The experience is common enough that it has its own genre of late-night search. With one person the phone is just a phone. With another, the same twelve hours of silence produces a full internal case file. Same person, same nervous system, two entirely different weeks.
The usual explanations both feel wrong. One says the calm relationship is proof of growth, which does not survive the next relationship. The other says the anxious one revealed the truth, which does not explain the calm.
.90
year-long test-retest for attachment to a parent
.50
the same measure, same year, for a romantic partner
388
adults measured every week for a year
30
days of daily measurement in the companion sample
How much attachment actually moves in a year
In 2011 Fraley, Vicary, Brumbaugh and Roisman ran two unusually intensive studies. One sample of 203 adults reported their attachment representations every day for a month. Another sample of 388 did it weekly for a year. The point was to settle an argument about whether attachment behaves like a fixed prototype or like something that revises itself continuously.
The answer was both, in a specific proportion. There is a stable factor underneath, and it is real, and it survives. And the relationship-specific readings varied enormously depending on which relationship was being measured. Attachment security with mothers and fathers held test-retest correlations around .90 across a year. Security with a romantic partner sat around .50 over the same intervals.
That gap is the whole subject. Two measurements of supposedly the same construct in the same people across the same year, and one of them barely moves while the other moves a great deal. Whatever is being measured with a romantic partner is not behaving like a trait. It is behaving like a reading taken in conditions.
What that means for the word secure
The four names describe quadrants on two dials, and the dials were designed to capture a general tendency. Security, as most people use the word about themselves, is not that. It is a description of a current state: how much alarm is running, how available the other person seems, whether a need can be brought forward without calculation first.
Fleeson's work on traits as density distributions of states makes the same point in the general personality literature. Across two to three weeks of ordinary life the typical person expresses nearly every level of every trait, while the average across those moments stays almost perfectly stable. The trait is the distribution. The moment is a draw from it.
| What is being described | How stable | What moves it |
|---|---|---|
| The general tendency, measured across relationships | Moderately stable across years | A long relationship, long therapy, major loss |
| The reading with one specific person | Varies substantially within a single year | That person's consistency, what is at stake, the state of the week |
| The state on a Tuesday night | Unstable by construction | Sleep, the last conversation, whether a reply arrived |
Most of the internet writes about the first row and most people are asking about the third.
Trust splits the same way and is measured worse. A questionnaire returns a general readiness to extend it, which belongs in the first row, while the thing being worried about almost always belongs in the second. What a trust test can and cannot tell you works through the gap, including the large dataset showing that relationship-specific reports carry more than twice the predictive weight of anything measured about the person alone.
Why the calm relationship feels like a different self
Whatever is happening in the relationship that produces the secure state, it is not a personality change. It tends to be a set of conditions, and they are more ordinary than the experience of them suggests.
The other person is predictable in the low-stakes ways. Not grand gestures, which do nothing for this, but the boring kind: replies arrive in a recognisable rhythm, plans hold, and the weather does not change without explanation. Predictability is the input the system was built to read, and it reads volume rather than intensity.
Repair happens early and without ceremony. In relationships where something goes wrong and gets addressed the same evening, the alarm never gets a chance to compound, because the evidence arrives before the case has been built. Where repair is slow or absent, the same person rebuilds the case every time and the reading drops.
And the cost of needing something is low. This is the one people notice last. In some relationships a request is just a request. In others it is an event, and after a few of those the requests stop being made, which is read from inside as being more relaxed and is in fact the system going quiet because asking has been priced out.
Where this reframe gets abused
Two ways, and both are common enough to be worth naming.
The first is the exception story. Feeling secure with one person becomes evidence that this person is uniquely right, which quietly converts a state produced by ordinary conditions into a fact about a soulmate, and makes the relationship load-bearing in a way that is difficult to survive. The conditions are reproducible. The reading is not proof of destiny.
The second is the opposite move. If security is just a state, then the anxious relationship was entirely the other person's doing and there is nothing to look at. Sometimes that is true. But the stable factor in the Fraley data is real, and it is what decides how much ambiguity a person can absorb before the alarm starts. Two people in the same relationship, given the same delayed reply, do not have the same evening.
The honest position sits in between, and it is less satisfying than either end. Something travels with a person and something is produced by the relationship, and the second is larger than the four-name vocabulary suggests.
What is actually changeable
The research on moving the general tendency is unglamorous and slow. A steady relationship given years, or a long therapeutic one, and in both cases the mechanism is the same: repeated evidence that the expectation was wrong. The entry on earned secure attachment goes through the evidence properly, including the part where the original earned-secure group reported more depressive symptoms than the continuously secure one, which is not how the term usually gets sold.
The state is more tractable. It responds to whether things get repaired the same evening, to whether requests are cheap, and to how much of the ambiguity is being interpreted rather than checked. None of that requires becoming a different kind of person, which is fortunate, because the timescale for that is decades and the timescale for this week is this week.
The person who is calm in one relationship and not in another has not been inconsistent. They have been measured twice, in two different rooms, with an instrument that was never claiming to be immune to the room.
Sources
- Fraley, R. C., Vicary, A. M., Brumbaugh, C. C., Roisman, G. I. (2011). Patterns of stability in adult attachment: an empirical test of two models of continuity and change. Journal of Personality and Social Psychology, 101(5).
- Fleeson, W. (2001). Toward a structure- and process-integrated view of personality: traits as density distributions of states. Journal of Personality and Social Psychology, 80(6).
- Roisman, G. I., Padrón, E., Sroufe, L. A., Egeland, B. (2002). Earned-secure attachment status in retrospect and prospect. Child Development, 73(4).
- Roberts, B. W., DelVecchio, W. F. (2000). The rank-order consistency of personality traits from childhood to old age. Psychological Bulletin, 126(1).
- Mickelson, K. D., Kessler, R. C., Shaver, P. R. (1997). Adult attachment in a nationally representative sample. Journal of Personality and Social Psychology, 73(5).