Contribution
It moves attachment research from between-person styles to within-person variation across relationships, and offers self-determination theory’s needs as what explains that variation: a partner is a secure attachment figure to the degree that the relationship satisfies the person’s needs.
Key insight
Security is a property of a relationship as much as of a person. Most of its variance lies across a person’s partners, and need satisfaction with a partner tracks security with that partner, relationship by relationship.
Assumptions
- Self-report throughout. Felt security (M. T. Greenberg’s inventory, Study
- or Bartholomew and Horowitz’s four styles (Studies 2–3), and a need satisfaction scale per partner (15 items in Study 1, 9 in Studies 2–3, the appendix).
- College samples. N = 136, 152 and 160 University of Rochester undergraduates, mostly women (pp. 370, 373, 377).
- Cross-sectional. “Although the data in the current studies were all cross-sectional, we used an analytic strategy that suggests causation” (p. 381).
- The relatedness confound. Because relatedness satisfaction is conceptually close to security, every analysis was repeated with autonomy and competence only (p. 370).
Key results
- Variance within persons. 56% in Study 1 (felt security), 79% in Study 2 (overall security across six figures, 72% and 81% for the self and other dimensions), about twice the between-person share in Study 3 (pp. 371, 374, 379). Error is included in the within-person share, as the authors say.
- Need satisfaction predicts security within persons. HLM slopes, all p < .001: 0.81 (0.79 without relatedness) in Study 1; 4.54 (4.11) for overall security in Study 2; 5.23 (4.31) in Study 3 (Table 1). In every specific relationship, residual need satisfaction predicted residual security (Table 3: betas .59–.79 in Study 2, .60–.75 in Study 3).
- The needs differ. Entered together, relatedness was the strongest predictor in 15 of 18 regressions; without it, autonomy was stronger than competence in 15 of 18; competence added independent variance only for mother and best friend (pp. 375–376, 380).
- Well-being. Mean security predicted well-being (β = .60, .45, .30). Variability across partners did not, except in Study 2 with six figures (β = −.15) and for the other dimension (β = −.20); meta-analytically d = −0.12, CI −0.31 to 0.07, for overall security, and d = −0.28, CI −0.51 to −0.06, for the other dimension (p. 380).
- Mediation. Need satisfaction partly mediated security’s relation to well-being in Study 2 (β .50 → .34) and fully in Study 3 (.33 → .05); combined, partial (d = .34). Security partly mediated need satisfaction’s relation to well-being in Study 2 only; combined, partial (d = .33) (pp. 377, 379, 381).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Most variance in adult attachment security lies within persons, across partners | strong for these measures and samples | variance decomposition in three studies, error included |
| C2 | Relationship-specific need satisfaction predicts relationship-specific security | strong as association | HLM and residual regressions in every relationship, with and without relatedness |
| C3 | Security is a function of need satisfaction | weak | cross-sectional; the authors say the relation is “bidirectional” and that third variables may drive both (p. 381) |
| C4 | Variability in security across partners does not harm well-being | moderate | null in two of three studies; negative for the other dimension across two |
Concepts
- Within-person variability — variation in a person’s attachment across relational partners, estimated by multilevel models (p. 370).
- Need satisfaction with a partner — how well autonomy, competence and relatedness are met “when they are with” that partner (p. 373).
Connections
- Patrick et al. 2007 (LIT-1971, NOTE-1515) summarize it; their account checks.
- Bowlby; Hazan and Shaver; Bartholomew and Horowitz, the attachment tradition it extends.
- Kasser and Ryan 1996 (LIT-1977), whose residualizing method it adapts (p. 375).
Bearing on the record
- QUESTION-369. The need that best predicts the outcome, relatedness, is the one whose items are closest to the outcome in content, so the authors repeat everything without it (p. 370). That is the interdefinition worry at the level of an instrument: a need claim tested by an outcome partly made of the need. With relatedness removed the results hold, which is the paper’s answer, for autonomy and competence only.
- Need satisfaction is relational. The variance is mostly across partners, so a need’s satisfaction is a property of a person in a relationship. This bears on THEORY-052 and CLAIM-368 only as a reminder that a person-level frustration score averages over relations that differ.
- No instruction for machine-learning practice.
Limitations
- Cross-sectional, self-report, North American undergraduates, as the authors say (p. 381).
- The within-person share includes measurement error.
- Mediation by the Baron–Kenny–Sobel method on cross-sectional data, which cannot fix direction.
Open questions
- Does a change in need satisfaction with a partner precede a change in security, as the theory says, or follow it? The authors call for longitudinal work (p. 381).