29.07.05 · psychology / social-psychology

Interpersonal attraction and close relationships: adult attachment, social-exchange theories, and the Sternberg triarchic model

shipped3 tiersLean: none

Anchor (Master): Berscheid-Walster 1969; Festinger-Schachter-Back 1950; Kelley-Thibaut 1959; Hatfield-Walster 1981; Hazan-Shaver 1987; Sternberg 1986 Psych. Rev.; Reis-Shaver 1988; Rusbult 1980; Aron-Aron 1986; Gottman-Levenson 1992; Bartholomew-Horowitz 1991; Waldinger-Schulz 2023 (Harvard Study of Adult Development)

Intuition Beginner

What makes people fall in love? Psychologists have studied the question for seventy years, and the biggest predictor is the simplest: proximity. You form relationships with the people you see often — a classmate, a coworker, a neighbour. The second predictor is similarity, meaning shared values, interests, and background. The third is reciprocal liking: we tend to like the people who like us. Beyond initial attraction, long-term relationships follow patterns too. Cindy Hazan and Phillip Shaver showed in 1987 that adult romantic relationships echo the attachment patterns we formed as infants with our parents: secure, anxious, or avoidant.

Robert Sternberg proposed in 1986 that love has three components — intimacy (closeness), passion (romantic and physical attraction), and commitment (the decision to maintain the relationship). Different combinations produce different kinds of love. Intimacy plus passion gives romantic love. Intimacy plus commitment gives companionate love, the warm affection of long-married couples. All three together gives what Sternberg called consummate love. The framework is a map rather than a measurement: it organises the experiences people report and predicts which kinds of relationships endure and which dissolve.

Why does this matter? Close relationships are the single largest predictor of adult well-being. The Harvard Study of Adult Development has followed the same cohort for more than eighty years, and the quality of a person's close relationships at age fifty predicts their physical health and longevity in late life more strongly than their cholesterol level, their income, or their social class. Understanding why some relationships thrive and others fail is one of the most consequential empirical questions in the behavioural sciences.

Visual Beginner

The picture collects the four core architectures of close-relationships research into a single diagram. The left panel shows Sternberg's triarchic triangle with intimacy, passion, and commitment at the vertices and the eight sub-types of love occupying the interior regions. The right panel arranges the four adult attachment styles on the two-dimensional model of Bartholomew and Horowitz, with model of self on the vertical axis and model of others on the horizontal. The bottom panel shows Gottman's "Sound Relationship House" stacked alongside the 5:1 positive-to-negative ratio and the four horsemen of divorce.

The arrow linking Sternberg's triangle to the attachment grid is the load-bearing claim of the unit: the three components of love are not free-floating dimensions but are scaffolded by the internal working models formed in infancy and measured by adult-attachment style. The arrow from the attachment grid to the Sound Relationship House expresses the empirical finding that secure attachment is a statistical precondition for the communication patterns that Gottman's coding identifies as predictive of stability.

Worked example Beginner

The Gottman marriage lab. For more than forty years, John Gottman and his collaborators brought couples into a laboratory in Seattle, asked them to discuss a topic of ongoing conflict for fifteen minutes, filmed the conversation, and followed up years later to record which couples had stayed together and which had divorced.

Step 1. Coders trained in the Specific Affect Coding System split each conversation into six-second slices and labelled every slice with the emotion expressed. Consider a stable couple: in fifteen minutes of conflict they accumulate $50$ positive and $10$ negative interactions, giving a ratio of $5 : 1$ . A couple heading toward divorce might accumulate $20$ positive and $20$ negative, giving a ratio of $1 : 1$ . The ratio, not the absolute number of negative events, is what predicts stability.

Step 2. Among the negative interactions, four patterns were over-represented in couples who later divorced. Criticism attacks the partner's character ("you are selfish") rather than the behaviour. Contempt conveys superiority through disgust, sarcasm, or eye-rolling. Defensiveness plays victim and rejects influence. Stonewalling withdraws from the interaction entirely. Gottman named these the four horsemen.

Step 3. In the original Seattle sample, the presence of the four horsemen — and especially contempt — predicted divorce with approximately $93%$ accuracy over a fourteen-year follow-up. Independent replications place the predictive accuracy in the $80$ to $90%$ range across different samples and coding procedures. The clinical applications integrate this coding into behavioural couple therapy.

What this tells us: marital stability is empirically trackable. The same conflict discussion, coded frame-by-frame, yields a quantitative signal that forecasts the relationship's trajectory years in advance. The signal is concentrated in the ratio of positive to negative exchanges and in the presence of contempt in particular.

Check your understanding Beginner

Exercise (easy, multiple choice).

In Sternberg's triarchic theory, which component of love is defined as "the decision to maintain the relationship in the long term"?

A. Intimacy B. Passion C. Commitment D. Reciprocal liking

Hint

Sternberg names three components. One captures closeness and sharing, one captures romantic and physical attraction, and the third captures the cognitive decision to stay with the partner over time.

Answer

C. Commitment. In Sternberg's (1986) triangular theory, commitment is the decision to love and to maintain that love over time; it develops more slowly than passion and persists after passion has faded. Intimacy (A) is the closeness and connectedness component. Passion (B) is the romantic and physical-attraction component. Reciprocal liking (D) is one of the antecedents of attraction and is not one of Sternberg's three components of love itself.

Exercise (easy, true-false).

The folk-saying claim that "opposites attract" is well supported by the empirical evidence on long-term relationship formation and stability.

Hint

Consider what the research on similarity in attitudes, values, personality, and background actually shows about which pairs of partners form stable relationships.

Answer

False. The empirical literature on attraction, beginning with Newcomb's 1961 study and synthesised in Berscheid and Reis's 1998 handbook chapter, finds that similarity in attitudes, values, personality, and background is a strong and reliable predictor of attraction and of long-term relationship stability. The "opposites attract" pattern is occasionally observed for superficial complementarity (e.g. one partner talks more, the other listens), but the bulk of the evidence shows that similarity is the rule. Couples who differ on core values, religion, or attitudes show higher rates of dissatisfaction and dissolution than matched couples.

Exercise (easy, numeric).

Gottman's marriage lab reports that stable couples maintain a positive-to-negative interaction ratio of about $5 : 1$ during conflict discussions. In a fifteen-minute conversation with $10$ negative interactions coded, how many positive interactions would the lab expect for a couple in the stable range? Give the answer as a whole number.

Hint

Multiply the number of negative interactions by the stable ratio.

Answer

$50$ . A $5 : 1$ ratio with $10$ negative interactions means $5 \times 10 = 50$ positive interactions. This number, not the absence of negativity, is what characterises the stable couples in Gottman's samples. Couples who maintain roughly five positive exchanges for every negative one — even during conflict — show the affective balance that predicts long-term stability. Couples near a $1 : 1$ ratio, by contrast, are at substantially elevated risk of divorce over the follow-up window.

Formal definition Intermediate+

Close-relationships research organises around four empirical targets: the antecedents of initial attraction, the structural types of adult attachment, the social-exchange accounting that governs commitment, and the triarchic decomposition of love itself. Each target has a canonical originator and a quantitative or quasi-quantitative formalism.

Definition (Antecedents of attraction). Five antecedents are empirically established as the principal predictors of whether one person will form a relationship with another ^{[BerscheidWalster1969]}:

Proximity. The single strongest predictor of relationship formation is physical and functional distance. Festinger, Schachter, and Back (1950) ^{[FestingerSchachterBack1950]} showed in their Westgate-Westgate West married-student housing study that residents of adjacent apartments were many times more likely to become friends than residents two doors apart, and that functional distance (the layout of stairwells, mailboxes, and footpaths) predicted friendship better than Euclidean distance.
Familiarity. Zajonc's (1968) mere-exposure effect shows that repeated exposure to a neutral stimulus increases liking for that stimulus, up to a satiation point. The effect is robust for faces, melodies, and ideographs.
Similarity. The Byrne (1971) attraction paradigm quantified the lawful-linear relationship between proportion of similar attitudes and reported attraction: $A = k \cdot p_{sim} + c$ , where $A$ is the attraction rating, $p_{sim}$ is the proportion of similar attitudes, and $k, c$ are free parameters.
Physical attractiveness. The matching hypothesis (Walster et al. 1966 computer-dance study) holds that people pair with others of similar perceived attractiveness; the original 1966 study found that attractiveness of the randomly-assigned partner, not similarity of attitudes or personality, predicted whether a second date was sought.
Reciprocal liking. Aronson (1969) and others demonstrated that the knowledge that the other person likes you is among the strongest single determinants of reported attraction, especially in long-term contexts where the cost of misplaced affection is high.

Definition (Adult attachment style, Hazan-Shaver 1987). Adult romantic attachment is a stable individual-differences dimension reflecting the internal working models of self and of others formed in infancy. The original Hazan-Shaver (1987) ^{[HazanShaver1987]} three-category measure, replicated cross-culturally, yields distribution estimates of approximately $60%$ secure, $20%$ anxious (also called anxious-ambivalent or preoccupied), and $25%$ avoidant in non-clinical adult samples; the residual is inconsistent responding. Bartholomew and Horowitz (1991) ^{[BartholomewHorowitz1991]} split avoidant into dismissive-avoidant (high self-regard, low regard for others) and fearful-avoidant (negative models of both), producing the four-category model.

Definition (Social-exchange and investment models). Kelley and Thibaut (1959) ^{[KelleyThibaut1959]} formalised interdependence theory by representing any dyadic interaction as an outcome matrix and defining each participant's evaluation of outcomes against two internal standards: the comparison level $CL$ (the outcome the participant believes they deserve, based on past experience) and the comparison level for alternatives $CL_{alt}$ (the best outcome available outside the relationship). Satisfaction is defined as outcome minus $CL$ ; dependence is defined as outcome minus $CL_{alt}$ .

Rusbult (1980) ^{[Rusbult1980]} extended this into the investment model, in which commitment $C$ is a function of satisfaction $S$ , investments $I$ (tangible and intangible resources tied to the relationship that would be lost on exit), and alternatives $A$ :

C = f (S, I, - A),

with $S, I, A$ each measured on multi-item psychometric scales. The investment model has been replicated across heterosexual and same-sex dating relationships, marriages, abusive relationships, and non-romantic commitments including jobs and religious affiliations (Rusbult, Martz, and Agnew 1998 meta-analysis).

Definition (Sternberg's triarchic theory of love). Love is the product of three orthogonal components: intimacy (the closeness, connectedness, and bondedness one feels in the relationship), passion (the romantic and physical attraction), and commitment (the decision to love and to maintain that love). The eight combinations define eight sub-types: non-love (none present), liking (intimacy alone), infatuation (passion alone), empty love (commitment alone), romantic love (intimacy + passion), companionate love (intimacy + commitment), fatuous love (passion + commitment), and consummate love (all three) ^{[Sternberg1986]}.

Definition (Closeness-intimacy, Reis-Shaver 1988). Reis and Shaver (1988) ^{[ReisShaver1988]} model intimacy as an interpersonal process rather than a state: an episode of intimacy consists of one partner's self-disclosure followed by the other's responsive reaction (a reaction that communicates understanding, validation, and care). Intimacy accumulates across episodes to form the perception of being close.

Counterexamples to common slips

"Opposites attract." The empirical literature finds the opposite. Similarity of attitudes, values, religion, intelligence, personality (especially the Big Five dimensions), and even physical appearance predicts relationship formation and stability. Complementarity of superficial traits (e.g. one partner more talkative) is occasionally observed but does not generalise to core values.
"Physical attractiveness is everything." Attractiveness dominates first impressions and the first minutes of interaction, but its predictive power declines sharply as the relationship develops. The matching hypothesis operates throughout: stable couples are typically matched on perceived attractiveness, and the rare mismatched couple experiences elevated relationship stress.
"Attachment styles are fixed from childhood." Attachment styles are moderately stable across the lifespan (test-retest correlations around $0.4$ over multi-year intervals in adult samples) but are not fixed. They change across relationships, after major life events, and after successful therapeutic intervention. The four-category model also has ongoing dimensional-versus-categorical debates.
"Gottman predicts divorce with $94%$ accuracy." The original Seattle sample produced a figure around $93%$ for the specific sample, coding, and follow-up window; independent replications typically find $80$ to $90%$ . The figure is conditional on the procedure and should not be cited as a universal constant.
"Love is purely a cultural construction." Jankowiak and Fischer (1992) ^{[JankowiakFischer1992]} found evidence of passionate love in $147$ of $166$ cultures surveyed, including foraging societies with no contact with Western media. Cross-cultural variation exists in the expression and institutionalisation of love, but the core features of romantic-attachment behaviour are universal.
"Online dating is fundamentally different." Online platforms change the search structure (the platform is the new neighbourhood) but the same principles of proximity (in the form of algorithmic exposure), similarity (filtering on shared characteristics), and reciprocal liking (mutual-message features) operate.
"Arranged marriages are loveless." Arranged marriages typically begin with low intimacy and high commitment, and many develop companionate love over time. The investment-model prediction — high initial commitment plus accumulating investments produces stable satisfaction — characterises many arranged-marriage trajectories as accurately as it characterises love-marriage trajectories.

Key model: adult attachment theory (Hazan-Shaver 1987) Intermediate+

Model (Adult romantic attachment). Internal working models of self and of others, formed in infancy through interaction with primary caregivers, are partially carried forward into adult romantic relationships and produce stable individual differences in the experience and expression of intimacy. Adult attachment is therefore continuous with infant attachment as characterised by Bowlby (1969, 1973) and Ainsworth, Blehar, Waters, and Wall (1978), with the secure / anxious / avoidant categories of the infant-caregiver Strange Situation mapping onto the secure / anxious / avoidant adult romantic-attachment categories of Hazan-Shaver self-report measurement.

Derivation. (i) Bowlby's control-systems foundation. Bowlby (1969) modelled infant attachment as a behavioural control system: the infant monitors the caregiver's accessibility, and attachment behaviour (crying, approaching, clinging) is activated when accessibility falls below a set-point. The internal working model is the infant's encoded representation of the caregiver's typical responsiveness — a prediction about the future of the relationship.

(ii) Ainsworth's empirical categories. Ainsworth et al. (1978) classified infants in the Strange Situation as secure (upset at separation, comforted at reunion, returns to play), anxious-ambivalent (severely upset at separation, resistant at reunion), or avoidant (indifferent at separation, ignores caregiver at reunion). The category is stable across years and is predicted by observed maternal responsiveness in the first year.

(iii) Hazan-Shaver's extension. Hazan and Shaver (1987) ^{[HazanShaver1987]} observed that adult romantic relationships satisfy the defining conditions of an attachment bond: proximity-seeking, separation protest, safe-haven behaviour under threat, and secure-base behaviour supporting exploration. They proposed that the internal working model of the infant-caregiver dyad is the cognitive template applied to adult romantic partners.

(iv) Self-report validation. Hazan and Shaver constructed a three-paragraph self-report measure ("Which of the following best describes your feelings in romantic relationships?") mapping onto secure, anxious, and avoidant categories. The distributions they reported (~ $60%$ secure, ~ $20%$ anxious, ~ $25%$ avoidant) match the infant Strange Situation distributions. Subsequent research finds that securely-attached adults report longer, more trusting relationships; anxiously-attached adults report preoccupation, jealousy, and unreciprocated love; avoidantly-attached adults report discomfort with closeness and emotional distance.

(v) Cross-cultural replication. The three-category distribution has been replicated in dozens of countries, including Japan, China, Israel, the Netherlands, and Brazil, with the secure category consistently in the majority. Variations exist — avoidant attachment is somewhat more common in individualist cultures, anxious attachment somewhat more common in collectivist cultures — but the basic structure replicates.

(vi) The four-category refinement. Bartholomew and Horowitz (1991) ^{[BartholomewHorowitz1991]} split avoidance into dismissive (positive model of self, negative model of others — "I do not need others") and fearful (negative model of both — "I want closeness but fear rejection"). The four-category model maps onto two underlying dimensions: model of self (positive versus negative) and model of others (positive versus negative), yielding the four quadrants of secure, preoccupied, dismissive, and fearful. The dimensional representation has gained support from factor-analytic studies and is the basis of the Experiences in Close Relationships (ECR) self-report instrument.

(vii) The longitudinal claim. Mikulincer and Shaver (2007) and subsequent longitudinal studies find adult attachment styles to be moderately stable over years but responsive to relationship experience and to therapy, refuting the strong-fixity hypothesis. $□$

Caveat. The continuous-mapping claim has been contested on several grounds. First, infant attachment is observational (the Strange Situation codes infant behaviour under structured stress), while adult attachment is mostly self-reported, with the methodological weaknesses of self-report. Second, the distribution match (~ $60/20/25$ ) is suggestive but does not establish individual-level continuity — a secure infant is more likely but not guaranteed to become a secure adult. Third, attachment styles show a contextual dependence (the same person may be secure with one partner and anxious with another) that the trait-like formulation underplays. Fourth, evolutionary psychology accounts (Belsky 1999) propose that early attachment is a calibration mechanism for reproductive strategy rather than a fixed template, predicting the contextual variability that the categorical model finds puzzling. The Hazan-Shaver model remains the dominant framework because it captures robust empirical regularities at modest resolution, but it is best treated as a moderate-stability individual-difference variable rather than a fixed destiny.

Bridge. The Hazan-Shaver extension builds toward 29.06.01 for the developmental-psychology origins of attachment in the Bowlby-Ainsworth infant-caregiver framework, and appears again in 29.07.01 as the close-relationships slice of the social-psychology survey that this unit deepens. The central insight — that internal working models scaffold the experience of intimacy throughout the lifespan — is exactly the structural fact that links the infancy Strange Situation to the adult Gottman conflict-coding cascade: the foundational reason secure adults maintain the $5 : 1$ positive-to-negative ratio under stress is that their working model permits the responsive self-disclosure Reis and Shaver (1988) identified as the engine of intimacy. Putting these together, the bridge is between the developmental attachment tradition of Bowlby-Ainsworth-Main, the adult-romantic attachment tradition of Hazan-Shaver-Bartholomew-Horowitz, and the close-relationships-process tradition of Reis-Shaver-Gottman; this is the synthesis that identifies the internal working model with the longitudinal predictor of marital stability.

Exercises Intermediate+

Exercise 1 (easy, multiple choice).

According to Festinger, Schachter, and Back's (1950) married-student housing study, the strongest predictor of which pairs of residents became friends was:

A. Similarity of occupation or academic field B. Similarity of age and family structure C. Physical and functional distance between apartments D. Matched personality traits on self-report scales

Hint

The study examined two housing estates (Westgate and Westgate West) and recorded friendships among residents. Consider which variable the housing layout itself would most strongly structure.

Answer

C. Physical and functional distance between apartments. Festinger, Schachter, and Back (1950) ^{[FestingerSchachterBack1950]} found that residents of immediately adjacent apartments were many times more likely to become friends than residents two or three doors apart, even when those apartments housed people matched on occupation, age, and family structure. Crucially, functional distance — the layout of staircases, paths, and mailboxes that determined how often residents encountered one another — predicted friendship better than straight-line Euclidean distance. Similarity variables (A, B, D) were also predictive of friendship intensity but were dominated by proximity effects.

Exercise 2 (easy, true-false).

In Rusbult's (1980) investment model, an increase in the perceived quality of alternatives to the relationship is predicted to increase commitment to the current partner.

Hint

Write out the commitment function $C = f (S, I, - A)$ and consider the sign of the partial derivative with respect to $A$ .

Answer

False. The investment model states that commitment is an increasing function of satisfaction $S$ and investments $I$ and a decreasing function of the quality of alternatives $A$ . An increase in the perceived quality of available alternatives (a better romantic prospect, an attractive option to remain single, financial independence) reduces commitment to the current partner. This is the formalisation of the colloquial observation that people who feel they have "better options" are more likely to leave unhappy relationships. The investment model has been validated in dozens of studies predicting stay-or-leave decisions in dating, marriage, and abusive relationships.

Exercise 3 (easy, numeric).

Hazan and Shaver (1987) reported in their original adult-attachment self-report study that approximately $60%$ of adult respondents endorsed the "secure" paragraph. Suppose a contemporary non-clinical sample of $500$ adults shows the same distribution. How many of the $500$ respondents would endorse the "secure" paragraph? Give the answer as a whole number.

Hint

Multiply the proportion by the sample size: $0.60 \times 500$ .

Answer

$300$ . With $60%$ secure, $0.60 \times 500 = 300$ respondents endorse the secure paragraph; the remaining $200$ distribute between anxious ( $\approx 100$ ) and avoidant ( $\approx 125$ ) under the original Hazan-Shaver proportions, with a small inconsistent residual. This distribution has been replicated in dozens of countries; the secure-majority finding is one of the most robust individual-differences regularities in adult social-personality psychology.

Exercise 4 (medium, numeric).

A particular couple in a Gottman-style fifteen-minute conflict discussion produces $75$ positive coded interactions and $15$ negative coded interactions. Compute the positive-to-negative ratio, and state whether the couple is in the "stable" range ( $\geq 5 : 1$ ) or the "at-risk" range ( $< 5 : 1$ ).

Hint

Divide the number of positive interactions by the number of negative interactions and compare to the $5 : 1$ threshold.

Answer

Ratio $5 : 1$ ; the couple is in the stable range, exactly at the threshold. Computing: $75/15 = 5.0$ , which is the canonical threshold for stable marriages in the Gottman coding scheme. Couples at or above $5 : 1$ are in the "stable" range, those between $1 : 1$ and $5 : 1$ are in an intermediate zone of escalating risk, and those below $1 : 1$ are at high risk of divorce over multi-year follow-up. The threshold is a population-level statistical regularity rather than an individual diagnosis, and a single conversation is not by itself a reliable predictor — but the regularity is one of the most-cited empirical findings in close-relationships research.

Exercise 5 (medium, symbolic).

Sternberg's triarchic theory posits three binary components: intimacy ( $I \in {0, 1}$ ), passion ( $P \in {0, 1}$ ), and commitment ( $C \in {0, 1}$ ). Each of the eight combinations is a named sub-type of love. Write the canonical names for the four combinations in which exactly two of the three components are present. Then state which combination of exactly two components has been most consistently associated with long-term marital satisfaction in longitudinal research, and explain why.

Hint

List the four pairwise combinations: $I + P$ , $I + C$ , $P + C$ are the three possible pairs (the fourth combination "exactly two present" is one of these three — but Sternberg's eight sub-types include four combinations with two present? Re-examine: with three binary components, the count of combinations with exactly $k$ present is $(k 3)$ . The combinations with exactly two present are $(2 3) = 3$ , not four.

Answer

With three binary components, there are $(2 3) = 3$ combinations with exactly two present (the prompt's "four" is a slip; the correct count is three):

$I = 1, P = 1, C = 0$ : romantic love (intimacy + passion, no commitment).
$I = 1, P = 0, C = 1$ : companionate love (intimacy + commitment, passion faded).
$I = 0, P = 1, C = 1$ : fatuous love (passion + commitment, no intimacy — the "whirlwind" marriage).

Of these three, companionate love ( $I + C$ ) has been most consistently associated with long-term marital satisfaction in longitudinal research. Passion declines predictably over the first years of a relationship (the typical half-life of self-reported passionate love in established relationships is on the order of $18$ to $36$ months), so sustained long-term satisfaction depends on the intimacy-plus-commitment pair that survives the decline of passion. Sternberg's framework predicts — and longitudinal samples confirm — that couples who maintain companionate love report high marital satisfaction decades into the relationship even after the passion component has substantially decayed.

Exercise 6 (medium, short answer).

Two adults describe their romantic relationships. Adult A says: "I find it relatively easy to get close to partners, I am comfortable depending on them and having them depend on me, and I do not often worry about being abandoned." Adult B says: "I am uncomfortable being without a partner, I find that partners are reluctant to get as close as I would like, and I often worry that my partner does not really love me." Identify each adult's attachment style under the Hazan-Shaver (1987) three-category model and under the Bartholomew-Horowitz (1991) four-category model, and explain the difference in the two classification systems.

Hint

Hazan-Shaver has three categories (secure, anxious, avoidant). Bartholomew-Horowitz splits one of these into two by crossing model-of-self with model-of-others.

Answer

Adult A is secure under both the Hazan-Shaver three-category model and the Bartholomew-Horowitz four-category model. Adult B is anxious (also called anxious-ambivalent) under Hazan-Shaver, and preoccupied (anxious-preoccupied) under Bartholomew-Horowitz. The two systems agree on secure and on the broader anxious-versus-avoidant distinction; the difference is in how avoidance is split. Bartholomew and Horowitz (1991) ^{[BartholomewHorowitz1991]} crossed model of self (positive versus negative) with model of others (positive versus negative) to produce four categories: secure (positive self, positive others), preoccupied (negative self, positive others — the adult B pattern, wanting closeness but fearing one is unlovable), dismissive-avoidant (positive self, negative others — "I do not need others"), and fearful-avoidant (negative self, negative others — wanting closeness but fearing betrayal). The Hazan-Shaver three-category model collapses dismissive and fearful into a single "avoidant" category, while the four-category model resolves them.

Exercise 7 (medium, short answer).

Reis and Shaver's (1988) closeness-intimacy model characterises intimacy as a two-step interpersonal process. State the two steps, identify the failure modes that can occur at each step, and explain why the model predicts that an avoidantly-attached adult will report low intimacy even with a willing partner.

Hint

The two steps are self-disclosure by the speaker and responsive reception by the listener. Consider what each attachment style brings to each step.

Answer

The Reis-Shaver (1988) ^{[ReisShaver1988]} model decomposes intimacy into: (1) self-disclosure by one partner — the communication of personally significant information, feelings, or vulnerabilities; and (2) partner responsiveness — the listener's reaction communicating understanding, validation, and care. Intimacy accumulates iteratively: a successful episode of disclosure-plus-responsiveness increases the likelihood and depth of subsequent disclosure.

The failure modes are: at step (1), disclosure can be absent, superficial, or withdrawn; at step (2), the response can be dismissive, contemptuous, problem-solving-when-empathy-was-needed, or absent (stonewalling). An avoidantly-attached adult under the Hazan-Shaver-Bartholomew-Horowitz framework holds a working model in which closeness is dangerous and dependence is to be avoided. This produces a chronic deficit at step (1) — the person does not disclose the vulnerable material that would permit a responsive partner to respond — and a secondary deficit at step (2) in the form of distancing responses when the partner discloses. The model therefore predicts that avoidant adults will report low intimacy even when paired with a securely-attached, willing, responsive partner, because the bottleneck is the disclosure step rather than the responsiveness step. Empirical studies using daily-diary and laboratory-interaction methods confirm this prediction.

Exercise 8 (hard, numeric).

Gottman's cascade model of dissolution treats the four horsemen as producing a cumulative state of "negative sentiment override" in which a couple perceives even neutral interactions as negative. Suppose the per-interaction probability of contempt during a conflict discussion is $p = 0.05$ and the couple engages in $n = 100$ coded interactions. Use the Poisson approximation to the binomial to estimate the probability that the couple exhibits at least one instance of contempt during the discussion. Report the probability to two decimal places.

Hint

The Poisson rate is $λ = n p = 100 \times 0.05 = 5$ . The probability of at least one event is $1 - P (0) = 1 - e^{- λ}$ .

Answer

Approximately $0.99$ (or $99%$ ). With $n = 100$ interactions and per-interaction contempt probability $p = 0.05$ , the Poisson rate is $λ = n p = 5$ . The probability of at least one contempt event is

P (\geq 1) = 1 - P (0) = 1 - e^{- λ} = 1 - e^{- 5} = 1 - 0.0067 \approx 0.9933.

So the probability of at least one contempt event in a 100-interaction conflict discussion, even at a relatively modest per-interaction contempt rate of $5%$ , exceeds $99%$ . This calculation is the quantitative backbone of Gottman's claim that contempt is the most diagnostic of the four horsemen: in any sufficiently long conflict discussion, the couple whose contempt rate is even modestly elevated will almost certainly display the pattern, and the cascading effect of contempt on the partner's physiology (heart-rate elevation, adrenaline release, attentional narrowing) makes the pattern self-perpetuating. The clinical recommendation — to interrupt and repair contempt as soon as it appears — follows directly.

Exercise 9 (hard, short answer).

Rusbult's investment model expresses commitment as $C = f (S, I, - A)$ , where $S$ is satisfaction, $I$ is investments, and $A$ is alternatives. The model is widely replicated in dating and marital samples, but abusive relationships present an apparent counter-example: many victims of domestic violence remain with abusive partners despite low satisfaction, severe physical and psychological costs, and available shelters and support services. Use the investment model to explain this pattern, and identify which of the three predictors is doing the explanatory work.

Hint

Consider how each of $S$ , $I$ , and $A$ is operationalised in abusive relationships specifically. Think about the costs of leaving (financial dependence, children, threats of retaliation, identity) and about how alternatives are perceived rather than how they objectively are.

Answer

The investment model predicts commitment as a function of satisfaction $S$ (low in abusive relationships), investments $I$ (often very high), and the perceived quality of alternatives $A$ (often very low). The paradox of why victims stay is resolved when the model is applied carefully to the three predictors:

Satisfaction $S$ is low. Victims report low relationship satisfaction; the model does not deny this. The satisfaction component pushes toward low commitment.
Investments $I$ are high. Victims typically share housing, finances, children, social networks, religious communities, and identity with the abusive partner. The dissolution of these investments on exit is enormous. The investments component pushes toward high commitment.
Perceived alternatives $A$ are low. Abusers systematically isolate their victims from friends and family, control finances, and undermine the victim's belief that they can survive independently. Threats of retaliation ("I will kill you if you leave," "I will take the children") further depress the perceived quality of the exit option. The alternatives component pushes toward high commitment.

The explanatory work is done primarily by $I$ and $A$ — that is, by structural and perceptual barriers to exit rather than by relationship satisfaction. Rusbult, Martz, and Agnew (1998) tested the model explicitly in a sample of women in shelters for domestic violence and found that commitment predicted the decision to return to the abusive partner, with investments and alternatives doing most of the predictive work. The investment model therefore is not falsified by abusive relationships; rather, it identifies the levers a clinician or social-service intervention must pull (raise perceived alternatives, disentangle investments) to enable exit.

Exercise 10 (hard, short answer).

Berscheid and Reis (1998) note that the five canonical antecedents of attraction (proximity, familiarity, similarity, physical attractiveness, reciprocal liking) operate on different timescales and at different stages of relationship development. Construct a staged model: for each of the four stages — (a) initial contact, (b) first several interactions, (c) the transition to an established relationship, and (d) long-term maintenance — identify which antecedent(s) are most influential, and explain why the dominant predictor shifts across stages.

Hint

Consider what information each stage makes available. At initial contact, only structural variables (how often you see the person) operate. At first interactions, surface variables (attractiveness, reciprocal liking) dominate. At the transition to established relationships, similarity of values becomes more weighty. At long-term maintenance, investments and attachment style take over.

Answer

A staged model:

(a) Initial contact. The dominant antecedent is proximity (Festinger-Schachter-Back 1950). No other variable operates meaningfully until two people are repeatedly co-present. The Westgate housing study showed that even small variations in functional distance (apartment adjacent versus two doors away) produce large differences in friendship formation. The structural fact of repeated exposure is the precondition for every subsequent stage.

(b) First several interactions. Physical attractiveness (the matching hypothesis and the halo effect) and familiarity (Zajonc mere-exposure) dominate. The computer-dance study found that attractiveness of the randomly-assigned partner predicted desire for a second date more strongly than personality similarity, because personality had not yet been observed. Repeated exposure to the person's face and voice operates through the mere-exposure mechanism to elevate liking independently of any other variable.

(c) Transition to an established relationship. Similarity of attitudes, values, and personality becomes the dominant predictor. The Byrne attraction paradigm shows that as the number of observed attitude dimensions increases, the proportion-of-similar-attitudes linearly predicts attraction ratings. By the time a relationship is consolidating, surface variables have been replaced by deeper structural fit. Reciprocal liking is the second-order predictor: knowledge that the other person likes you becomes especially weighty at this stage because the costs of misplaced affection (and the benefits of mutual commitment) are now substantial.

(d) Long-term maintenance. The original five antecedents are largely replaced by attachment style (Hazan-Shaver; predicts how the partners respond to stress and to each other's disclosure), investments and alternatives (Rusbult; predict whether commitment is sustained), and interactive process variables (Reis-Shaver self-disclosure-and-responsiveness; Gottman positive-to-negative ratio and the four horsemen). The shift across stages reflects the deepening of information available: structural variables give way to surface variables give way to deep-fit variables give way to process variables. Each stage's predictor is the one most diagnostic at the level of information the relationship has accumulated.

Interpretive debates and developments Master

Result 1 (Berscheid-Walster 1969: the founding monograph). Berscheid and Walster's Interpersonal Attraction ^{[BerscheidWalster1969]} consolidated two decades of post-war research into the first systematic framework, organised around the five antecedents (proximity, familiarity, similarity, attractiveness, reciprocal liking). The 1969 monograph is the canonical origin point of the field and is the reference against which all subsequent syntheses position themselves; Berscheid and Reis's 1998 Handbook of Social Psychology chapter is the modern descendant.

Result 2 (Festinger-Schachter-Back 1950: proximity and functional distance). The Westgate-Westgate West married-student housing study ^{[FestingerSchachterBack1950]} is the empirical foundation of the proximity effect. The finding that functional distance (the layout of footpaths and staircases) predicted friendship formation more strongly than Euclidean distance is the load-bearing observation: it establishes that proximity operates through encounter frequency rather than through any intrinsic property of physical closeness.

Result 3 (Kelley-Thibaut 1959: interdependence theory). Kelley and Thibaut's The Social Psychology of Groups ^{[KelleyThibaut1959]} introduced the outcome-matrix representation of dyadic interaction and the comparison-level / comparison-level-for-alternatives framework. The formalisation is the bridge from informal social-exchange thinking (Homans 1958; Blau 1964) to the quantitative investment model of Rusbult, and provides the analytical machinery that subsequent close-relationships research has refined rather than replaced.

Result 4 (Hatfield-Walster 1981: passionate love as a labelling problem). Hatfield and Walster's A New Look at Love ^{[HatfieldWalster1981]} drew on Schachter-Singer (1962) two-factor emotion theory to model passionate love as physiological arousal plus a cognitive label attributing the arousal to romance. The model explains the "suspension bridge" effect (Dutton-Aron 1974): participants approached by an attractive confederate on a fear-arousing suspension bridge showed higher attraction ratings than those approached on a stable bridge, because the fear-induced arousal was mis-attributed to romantic attraction.

Result 5 (Hazan-Shaver 1987: adult romantic attachment). Hazan and Shaver's Journal of Personality and Social Psychology paper ^{[HazanShaver1987]} is the most-cited single paper in close-relationships research and is the load-bearing reference of this unit. The three-paragraph self-report measure produced the empirical foundation for the adult-attachment field. The 1987 paper's contribution is the bridging claim — that the Bowlby-Ainsworth infant-caregiver framework applies to adult romantic bonds, with the same three-category structure, the same internal-working-model mechanism, and the same population distribution — supported by initial self-report data and opening a research programme that has since generated thousands of studies.

Result 6 (Sternberg 1986: the triarchic theory). Sternberg's Psychological Review paper ^{[Sternberg1986]} proposed the intimacy-passion-commitment cube and the eight sub-types. The theory is partly descriptive (it organises the experiences people report) and partly predictive (it predicts which kinds of love endure: companionate love with its intimacy-plus-commitment structure outlasts passion-driven romantic love). The empirical status of the three components as orthogonal dimensions is contested — factor-analytic studies find the components to be moderately correlated rather than orthogonal — but the organising framework remains the standard pedagogical decomposition of love.

Result 7 (Reis-Shaver 1988: the closeness-intimacy process). Reis and Shaver's chapter ^{[ReisShaver1988]} reframed intimacy from a static property of a relationship to a process constituted by iterative episodes of self-disclosure plus responsive reception. The process model explains why intimacy can be lost (when disclosure-or-responsiveness fails) and regained (when the cycle resumes), and predicts that secure attachment produces more and deeper disclosure episodes and therefore higher accumulated intimacy.

Result 8 (Rusbult 1980: the investment model). Rusbult's Journal of Experimental Social Psychology paper ^{[Rusbult1980]} operationalised commitment as a function of satisfaction, investments, and alternatives. The model has been validated in more than fifty studies across dating, marital, abusive, homosexual, friendship, job, and religious-commitment samples (Le and Agnew 2003 meta-analysis across $52$ studies, $n > 11, 000$ , mean effect size $r = 0.51$ between commitment and its three predictors). The investment model is the most successful quantitative prediction in close-relationships research.

Result 9 (Aron-Aron 1986: the self-expansion model). Aron and Aron's Love and the Expansion of Self ^{[AronAron1986]} proposed that close relationships serve a self-expansion motive: the inclusion of the other's resources, perspectives, and identity in the self produces the characteristic sense of growth associated with falling in love. The model explains the rapid cognitive reorganisation that accompanies new love (the partner's traits become accessible as one's own) and predicts that relationship satisfaction tracks perceived expansion opportunities.

Result 10 (Gottman-Levenson 1992: the Seattle marriage lab). Gottman and Levenson's Journal of Personality and Social Psychology paper ^{[GottmanLevenson1992]} reported the longitudinal prediction of marital dissolution from observed conflict-discussion behaviour, with contempt, criticism, defensiveness, and stonewalling (the four horsemen) carrying most of the predictive signal. The 5:1 positive-to-negative ratio of stable couples and the diagnostic primacy of contempt emerged as the most-cited empirical findings in close-relationships research of the late twentieth century. The Seattle lab's methodology — observational coding rather than self-report — established a methodological standard that subsequent couple-research has adopted.

Result 11 (Waldinger-Schulz 2023: the Harvard Study of Adult Development). Waldinger and Schulz's 2023 synthesis ^{[WaldingerSchulz2023]} of the $85$ -year Harvard Study of Adult Development (the longest-running study of human development ever conducted) reports that the quality of close relationships at age $50$ predicts physical health at age $80$ more strongly than any other measured variable including cholesterol level, income, social class, or exercise. The finding reframes close-relationships research from a sub-discipline of social psychology to a load-bearing predictor of population health, with implications for public-health policy.

Synthesis. The foundational reason adult-attachment research has held for forty years is that the Hazan-Shaver working-model framework captures the developmental origin of the intimacy process that Reis-Shaver characterised and that Gottman's cascade measures empirically, and this is exactly the structural fact that links the five antecedents of attraction (Berscheid-Walster) to the long-term outcomes the Harvard Study measured. The bridge is between Bowlby-Ainsworth's infancy framework, Hazan-Shaver's adult extension, Reis-Shaver's process model, Rusbult's commitment calculus, and Gottman's conflict-coding cascade — five convergent lines of evidence on the same underlying architecture. Putting these together, the central insight is that close relationships are scaffolded by internal working models, maintained by iterated self-disclosure-and-responsiveness, defended by the 5:1 affective ratio, and that this entire cascade identifies adult-attachment security with the longitudinal predictor of marital stability and of late-life health. The pattern generalises across cultures (the cross-cultural attachment-style distributions of Result 5; the Jankowiak-Fischer universality finding) and appears again in 29.06.01 for the developmental origins of attachment and in 29.07.01 for the social-psychology survey context.

Full argument set Master

Proposition (Investment-model commitment as a sufficient statistic for stay-or-leave decisions). Under the assumptions that (i) commitment $C$ is a continuous real-valued latent variable, (ii) $C$ is an increasing function of satisfaction $S$ and investments $I$ and a decreasing function of alternatives $A$ , and (iii) the decision to remain in or exit a relationship is monotone in $C$ (there exists a threshold $C^{*}$ such that the individual remains if and only if $C \geq C^{*}$ ), the triple $(S, I, A)$ is a sufficient statistic for predicting stay-or-leave behaviour in the sense that no additional variable adds incremental prediction once $(S, I, A)$ is conditioned on.

Proof. Let $C = f (S, I, - A)$ be a strictly increasing function in each argument. The decision rule is "stay iff $C \geq C^{*}$ ." Suppose there exists an additional variable $X$ (e.g. attachment style, social network approval, religious belief) that adds incremental prediction to stay-or-leave behaviour $D$ beyond $(S, I, A)$ . Then $P (D ∣ S, I, A, X) \neq = P (D ∣ S, I, A)$ for some $(S, I, A, X)$ . By the decision rule, $P (D = stay ∣ S, I, A, X) = P (f (S, I, - A) \geq C^{*} ∣ S, I, A, X)$ . If $f$ and $C^{*}$ are deterministic functions of their arguments, then $P (f (S, I, - A) \geq C^{*} ∣ S, I, A, X) = P (f (S, I, - A) \geq C^{*})$ for all $X$ , contradicting the assumption of incremental prediction. The only way $X$ can add prediction is by influencing $f$ , $C^{*}$ , or the decision rule itself — that is, by being a moderator of the commitment process rather than an independent cause of the decision.

The empirical content of the proposition is therefore the moderation claim: $(S, I, A)$ is sufficient conditional on the moderating structure being stable. Le and Agnew's (2003) meta-analysis of $52$ studies across $n > 11, 000$ participants found that the triple $(S, I, A)$ accounted for an average of $r^{2} \approx 0.26$ of the variance in commitment and that this proportion was stable across relationship type (dating, marital, abusive, same-sex, friendship, job, religious), supporting the moderation-stability assumption. The residual variance — $74%$ — is attributable to measurement error, to individual differences in threshold $C^{*}$ , and to moderators (e.g. attachment style, religiosity) that operate on the commitment-to-decision mapping rather than on the commitment itself. The proposition is the formal statement of why the investment model is the dominant framework: it provides a sufficient statistic for the prediction that no competitor has surpassed. $□$

Proposition (Adult attachment as a continuity claim with bounded stability). The Hazan-Shaver (1987) extension of Bowlby-Ainsworth infant attachment to adult romantic attachment is a continuous-mapping claim $M : I \to A$ from infant Strange Situation categories $I = {secure_{i}, anxious_{i}, avoidant_{i}}$ to adult attachment categories $A = {secure_{a}, anxious_{a}, avoidant_{a}}$ such that (i) $M$ preserves category (secure infant maps to secure adult, etc.), (ii) the population-level marginal distribution is preserved ( $P (secure_{i}) \approx P (secure_{a}) \approx 0.60$ , etc.), and (iii) the per-individual mapping is moderately but not perfectly stable, with test-retest correlations in the range $r \approx 0.3$ to $0.5$ over multi-year intervals.

Proof. (i) Category preservation. The internal-working-model mechanism predicts that an infant who encoded the caregiver as responsive develops a working model of others as available and a working model of self as worthy of care; the same working model, applied to the adult romantic partner, produces the secure-adult profile of comfortable closeness. Symmetric arguments hold for anxious (inconsistent caregiver $\to$ working model of others as conditionally available $\to$ adult preoccupation with abandonment) and avoidant (rejecting caregiver $\to$ working model of others as unavailable $\to$ adult distancing). The semantic content of the three categories is preserved across the lifespan, supporting the categorical-preservation claim.

(ii) Marginal preservation. Hazan-Shaver (1987) reported adult-attachment distributions of approximately $60%$ secure, $20%$ anxious, $25%$ avoidant, with $5%$ inconsistent. Ainsworth et al. (1978) reported infant Strange Situation distributions of approximately $60%$ secure, $20%$ anxious-ambivalent, $20%$ avoidant in middle-class American samples. The match is approximate (the avoidant proportion is somewhat higher in adult samples, perhaps because some anxious infants develop avoidant adult strategies). Replications across cultures confirm the broad pattern: secure-majority, anxious-minority, avoidant-minority in dozens of countries.

(iii) Bounded stability. Longitudinal studies of attachment from infancy to adulthood (e.g. Waters, Merrick, Treboux, Crowell, and Albersheim 2000, $n = 50$ , twenty-year follow-up) find test-retest concordance of approximately $60$ to $70%$ ( $κ \approx 0.4$ , equivalent to $r \approx 0.4$ for the underlying dimension), which is substantial but far from deterministic. The loss of stability is concentrated in individuals who experienced significant life events (parental divorce, trauma, therapy) between measurement occasions — a finding consistent with the working-model mechanism, since working models update under disconfirming experience.

The continuity claim is therefore bounded: it holds at the population level (marginal preservation), at the categorical level (semantic preservation), and at the individual level at moderate strength ( $r \approx 0.4$ ). The proposition rules out both the strong-fixity claim ( $r = 1$ , attachment is destiny) and the strong-discontinuity claim ( $r = 0$ , adult attachment is unrelated to infant attachment); the empirical literature places the true stability at intermediate values, consistent with the working-model-updating mechanism. $□$

Connections Master

Social psychology survey 29.07.01 is the chapter anchor of which this unit is the deep extension on close relationships. The survey unit's brief treatment of attraction, attachment, and relationship processes is unpacked here into the full Berscheid-Walster antecedents, Hazan-Shaver attachment framework, Sternberg triarchic decomposition, Rusbult investment model, Reis-Shaver intimacy process, and Gottman conflict cascade. The five antecedents of attraction identified in the survey are formalised in this unit's Formal Definition section, and the survey's passing reference to "attachment theory applied to adult relationships" is developed into the full Hazan-Shaver continuity claim and its four-category Bartholomew-Horowitz refinement.
Developmental psychology across the lifespan 29.06.01 supplies the Bowlby-Ainsworth infant-caregiver attachment framework without which the Hazan-Shaver (1987) extension is incomprehensible. The Strange Situation categories (secure, anxious-ambivalent, avoidant) characterised by Ainsworth, Blehar, Waters, and Wall (1978) are the empirical foundation on which adult attachment is built; the present unit's Key Model section derives the adult-romantic extension from the infant-caregiver original. The continuity claim — that internal working models formed in infancy scaffold adult intimacy — is the load-bearing cross-unit dependency between 29.06.01 and this unit.
Working memory and the Baddeley-Hitch model 29.05.04 is the cognitive-systems peer that provides the comparative framework for the present unit's construct-validation strategy. Both units share the same empirical architecture: a multi-component system (Baddeley-Hitch central executive + slave systems; adult attachment working-model-of-self + working-model-of-others) is inferred from behavioural dissociations, validated by individual-differences psychometrics, and refined by neuroimaging evidence on the underlying substrate. The cross-domain methodological parallel between working-memory research and attachment research is the structural fact that licenses both as paradigmatic products of the cognitive revolution.

Historical & philosophical context Master

Ellen Berscheid and Elaine Hatfield (then Walster) established interpersonal attraction as a distinct empirical field with the 1969 monograph Interpersonal Attraction ^{[BerscheidWalster1969]}, synthesising post-war research on proximity, similarity, and physical attractiveness into a systematic framework. The proximal origin is Festinger, Schachter, and Back's 1950 married-student housing study ^{[FestingerSchachterBack1950]}, which established the proximity effect and the functional-distance refinement. Harold Kelley and John Thibaut's 1959 The Social Psychology of Groups ^{[KelleyThibaut1959]} introduced interdependence theory and the comparison-level framework that became the analytical engine of close-relationships research.

The attachment strand began with John Bowlby's three-volume Attachment and Loss (1969, 1973, 1980), which reframed infant-caregiver bonding as a biological control system rather than a secondary drive, and Mary Ainsworth's empirical operationalisation in the Strange Situation (Ainsworth, Blehar, Waters, and Wall 1978). Cindy Hazan and Phillip Shaver's 1987 Journal of Personality and Social Psychology paper ^{[HazanShaver1987]} extended the framework to adult romantic relationships, generating the most-cited line of research in close-relationships psychology. Kim Bartholomew and Leonard Horowitz's 1991 J. Pers. Soc. Psychol. paper ^{[BartholomewHorowitz1991]} refined the three-category model into the four-category model that is now standard in adult-attachment measurement.

Robert Sternberg's 1986 Psychological Review triarchic theory ^{[Sternberg1986]} provided the decomposition of love into intimacy, passion, and commitment that organised the experience-near description of love types. Harry Reis and Phillip Shaver's 1988 chapter ^{[ReisShaver1988]} reframed intimacy as an iterated process of self-disclosure and responsiveness. Caryl Rusbult's 1980 Journal of Experimental Social Psychology paper ^{[Rusbult1980]} introduced the investment model that became the dominant quantitative framework for commitment. Arthur and Elaine Aron's 1986 Love and the Expansion of Self ^{[AronAron1986]} proposed the self-expansion motive as the engine of relationship formation. John Gottman and Robert Levenson's 1992 J. Pers. Soc. Psychol. paper ^{[GottmanLevenson1992]} established the Seattle marriage lab as the methodological paradigm for observational couple research and identified the four horsemen as predictors of dissolution. Robert Waldinger and Marc Schulz's 2023 synthesis ^{[WaldingerSchulz2023]} of the eighty-five-year Harvard Study of Adult Development reported the relationship-quality finding that reframes close-relationships research as a load-bearing predictor of population health.

Bibliography Master

Ainsworth, Mary D. Salter, Mary C. Blehar, Everett Waters, and Sally Wall. Patterns of Attachment: A Psychological Study of the Strange Situation. Hillsdale, NJ: Lawrence Erlbaum, 1978.
Aron, Arthur, and Elaine N. Aron. Love and the Expansion of Self: Understanding Attraction and Satisfaction. New York: Hemisphere, 1986.
Bartholomew, Kim, and Leonard M. Horowitz. "Attachment Styles Among Young Adults: A Test of a Four-Category Model." Journal of Personality and Social Psychology 61, no. 2 (1991): 226–244.
Berscheid, Ellen, and Elaine Hatfield (Walster). Interpersonal Attraction. Reading, MA: Addison-Wesley, 1969.
Bowlby, John. Attachment and Loss, vol. 1, Attachment. New York: Basic Books, 1969.
Byrne, Donn. The Attraction Paradigm. New York: Academic Press, 1971.
Festinger, Leon, Stanley Schachter, and Kurt Back. Social Pressures in Informal Groups: A Study of Human Factors in Housing. Stanford: Stanford University Press, 1950.
Gottman, John M. The Seven Principles for Making Marriage Work. New York: Crown, 1999.
Gottman, John M., and Robert W. Levenson. "Marital Processes Predictive of Later Dissolution: Behavior, Physiology, and Health." Journal of Personality and Social Psychology 63, no. 2 (1992): 221–233.
Hatfield, Elaine, and G. William Walster. A New Look at Love. Reading, MA: Addison-Wesley, 1981.
Hazan, Cindy, and Phillip Shaver. "Romantic Love Conceptualized as an Attachment Process." Journal of Personality and Social Psychology 52, no. 3 (1987): 511–524.
Jankowiak, William R., and Edward F. Fischer. "A Cross-Cultural Perspective on Romantic Love." Ethnology 31, no. 2 (1992): 149–155.
Kelley, Harold H., and John W. Thibaut. The Social Psychology of Groups. New York: Wiley, 1959.
Le, Benjamin, and Christopher R. Agnew. "Commitment in Relationships: An Accumulation of Theory and Research, Part I: The Investment Model." Personal Relationships 10, no. 1 (2003): 37–57.
Mikulincer, Mario, and Phillip R. Shaver. Attachment in Adulthood: Structure, Dynamics, and Change. New York: Guilford Press, 2007.
Reis, Harry T., and Phillip Shaver. "Intimacy as an Interpersonal Process." In Handbook of Personal Relationships, edited by Steve W. Duck, 367–389. Chichester: Wiley, 1988.
Rusbult, Caryl E. "Commitment and Satisfaction in Dating Associations: A Test of the Investment Model." Journal of Experimental Social Psychology 16, no. 2 (1980): 172–186.
Rusbult, Caryl E., John M. Martz, and Christopher R. Agnew. "The Investment Model Scale: Measuring Commitment Level, Satisfaction Level, Quality of Alternatives, and Investment Size." Personal Relationships 5, no. 4 (1998): 357–391.
Sternberg, Robert J. "A Triangular Theory of Love." Psychological Review 93, no. 2 (1986): 119–135.
Waldinger, Robert, and Marc Schulz. The Good Life: Lessons from the World's Longest Scientific Study of Happiness. New York: Simon & Schuster, 2023.
Waters, Everett, Judith Merrick, Dominique Treboux, Judith Crowell, and L. Albersheim. "Attachment Security in Infancy and Early Adulthood: A Twenty-Year Longitudinal Study." Child Development 71, no. 3 (2000): 684–689.
Zajonc, Robert B. "Attitudinal Effects of Mere Exposure." Journal of Personality and Social Psychology 9, no. 2, pt. 2 (1968): 1–27.

Prerequisites

29.07.01

Tier anchors

beginner: Aronson, Wilson, and Akert, Social Psychology (8th ed., Pearson, 2013), Ch. 10 (Interpersonal Attraction); Gottman-Silver, What Makes Love Last? (Simon & Schuster, 2012); Esther Perel, TED talks on relational desire (2013, 2018)
intermediate: Berscheid and Reis, 'Attraction and Close Relationships,' in Gilbert, Fiske, and Lindzey (eds.), The Handbook of Social Psychology (4th ed., McGraw-Hill, 1998); Gottman, The Seven Principles for Making Marriage Work (Crown, 1999); Mikulincer and Shaver, Attachment in Adulthood (Guilford, 2007)
master: Berscheid-Walster 1969; Festinger-Schachter-Back 1950; Kelley-Thibaut 1959; Hatfield-Walster 1981; Hazan-Shaver 1987; Sternberg 1986 Psych. Rev.; Reis-Shaver 1988; Rusbult 1980; Aron-Aron 1986; Gottman-Levenson 1992; Bartholomew-Horowitz 1991; Waldinger-Schulz 2023 (Harvard Study of Adult Development)

References

Festinger, Leon, Stanley Schachter, and Kurt Back — Social Pressures in Informal Groups: A Study of Human Factors in Housing (Stanford University Press, 1950), Ch. 1-3 (proximity and functional distance in married-student housing) · Ch. 1-3; the Westgate-Westgate West dorm study
Hazan, Cindy, and Phillip Shaver — Romantic Love Conceptualized as an Attachment Process, Journal of Personality and Social Psychology 52(3), 511-524 (1987) · full text; the three-category self-report attachment measure
Sternberg, Robert J. — A Triangular Theory of Love, Psychological Review 93(2), 119-135 (1986) · full text; intimacy, passion, commitment and the eight sub-types
Rusbult, Caryl E. — Commitment and Satisfaction in Dating Associations: A Test of the Investment Model, Journal of Experimental Social Psychology 16(2), 172-186 (1980) · full text; commitment = satisfaction + investments - alternatives
Gottman, John M., and Robert W. Levenson — Marital Processes Predictive of Later Dissolution: Behavior, Physiology, and Health, Journal of Personality and Social Psychology 63(2), 221-233 (1992) · full text; the Seattle marriage lab and the four-horsemen coding
Kelley, Harold H., and John W. Thibaut — The Social Psychology of Groups (Wiley, 1959), Ch. 5-7 (interdependence theory; comparison level, comparison level for alternatives) · Ch. 5-7; outcome matrix, CL, CL_alt
Bartholomew, Kim, and Leonard M. Horowitz — Attachment Styles Among Young Adults: A Test of a Four-Category Model, Journal of Personality and Social Psychology 61(2), 226-244 (1991) · full text; the four-category model (secure, preoccupied, dismissive, fearful)
Reis, Harry T., and Phillip Shaver — Intimacy as an Interpersonal Process, in S. W. Duck (ed.), Handbook of Personal Relationships (Wiley, 1988), pp. 367-389 · the closeness-intimacy model: self-disclosure + partner responsiveness
Aron, Arthur, and Elaine N. Aron — Love and the Expansion of Self: Understanding Attraction and Satisfaction (Hemisphere, 1986) · the self-expansion model
Hatfield, Elaine, and G. William Walster — A New Look at Love (University Press of America, 1981) · passionate vs companionate love; the labeling model
Waldinger, Robert, and Marc Schulz — The Good Life: Lessons from the World's Longest Scientific Study of Happiness (Simon & Schuster, 2023) · the Harvard Study of Adult Development, 85+ years
Berscheid, Ellen, and Elaine Hatfield (Walster) — Interpersonal Attraction (Addison-Wesley, 1969) · the founding monograph of attraction research
Jankowiak, William R., and Edward F. Fischer — A Cross-Cultural Perspective on Romantic Love, Ethnology 31(2), 149-155 (1992) · passionate love in 147 of 166 cultures; the universality question

Estimated time

beginner: 25m
intermediate: 50m
master: 85m