet fixed or determined), Plessner argued that human beings possess excentric positionality: they live from an embodied center while simultaneously being able to take up a position toward that center, toward their own body, conduct, and relationships, such that nothing about being human simply coincides with a naturally given position. Plessner developed this claim primarily with reference to a single subject's relation to itself. Sexual partner choice reveals what happens when Plessner's problem doubles: when the object of one excentric being's indeterminacy is another excentric being, whose own self-relating evaluation is recursively coupled to the first person's.
That doubling is the argument of this essay, and it is worth stating as a single thesis before the evidence and the theoretical apparatus are laid out in full. Sexual partner formation is unusually undetermined not because humans lack preferences, or because attraction is random, but because the person being valued is another recursively valuing being whose response enters into, and can transform, the valuation itself. Everything that follows (the empirical literature on preference and prediction, the machinery of kinship and marriage, the architecture of dating platforms, the statistics of assortative mating, the politics of who gets to keep a relational future open) is best read as evidence for where this doubled indeterminacy shows up, where it is successfully managed, and where it refuses management altogether.
Against "Choice": Separating Valuation, Formation, and Stabilization
Before going further, the vocabulary this essay has been using needs to be treated with some suspicion. "Partner choice," "mate choice," and "partner selection" all borrow, more or less consciously, from a decision-theoretic picture: an agent surveys a set of options and picks one. The picture is not simply imprecise; it actively obscures the phenomenon it is meant to describe, because it compresses several recursively distinct processes into a single decisional event.
Consider what actually happens across an ordinary human life. A person can feel erotic desire for someone they never approach. They can form a relationship (spend time with, become attached to, build a shared life with) someone toward whom their initial attraction was moderate at best. They can marry a third person entirely, for reasons that include but are not exhausted by desire: family approval, timing, practical compatibility, the accumulated momentum of years together. And they can, within that marriage, continue to feel real erotic pull toward a fourth person they have never touched. None of this is exotic or pathological; it is close to the modal human experience of adult sexual and romantic life. Yet a single vocabulary of "choice" invites the assumption that these are all variants of one operation, differing only in degree: as though desiring A, partnering with B, marrying C, and wanting D were four data points on the same preference curve, when they are in fact four outputs of at least three distinguishable processes.
This essay will use three terms with some care. Erotic valuation names the felt pull toward a particular person (attraction, desire, fascination), which can arise, intensify, or dissolve largely independent of any decision. Partner formation names the dyadic process by which two people actually build an ongoing relationship: the escalating exchange of attention, disclosure, and reciprocal responsiveness that this essay will later analyze as interrecursive. Institutional stabilization names the symbolic and legal acts (a declared relationship, an engagement, a marriage, a formally negotiated non-monogamous agreement) that fix a partnership's public status and its consequences for property, kinship, and recognition, independent of what either party currently feels.
The failure of these three processes to coincide is not a nuisance for a theory of partner choice, something to be bracketed as noise around the main signal. It is some of the best evidence available that partner choice, understood as decisionistic selection, is the wrong model. If erotic valuation, dyadic formation, and institutional stabilization were really stages of one underlying preference-satisfaction process, their divergence would be a standing anomaly requiring explanation case by case. On the account developed here, their divergence is exactly what should be expected, because each belongs to a different mediation and operates at a different level of recursivity: valuation can remain largely embodied and unreflective (what will later be called L1–L2), formation is thoroughly interrecursive and unfolds over time, and stabilization is an act of explicit multisymbolic classification (L4) that can be performed in a single afternoon at a courthouse. A theory adequate to the phenomenon has to be built to expect the gaps between these three, not to treat them as failures of an otherwise unified choice.
Plessner's Anthropological Laws in the Anatomy of Courtship
Plessner's philosophical anthropology does more work for this argument than a single borrowed phrase about the "unfixed animal." In Levels of Organic Life and the Human, he derives from excentric positionality three so-called anthropological laws, and each finds an unusually concrete instantiation in courtship.
The first is the law of natural artificiality. Because the human being does not simply coincide with its own embodied center, it cannot simply act out of that center the way other organisms enact their environmentally fitted behavior. To realize what it "naturally" is, the human being must supplement itself artificially (with tools, institutions, symbolic forms) not as an addition to an otherwise complete nature but as the very means by which an incomplete nature becomes livable. Courtship is naturally artificial in exactly this sense. No courtship is ever a raw discharge of instinct meeting raw instinct; even the most apparently spontaneous flirtation is staged through learned gesture, cultural script, calibrated timing, and an acquired sense of what a glance held a half-second too long communicates. The elaborate apparatus of courtship (the asking-out, the date, the gift, the introduction to friends, the meeting of parents) is not an external constraint layered onto a natural mating process that would otherwise proceed unaided. It is the artificial scaffolding through which something experienced as natural desire is actually able to become a relationship at all. Strip the scaffolding away and what remains is not a purer, more authentic form of attraction; it is, empirically, mutual paralysis, the well-documented awkwardness of two people who want something from each other and have no available form through which to want it.
The second is the law of mediated immediacy. Attraction is among the most immediate-feeling experiences available to a person: a glance, a scent, a particular laugh can seem to strike directly through, bypassing reflection altogether. Plessner's claim is not that this feeling is illusory but that its immediacy is itself an achievement of mediation, not evidence against mediation. The glance registers as significant because it has already been filtered through language, cultural category, and personal narrative: through the prior sedimentation of what counts as attractive, through the story a person tells about their own "type," through socially available scripts for what a lingering look means. Mediated immediacy explains why attraction so often feels like a discovery ("I just knew") while remaining, on inspection, saturated with cultural and biographical mediation that the feeling of immediacy actively conceals from the person having it.
The third is the law of the utopian standpoint. Plessner argues that the excentric human being lacks any final, secure position it can simply occupy; its relation to the world remains constitutively open, without a resting place that would close the question of what it is. Applied to partnership, this law names something the empirical and institutional material in later sections will confirm repeatedly: no symbolic determination of a relationship (boyfriend, fiancé, spouse) ever finally exhausts or grounds it. The relationship retains a residual homelessness, an inability to arrive at a settled final position, which is precisely why relationships require perpetual renewal, reassurance, and negotiation, and why they remain vulnerable to being recursively reopened even after their most formal closure. A wedding does not resolve the utopian standpoint of the couple it creates; it stabilizes a field within which that standpoint continues to operate.
Plessner's earlier, more directly social work adds a further resource that his philosophical anthropology alone does not supply. In The Limits of Community, written as a critique of interwar ideals of total, transparent Gemeinschaft, Plessner defended distance, role, ceremony, and tact as necessary protections against an intimacy so complete it becomes unbearable: what he called, approvingly, the dignity of reserve. This is not a minor addendum; it connects directly to the concept this essay will develop as interrecursive load. If A's awareness of B's evaluation can transform A's own evaluation, and B's awareness of A's awareness can transform it again, then a relationship with no buffering devices at all (total, immediate, continuously updated mutual transparency) would not be maximally intimate. It would be recursively unbearable, a runaway loop of evaluation evaluating evaluation with no point of rest. Courtship etiquette, discretion, the ritualized structure of a first date, the social permission to not say too much too soon, the plausible deniability built into flirtation, and the slow, conventionally paced reveal of intention all function as devices that reduce recursive load rather than as barriers standing between two people and some more authentic, unmediated encounter. They make otherwise unmanageable mutual legibility tolerable enough to proceed. This is Plessner's own insight, developed for social life in general, arriving at its most intimate and highest-stakes application.
What the Evidence Shows: Preferences, Predictors, and Their Limits
Having distinguished valuation, formation, and stabilization, and having given Plessner's laws a concrete anchor in courtship, it becomes possible to read the empirical literature with more precision about which of these three processes any given finding actually concerns. Most of the field's best-established results are results about erotic valuation and early formation; comparatively little of it speaks directly to stabilization, and this asymmetry matters for how confidently the findings can be extended.
The contemporary consensus rejects a simple checklist model in which a person carries a fixed list of desired traits, scans a stable population, and selects the best match. Certain broad dimensions recur across methods (physical appearance, warmth and trustworthiness, competence and ambition, and similarity of values and life goals), but these labels conceal considerable variation in what is actually measured. A revealing experimental study by Li, Bailey, Kenrick, and Linsenmeier, using a forced "mate budget" that required participants to trade traits against one another, found that both sexes treated dependability and warmth as necessities (traits with a threshold below which nothing else compensates) while treating physical attractiveness and status as luxuries, valuable mainly once the necessities were secured. Different traits, in other words, behave like different kinds of goods, with different thresholds and different marginal value, which already complicates any single ranked list.
The relationship between stated ideals and actual selection has been one of the field's most productive puzzles. Eastwick and Finkel's speed-dating research found that participants' abstract preferences often failed to predict whom they actually chose after brief live interactions, suggesting limited introspective access to the drivers of in-person attraction. Later work refined rather than overturned this finding: Li, Meltzer, and McNulty, deliberately including low-attractiveness and low-status candidates in speed-dating and messaging designs, found that men were more likely to reject low-attractiveness partners in long-term conditions while women were more likely to reject low-status partners, and that individual differences in stated priorities predicted actual choices specifically when the relevant trait varied enough within the available pool. Predictive power, that is, depends heavily on measurement, on whether a trait genuinely varies among the people actually on offer, and on whether the goal is short-term or long-term, a distinction with a long history in evolutionary approaches to mating psychology.
Sex differences occupy an outsized place in public discussion and deserve a precise statement here. Buss's classic thirty-seven-culture study found that women, on average, placed relatively more weight on a partner's financial prospects, while men placed relatively more weight on youth and physical attractiveness, a pattern subsequently replicated, with real variation, across many countries. A large 2020 replication across forty-five countries confirmed persistent average differences in several preferences and in preferred partner age, while finding that age-gap preferences specifically narrowed in more gender-equal countries, a moderation effect rather than the wholesale disappearance a purely social-role account might predict. Two qualifications matter. Group averages coexist with enormous overlapping distributions, so that headline claims about what "men want" and "women want" badly misdescribe most actual people. And Eastwick and colleagues' meta-analysis found that although men and women differ in how much importance they report assigning to attractiveness and earning prospects, the correlations between those traits and actual romantic evaluation of real targets were statistically similar across the sexes. A difference in stated priority is not the same magnitude of thing as a difference in what actually predicts attraction to a real person, and treating a questionnaire average as a behavioral law collapses exactly the distinction between erotic valuation and its self-report that this essay has been at pains to keep open.
Assortative mating supplies the clearest population-level regularity in the field. Horwitz and colleagues' meta-analysis of 480 partner correlations across twenty-two traits, supplemented by UK Biobank data on 133 traits, found substantial partner resemblance for political values (approximately .58), religiosity (.56), and educational attainment (.55), alongside much smaller resemblance for personality traits such as extraversion (roughly .08). The causal interpretation of these numbers will be treated carefully in a later section; for now it is enough to note that they are robust and real, whatever produces them.
A brief methodological caution belongs here as well. Biological research on partner choice (immune-gene compatibility, hormonal and cycle-linked shifts in preference) has produced a genuinely mixed record that public science writing has often outrun. A meta-analysis of preference for immune-gene dissimilarity found substantial methodological heterogeneity and no support for a simple universal pattern. Claims of systematic cycle-linked shifts in women's mate preferences, once treated as well established, were subsequently challenged by a rigorous critique arguing that stronger designs did not reliably reproduce the reported effects, prompting a broader field-wide reckoning about measurement and replication. Vivid biological narratives about hidden determinants of desire have historically moved faster than their evidence, and any theory of partner choice (evolutionary, sociocultural, or the process-theoretic account developed in this essay) should hold its strongest causal claims to the same standard of convergent, replicated evidence.
The Center of the Argument: Interrecursive Indeterminacy
Everything reviewed so far treats the potential partner, at least implicitly, as a target: a bundle of measurable attributes that a chooser evaluates, more or less accurately, more or less consistently. This section makes the case that this framing, however empirically productive, is where the account must be left behind, because it cannot represent the single most distinctive feature of the phenomenon: the target is itself evaluating the evaluator, and knows it.
An ordinary contrast makes the point vivid. A person can prefer mango to apple without the mango knowing it has been preferred, still less revising its own qualities in response. A house does not become more charming because a buyer has begun to consider it. Another human being does. I can be attracted to someone; I can also wonder whether they are attracted to me, and that wondering is not idle, because their apparent interest can increase my own, and their apparent indifference can either extinguish my interest or, in the familiar and slightly uncomfortable way of human desire, intensify it. I might conceal my attraction because I believe that revealing it would change theirs. And I might be right. They might notice the concealment. I might notice that they have noticed.
The relationship-science literature already has a name for part of this: reciprocal liking, the well-established finding that mutual interest functions as a strong, sometimes decisive, predictor of relationship formation. It is important to be precise about why the claim of this essay is stronger than reciprocity, because a sympathetic reviewer trained in that literature could reasonably object that everything just described is exactly what reciprocal liking and ordinary attainability updating already predict: learning that someone likes you makes them seem more attainable, and attainability itself is attractive. That objection has to be taken seriously, and meeting it requires an experiment that reciprocal-liking and attainability accounts cannot straightforwardly assimilate.
Here is that experiment, stated as a design rather than a metaphor. Hold a target's stable attributes constant across conditions (same profile, same photograph, same biographical facts) and instead manipulate recursive depth: what each participant is told about what the other knows. In a first condition, A simply learns that B likes A. In a second, A learns that B knows that A likes B. In a third, A learns that B knows that A knows that B likes A. A simple attainability-updating account predicts that most of the relevant work is done at the first level: once A knows B is interested, attainability has already updated, and further nesting should add little beyond diminishing informational returns, since deeper orders convey no new fact about B's stable attributes or baseline interest. The interrecursive account predicts something different: that attraction, self-disclosure, anxiety, self-presentation, and felt uncertainty should continue to shift in systematic, non-trivial ways across recursive depth, because what changes at each order is not information about B's fixed qualities but the felt exposure of being known-to-be-known, which is precisely the kind of variable an informational-updating model has no natural place for. If the data show attraction and self-presentation moving in predictable, patterned ways as recursive depth increases (beyond what a plateauing attainability curve would predict), that is evidence for a mechanism ordinary preference theory cannot assimilate nearly as easily. If they do not, the interrecursive claim needs real qualification. Either outcome would be informative, which is what makes this the flagship test of the argument rather than merely an illustration of it.
The graded levels of recursivity already implicit in this discussion give the claim further texture. At the most absorbed level (L1), attraction is largely embodied and unreflective: one moves closer, softens one's voice, mirrors posture, without narrating any of it. At a more reflective level (L2), something registers as a felt difference: an unusual awareness of another's presence, an inexplicable eagerness to prolong a conversation. At a further level (L3), the experience becomes explicitly questionable: Do I like this person? Do they like me?; and the asking is not neutral: a person comfortably enjoying another's company can become newly awkward the moment they ask themselves whether attraction is involved, while an ambiguous feeling can crystallize and intensify the moment it is put into words. Beyond this lies symbolic classification (L4: boyfriend, situationship, soulmate, unsuitable match), through which an unstable, still-forming process becomes legible to both parties and to everyone around them. And beyond even that lies a level (L5) at which the classifications themselves become questionable: why do I keep assuming monogamy; why does marriage matter to me; what does it mean that I keep describing people as "not my type" right before falling for them? None of these levels is a stage a relationship simply passes through once. A single evening can move from L1 absorption to L3 anxious self-questioning and back again several times, which is exactly what the recursive-depth experiment is designed to detect under controlled conditions rather than merely describe anecdotally.
The Recursive Career of a Partnership
If interrecursivity is the mechanism, the next task is to show its trajectory: to trace one sequence through which a relationship actually moves, from a wholly unspecified future partner to a fully stabilized institutional fact. The stages below are not strictly linear in any individual case, and a relationship can loop backward from any stage to an earlier one, but laying them out in order clarifies what "partner choice" is actually a name for: not a decision but a career.
field of possible others
│
▼
first salience (someone becomes noticeable at all; L1/L2)
│
▼
attraction (embodied and/or reflective pull; L1–L3)
│
▼
recognition of possible reciprocity ("might they feel this too?")
│
▼
mutual interrecursive escalation (each responds to signs of the other's response)
│
▼
provisional pairing (dating, seeing each other, undeclared)
│
▼
symbolic naming ("we're together"; L4 classification)
│
▼
institutional stabilization (engagement, marriage, formal agreement)
│
▼
habitualization (the relationship recedes into absorbed routine; back toward L1)
│
▼
destabilization or renewed recursion (conflict, doubt, rupture, or deliberate renewal)
Reading the empirical literature against this schematic clarifies where most existing research actually sits. Speed-dating studies capture first salience and early attraction with real precision. Reciprocity research captures the transition from attraction to recognition of possible reciprocity. Studies of dating-app behavior capture provisional pairing and symbolic naming, often under conditions of unusual ambiguity: the entire vocabulary of the "situationship" is arguably a cultural response to a stage in this career that used to resolve faster and now, for many people, does not. Longitudinal couples research captures the transition from symbolic naming through institutional stabilization into habitualization, and machine-learning syntheses of that literature find that relationship-specific dynamics (how two people actually behave together) predict later quality better than either partner's individual traits considered alone, a finding entirely consistent with treating the relationship itself, not either party's fixed profile, as the primary object under formation.
The schematic also reframes marriage's place in the argument. Marriage is not the outcome of mate choice, in the sense of being the terminal step in a preference-satisfaction process that began at first salience. It is a very late, deliberately durable attempt to stabilize a relationship that has already been recursively generated through several prior stages, none of which marriage itself produces. This is not a semantic quibble. It changes what marriage should be expected to do and what it should not: a wedding cannot manufacture the interrecursive history that precedes it, and its failure to do so is not evidence against the institution but a direct consequence of what institutional stabilization actually stabilizes: status, obligation, and recognition, not the underlying recursive process itself, which continues, as the final two stages of the schematic show, straight through the ceremony and out the other side.
The Five Mediations as a Diagnostic Instrument
A further way to see why no simple model of matching can succeed is to notice that existing partner-choice research, taken as a whole, tends to specialize in exactly one of five irreducible mediations through which human evaluation occurs (multisensorial embodiment, being-with, multimaterial forming, multiversal dwelling, and multisymbolism) while implicitly treating its chosen mediation as though it were the whole of the phenomenon.
Physical-attractiveness research is, almost by definition, research into multisensorial embodiment: the moderate-to-strong association between rated attractiveness and romantic evaluation, on the order of r = .40 across a large meta-analytic literature, is a genuine and well-replicated finding about embodied response. Reciprocity and interaction research (speed dating, live conversation, dyadic behavioral coding) gets closer to being-with: the finding that mutual responsiveness in real time predicts outcomes better than static profile comparison is a finding about the interactional mediation specifically. Status and resource theories capture multimateriality: the same meta-analytic literature finds earning prospects predicting romantic evaluation at a much smaller r = .10, and market-value research linking self-perceived desirability to the selectivity of stated preferences is explicitly a theory of relative material and social position. Proximity and network research captures multiversal dwelling: the classic sociological finding that weak ties and network structure determine who becomes available for contact at all is a finding about the spatial and social organization of encounter, prior to any evaluation of traits. And stated-preference research, along with the entire architecture of dating-platform profiles, captures multisymbolism most directly: age, height, occupation, declared politics, and photographs are precisely the sedimented symbolic residue of a person, extracted for search and comparison.
The diagnostic point is not that each of these research traditions is wrong about its own domain; each is often quite well supported within it. The point is that none of them, taken separately or even summed, is sovereign, because the five mediations can diverge from one another in ways that no single-mediation theory can represent. A person can be bodily compelling and interpersonally excruciating to be around. A person can be a poor fit on every declared symbolic criterion and turn out, on actual encounter, to be extraordinarily easy to be with. A person can be materially impractical (the wrong city, the wrong income, the wrong life stage) while being embodied and interactional dynamite. This is stronger than the common and rather weak observation that "partner choice is complex." It is the specific claim that attraction has no single sovereign currency into which embodiment, interaction, material practicality, geography, and symbolic classification can all be converted and compared, which is exactly why a person can be, in the everyday phrase, wrong on paper and right in the room, or the reverse.
"Perfect on Paper": Dating Platforms as a Natural Experiment in Symbolic Sedimentation
That everyday phrase deserves to be treated as more than a colloquialism. It names, with real precision, a naturally occurring experimental test of the limits of symbolic determination, and contemporary dating platforms have inadvertently built the largest version of that test ever run.
A profile is an act of extraction: a living, developing, interrecursively responsive person is converted into a small set of searchable, comparable, symbolic fields: age, height, occupation, education, declared politics, a handful of photographs, a short statement of intent. This is multisymbolism operating at industrial scale, and it is worth taking seriously as a genuine technology, not merely a debased substitute for "real" encounter. Castro and Barrada's systematic review of seventy studies on the psychosocial correlates of app use, though noting the field's continuing methodological immaturity, documents real and measurable effects of this technology on exposure, motive, and behavior. Bruch and Newman's large-scale analysis of interaction across several American cities found a robust pattern of aspirational pursuit (users disproportionately contacting partners rated more desirable within the local platform market, with reply rates falling as the gap widened), evidence that platform-mediated symbolic comparison genuinely organizes behavior at scale, not merely that it exists. Hitsch, Hortaçsu, and Ariely, analyzing an online dating site directly, found substantial matching and sorting on stated attributes, showing that symbolic self-presentation does real predictive work in who contacts whom.
And then two people meet, and the entire apparatus confronts an event it did not design and cannot fully contain. A person who satisfies every declared criterion can produce nothing on meeting; a person who violates several stated preferences can prove overwhelmingly compelling. This is not a failure of the platforms at the task they were actually built for (organizing exposure, filtering an otherwise unmanageable field, converting an intractable search problem into a browsable one), and the sorting and matching effects documented above are real. It is evidence of exactly the limit this essay has been arguing for throughout: symbolic sedimentation (L4) can predict a great deal about exposure and initial contact while remaining structurally unable to predict the interrecursive process that only the encounter itself can generate. "Perfect on paper, wrong in person" is the everyday name for a finding this essay would state more precisely: multisymbolic compatibility is neither necessary nor sufficient for interrecursive escalation, because the two operate through different mediations that platforms can align on paper but cannot align in the room.
Statistics Without Overreach
It is tempting, at this point in the argument, to say that population statistics tell us little about the living process this essay has been describing. That temptation should be resisted, because it overstates a claim that a narrower, more defensible version can establish just as effectively. Population-level regularities are real, and they are informative; the mistake is not in trusting them but in asking them a question they are not built to answer.
Horwitz and colleagues' assortative-mating correlations (.58 for political values, .56 for religiosity, .55 for education) are robust, real, and worth taking seriously as evidence of genuine structure in how partnerships form. What such a correlation cannot do, and does not claim to do, is recover the generative history of any individual case. The same review distinguishes several mechanisms that could each produce identical population-level resemblance: phenotypic homogamy, in which people directly prefer similarity; social homogamy, in which people meet and pair within strata already organized by the trait: schools, professions, neighborhoods; genetic homogamy, involving correlated heritable components; and convergence, in which partners become more alike after the relationship has already formed. A correlation of .55 is compatible with any mixture of these four mechanisms, and disentangling them requires data (pre-formation measurement, network information, repeated assessment over time) that a cross-sectional correlation simply does not contain.
The precise claim, then, is this: population statistics can accurately represent the distribution of outcomes produced by millions of recursive trajectories without thereby recovering the generative recursive architecture of any one of them. This is a narrower and more defensible proposition than the sweeping claim that statistics "operate only on sedimented outcomes" and have nothing to say about process; longitudinal and time-series methods, appropriately designed, can preserve real temporal structure, and this essay's own proposed research program in a later section depends on exactly that possibility. What a distribution cannot do, however carefully collected, is tell you how one particular stranger became, to one particular person, irreplaceable. Both facts are true at once, and the field's occasional confusion between them (treating a robust correlation as if it settled a question about individual process, or dismissing a robust correlation because it cannot settle that question) is avoidable once the two levels of description are kept explicitly apart.
Power as a Mechanism of Recursive Closure
It would be easy to treat questions of inequality, sexual orientation, race, and safety as a final corrective section: a reminder, appended after the main argument, that the preceding account should not be assumed to generalize automatically to everyone. That placement would badly undersell what this material actually contributes, because power belongs inside the theory, not alongside it as a caveat.
The core claim of this essay has been that sexual partnership is unusually undetermined because it depends on an interrecursive process that neither party can unilaterally control: each must remain open to being changed by the other's regard, and the relationship's eventual shape is generated by that mutual openness rather than discovered in advance. Power, on this account, is best defined as the unequal capacity to close another person's recursive possibilities before that process can run its course, or to force it to run on terms set by a third party. A forced marriage does not merely restrict a person's field of eligible partners in the way an incest taboo or a religious rule does; it attempts to substitute institutional stabilization directly for the interrecursive process that would otherwise precede and generate it, foreclosing the escalation stage of the career traced earlier before it can begin. Economic dependence that makes leaving a relationship materially impossible does not merely lower a person's exit options; it forecloses the destabilization-or-renewed-recursion stage that keeps any relationship, however institutionally settled, still alive as a relationship rather than merely as an arrangement. Criminalization of same-sex relationships, or their pathologization within medicine and religion, does not simply stigmatize an already-formed relationship; it attempts to prevent the earlier stages of the recursive career (salience, attraction, recognition of possible reciprocity) from occurring safely or openly at all, which is a more specific and more damaging mechanism than "social disapproval," because it targets the process itself rather than merely its eventual social reception.
This reframing changes how the literature on sexual orientation, gender diversity, and consensual non-monogamy should be read here. These are not populations added to the analysis for the sake of completeness; they are cases in which the ordinary machinery of recursive closure has operated with unusual force, and studying them directly is one of the clearest ways to see that machinery at work. A dating app that functions as a mild convenience for a heterosexual user embedded in a dense social network can function as genuinely essential infrastructure for a sexual-minority user in a hostile or sparse local environment, precisely because it may be one of the few channels through which the earlier stages of the recursive career can occur safely. Reviews of consensual non-monogamy argue that studying such configurations directly, rather than treating them as a deviation from a monogamous baseline, sharpens rather than merely extends relationship science, in part because they make visible how many overlapping, differently negotiated recursive processes a single person can sustain at once, a possibility the standard dyadic model quietly forecloses by assumption rather than by argument. Racialization structures neighborhoods, schools, and friendship networks well before anyone opens a dating profile, and it becomes visible rather than created in the message and response data that platforms generate; work on this pattern is best read as evidence of how existing social categorization interacts with the architecture of encounter, not as evidence of a purely internal, unmediated preference operating independent of that architecture. Safety considerations (the threat of violence, harassment, coercion, involuntary disclosure, or reputational sanction), unevenly distributed by gender, sexuality, race, disability, and age, can appear in aggregated behavioral data as ordinary "selectivity" when they are in fact adaptive responses to real risk of recursive foreclosure, a distinction any adequate research design has to build in rather than infer after the fact.
A theory of partner choice that treats power as external context is a weaker theory than one that treats power as a set of specific mechanisms for closing, redirecting, or foreclosing an interrecursive process before it can generate its own outcome. The latter is what this section has tried to state.
Kinship, Romantic Modernity, and the Relocation of Recursive Workload
Institutions are the general case of which the previous section's coercive examples are the extreme end, and it is worth returning to the more ordinary institutional machinery (kinship, arranged marriage, and romantic ideology) with the fuller theoretical apparatus now in place.
Kinship systems, caste rules, and family involvement in partner selection are conventionally analyzed through an opposition between social constraint and individual freedom: how much "choice" does a person really have. That framing is not so much wrong as it is asking the wrong question. The more productive question is what work these institutions perform, and the answer proposed here is that they redistribute the recursive labor of partner selection. Family-assisted courtship and arranged marriage can pre-stabilize eligibility, expectations, and consequences before two prospective spouses ever encounter each other as such, externalizing a portion of what a couple in a different institutional setting would otherwise have to negotiate from scratch: is this person acceptable to my family, will this union be socially recognized, what obligations follow from it. This does not eliminate the underlying interrecursive process; the resulting relationship remains capable of failing or flourishing on its own terms, and research on such arrangements is heterogeneous enough (varying by consent, gendered power, exit options, class, and religious context) that no single verdict on the institution as a whole is defensible. What changes is how much of the evaluative and negotiative burden falls to the two people directly rather than to the institutions surrounding them.
Contemporary romantic ideology looks, at first glance, like the straightforward reverse: less institutional determination, more individual freedom. The stronger claim this essay wants to make is that romantic modernity does not simply replace determination with freedom. It relocates determination from kinship institutions into individual self-recursivity. "Marry the person you love" sounds like the absence of external constraint, but it quietly imposes an immense epistemic and emotional obligation on the individual, who must now discover, unaided by any external institution, whom they truly love, whether this is settling, whether the chemistry is sufficient, whether a better match exists, questions no simple internal faculty can answer independently of the relationships themselves, which means romantic freedom generates its own characteristic and often exhausting form of self-scrutiny.
Dating platforms then produce something close to a bizarre culmination of this trajectory rather than merely its latest chapter: maximum apparent choice, maximum symbolic preclassification, and maximum selfrecursive responsibility, all operating simultaneously. The field of theoretically available partners expands enormously; each of them is rendered, through profile architecture, into an even more extensively symbolized object than a family-introduced prospect ever was; and the entire burden of navigating this expanded, hyper-classified field (filtering, comparing, doubting, second-guessing, managing the well-documented fatigue of repeated low-stakes rejection) falls on the individual user, alone, with no institution left to share the recursive workload. This is a considerably stronger and more specific anthropological description of the peculiar exhaustion many people report about modern dating than the familiar diagnosis of mere "choice overload." The problem is not simply that there are too many options. It is that romantic modernity has combined an unprecedented expansion of the symbolic field with an unprecedented removal of the institutional buffers (kinship, community vetting, ritual pacing) that used to absorb part of the recursive load Plessner's laws of natural artificiality and mediated immediacy show every courtship must somehow carry.
Marriage as the Inversion Point
The material assembled across the preceding sections traces a single, striking temporal structure, and it is worth making that structure explicit as a figure rather than leaving it implicit in scattered observations about marriage.
radical prospective indeterminacy
(the future partner may not yet be born, may be a
total stranger, is not predictable from any fact
available at birth or through most of childhood)
│
▼
recursively generated valuation
(first salience → attraction → mutual escalation;
a relationship comes into existence that did not
exist before, produced by the encounter itself)
│
▼
symbolic / institutional stabilization
(naming, engagement, marriage, formal agreement;
L4 classification fixes status, obligation, and
public recognition)
│
▼
massive downstream determination
(residence, migration, finances, children, kinship
ties, daily routine, caregiving, who is present at
illness and death)
The stranger who plays no role whatsoever in a person's first twenty years can, at twenty-one, become the organizing structure around which the rest of that life is arranged. This is not a loose rhetorical flourish; it is the precise inversion the figure displays. The domain that begins furthest toward the open end of any plausible scale of human determination (a domain in which, at birth, literally nothing is fixed, not even the future partner's existence) ends, for most people who marry or form a lasting partnership, as one of the most determining structures their life will contain, reorganizing where they live, whether they migrate, how their money is spent, which relatives become significant, and who is present at the end.
The trajectory is therefore not, as a simpler account might suggest, indeterminacy resolving smoothly into choice. It is radical prospective indeterminacy giving way to a recursively generated valuation that did not exist in advance, which is then symbolically and institutionally stabilized, which then produces determination on a scale wildly disproportionate to how open the starting point was. Nothing about the initial indeterminacy predicts the eventual weight of the outcome; the weight is added entirely by what the middle two stages of the figure do. This is, among other things, a direct answer to the question this essay opened with, about why the newborn-cannot-predict-their-spouse observation, taken alone, understates the phenomenon. The interesting fact is not that the starting point is open. It is that from an opening this radical, human beings routinely produce an ending this determined, and that the mechanism connecting the two is neither pure chance nor pure discovery, but the recursive, then stabilized, generation traced above.
Cultural Elaboration Tracks Residual Indeterminacy: Toward a General Hypothesis
The kinship, marriage, and courtship material assembled in this essay supports a proposition considerably larger than an account of dating, and it is worth stating that proposition explicitly, with appropriate caution, because it is where this essay stops being only an anthropology of sexual partner choice and becomes a claim in general philosophical anthropology.
The hypothesis is this: the greater the stakes of a domain of human life, and the less completely its outcome can be determined in advance by available information, the more elaborate the stabilizing machinery (symbolic, material, and institutional) that human societies tend to construct around it. Sexuality is an unusually clear case of the pattern, because it combines extremely high stakes (reproduction, household formation, kinship, inheritance, lifelong companionship or its absence) with an unusually stubborn residual indeterminacy, of exactly the interrecursive kind this essay has been describing, that no amount of institutional pre-sorting fully eliminates. But it is very unlikely to be the only case, and naming the others, even briefly, shows what kind of hypothesis this is.
Death is a second case: enormously consequential, resistant to precise individual prediction even with extensive medical information, and surrounded by some of the most elaborate symbolic and ritual machinery any human society constructs: funerary rites, inheritance law, mourning conventions, religious doctrine concerning an afterlife that, by definition, no living person can verify. Serious illness is a related case, at a smaller scale: diagnosis narrows uncertainty without eliminating it, and the surrounding apparatus of medical ritual, prognosis-communication protocols, and support institutions can be read as machinery for making a genuinely open, high-stakes outcome livable in the interim. Political succession is a third case: who will hold power after a given leader is often a matter of enormous consequence and genuine indeterminacy even under formally settled rules, which is presumably why succession has historically attracted some of the most elaborate ritual, legal, and sometimes violent institutional attention of any political process. Financial markets may be a fourth, more abstract case: prices aggregate enormous amounts of distributed information yet remain famously resistant to individual prediction, and the vast institutional apparatus of contracts, insurance, regulation, and hedging that surrounds markets can be read, at least in part, as machinery for making a domain that cannot be individually determined in advance collectively tolerable to inhabit.
None of these parallels is developed here beyond a gesture, and each would require its own sustained argument to establish rather than merely suggest. The point of naming them is narrower and more defensible: to show that the mechanism this essay has traced in detail for sexual partnership (elaborate cultural machinery clustering not where human life is most determined but precisely where consequential indeterminacy is greatest and least eliminable) is not a peculiarity of sexuality alone. If the hypothesis holds even approximately across these other domains, then the extraordinary cultural elaboration surrounding sex is not primarily evidence that sexuality is culturally constructed all the way down, as some strong constructionist readings would have it. It is evidence of something closer to the opposite: that culture gathers, with unusual density, precisely at the places where life will not stay still.
Toward a Research Program
A theoretical account that offers only interpretive richness, however genuine, risks becoming unfalsifiable, and the task of this section is to specify what evidence would actually distinguish the claims made here from ordinary learning, incomplete information, or standard reciprocal-feedback models already present in relationship science.
The flagship study is the recursive-depth experiment already described in full: holding a target's stable attributes constant while manipulating what each participant is told about what the other knows about their mutual interest, across at least three orders of nesting, and measuring attraction, self-disclosure, anxiety, self-presentation, and felt uncertainty at each level. A simple attainability-updating account predicts a plateau after the first order; the interrecursive account predicts continued, patterned movement across orders. This is the single cleanest test available of whether interrecursivity is a distinct mechanism or a redescription of reciprocal liking, and it should be run before any of the studies below, since a null result here would require real revision of the theory's strongest claim.
A second study could examine the specific claim, developed in the section on the five mediations, that felt attraction sometimes contradicts a person's own symbolic self-classification: the everyday experience of "this isn't normally my type." Following participants prospectively across multiple real encounters, recording in-the-moment attraction alongside stated type and relationship labels, and conducting qualitative interviews about instances of felt discrepancy, would let researchers test whether such discrepancies reliably trigger explicit self-reflection and later recategorization, an observable process, not merely a philosophical possibility.
A third study could test institutional stabilization directly, following couples through cohabitation, marriage, formally negotiated non-monogamous agreements, or family-assisted courtship, measuring coordination, perceived certainty, explicit negotiation, conflict, and partner-specific evaluation before and after the formalizing event. The prediction, drawn directly from the recursive-career schematic, is specific: formalization should generally increase coordination and reduce some forms of uncertainty while leaving, and in some cases redirecting rather than eliminating, ongoing interrecursive change captured by the habitualization and destabilization stages. Comparative work of this kind must build in direct measures of consent and power, for the reasons given in the discussion of recursive closure, since terms like "arranged marriage" and "cohabitation" each cover heterogeneous arrangements whose meaning cannot be assumed constant across cases.
A fourth line of research would examine the platform environment directly, in partnership with dating services where possible, comparing profile-level compatibility, algorithmic exposure, in-app behavior, early in-person interaction, and eventual relationship outcomes, in order to calibrate precisely where multisymbolic sedimentation predicts outcomes well and where its predictive value drops off sharply once two people are in a room together. Existing behavioral-trace research already shows what such calibration can reveal, from aspirational pursuit patterns to the sorting effects of stated search criteria.
Across all four studies, the same standard applies. Any claim distinctive to the account developed in this essay (not merely that reciprocity matters, or that institutions constrain choice, both already well established) needs to be operationalized against a plausible rival explanation drawn from existing relationship science: ordinary updating on new information, standard models of reciprocal liking, or latent stable preferences that had simply not yet been elicited. A theory that cannot specify what evidence would count against it, however elegant its vocabulary, has not yet earned a place alongside the empirical literature it hopes to inform.
Conclusion: Ontological Production, Not Prediction
It would be possible to end this essay by saying that a future partner remains, until the relationship happens, genuinely unwritten: hard to predict, resistant to determination in advance. That formulation, however, would understate the claim this essay has actually been building toward, because it can still be heard as a statement about the limits of prediction: as though the right relationship already existed somewhere, in some settled form, merely waiting to be discovered by two people clever or lucky enough to find it. Nothing in the argument developed here supports that picture, and the strongest version of the claim explicitly denies it.
The future relationship is not merely unknown because some hidden answer has not yet been located. In the sense this essay has tried to make precise across its central sections, it does not yet exist. It will be produced, not disclosed, not discovered, but produced, through recursive encounters between two beings who are themselves changed by those very encounters, in a process that no amount of prior information about either party's fixed attributes could have specified in advance, because the object in question, the relationship itself, has no determinate identity independent of the process that generates it. This is the claim that actually separates the account offered here from prediction science, from preference theory, and from generic invocations of complexity that gesture at openness without specifying what kind. Prediction science asks how well a known set of variables forecasts an outcome that already has a determinate identity waiting to be forecast. This essay has argued that, for sexual partnership specifically, that presupposition fails at the point where the object of evaluation is itself another evaluator, and the failure is not a temporary gap awaiting better data. It is structural.
Return, in closing, to Plessner. The human being is not fixed (not settled once and for all into a single, naturally given position) and this essay has tried to show that this condition should not be read as a claim that human life is simply open or free-floating. Humans are among the most thoroughly classified, documented, and institutionally determined animals that exist, and the argument here has traced, in detail, how much elaborate machinery (kinship, courtship convention, marriage law, dating-platform architecture) human societies construct specifically around sexual partnership. The genuine discovery is not the absence of this determination. It is that none of it ever finally exhausts the recursive life it tries to fix, because the object being fixed is another being capable of exceeding any fixing applied to it, in exactly the way the first being is. Sexual partner choice sits at an unusually extreme point on this general condition, because the object of the choice is another animal undergoing the identical process from the other side, reading the first person's regard while being read in turn, each capable of being altered by what the other appears to feel about them.
That doubling is the thesis this essay opened with, and it can now stand as its conclusion in a single sentence: the unfixed animal constructs elaborate systems for fixing sexual partnership because another unfixed animal can never be reduced to a fixed object of choice. Two beings, neither fully determined, each capable of being changed by the other's regard, nonetheless construct between them one of the most durable structures either will ever inhabit (a structure that reorganizes residence, kinship, finances, care, and the company kept at the end of a life) and remain, even inside that structure, never quite finished with each other. That is not a limitation on the science of partner choice waiting to be overcome by better measurement. It is the phenomenon the science, read carefully enough, has been describing all along.