Relationship

What four trials of emotionally focused couples therapy actually reported

Published 2026-04-28 · 6 min read · By LovePinnacle Editorial

Quick answer

A 1999 review of four trials reports effect sizes of 1.28 overall and 1.31 on the Dyadic Adjustment Scale. In most of the distress studies, over half of treated couples were no longer distressed. The authors also say routine clinics may not match.

Emotionally focused couples therapy first appeared in the research literature in 1985, in a paper by Johnson and Greenberg. Fourteen years later, S. M. Johnson, J. Hunsley, L. Greenberg, and D. Schindler published "Emotionally focused couples therapy: Status and challenges" in Clinical Psychology: Science and Practice. That 1999 paper is where the assumptions and the trial numbers on this page come from. It is not a promise that you, or any couple you care about, will recover.

ICEEFT, the International Centre for Excellence in Emotionally Focused Therapy, calls EFT a humanistic approach that draws primarily from attachment theory. Its "What is EFT?" page applies the work to individuals, couples, and families, and it names those uses EFIT, EFCT, and EFFT. The page says peer-reviewed outcome research supports efficacy. It also uses a success slogan and gives no percentage. The sources cited here will not repeat that slogan. Wherever a number shows up below, it comes from the 1999 paper, and it stays tied to the limit the authors placed on it.

Assumption one: feeling and the pattern lock

The first assumption says emotional responses and interaction patterns determine each other, and both have to be addressed. You don't get to pick a feeling and ignore the loop, or coach the loop and ignore what each person feels. In the paper's framing, the two keep producing each other. A sharp sentence changes the feeling. The feeling changes the next sentence. Under this assumption, a therapy that only teaches a script has left half the work undone. A therapy that only names a feeling, and never touches what the two of you actually do, has left the other half undone.

Assumption two: stuck, not merely unskilled

The second assumption says partners are stuck in patterns that block responsiveness. They are not merely "unskilled." That word carries a lot of weight. If the problem were only a missing phrase, a class in phrasing would be the entire treatment. The authors' claim is harsher and kinder at the same time. The pattern itself gets in the way of responding. You can know a better sentence and still be unable to say it while the pattern is running. Skill talk, under this assumption, underestimates the trap.

You might be thinking this means nobody needs to learn a new way to talk. The assumption does not say that. It says a lack of skill is the wrong main story. New interactions are required later, in the third assumption. The second assumption only rejects the idea that you are a couple with a small vocabulary and an otherwise open bond.

Assumption three: new feeling and a new interaction

The third assumption says emotion is central to defining and redefining the relationship. New emotional experience and new interactions are both needed. One without the other does not satisfy the assumption. A new feeling that never shows up between you stays private. A new interaction with no new feeling behind it can look like a drill. The paper wants both, because emotion is how the relationship gets defined to begin with, and how it gets defined again.

This is still a theory of change inside a review, not a session plan you can run at the kitchen table tonight. The sources cited here contain no steps, no session counts, and no homework. If a site hands you a three-stage script and says it comes from this paper, that script is not in the published paper.

Assumption four: adult intimacy as attachment

The fourth assumption says adult intimacy is best understood as an attachment process. The published paper does not walk you through a childhood history or a set of types. It states the view directly: the closeness adults want from each other belongs with attachment, not off to the side as a separate skill. That view is why the first three assumptions put emotion and responsiveness at the center of the work. For a plainer look at everyday closeness, signs of emotional intimacy is a different page, and it is not a result from these four trials.

Four randomized trials, and what was pooled

The meta-analysis covered four randomized trials of distressed couples: Goldman and Greenberg, 1992; James, 1991; Johnson and Greenberg, 1985a; and Walker and colleagues, 1996. Across outcome measures, the overall mean effect size was 1.28. Using only the Dyadic Adjustment Scale, the weighted mean effect size was 1.31, with a Z of 6.42.

The four DAS effect sizes, printed with the comparison counts next to them, were not the same size. Johnson and Greenberg, 1985a, is listed at 2.19, with 15 versus 15. James, 1991, is listed at 0.70, with 14 versus 14. Goldman and Greenberg, 1992, is listed at 1.52, with 15 versus 15. Walker and colleagues, 1996, is listed at 1.27, with 16 versus 16. The average is real, and so is the spread. A mean of 1.31 can hide a trial at 0.70 and a trial at 2.19. If you remember only the mean, you will sound more certain than the table does.

In the marital-distress studies, EFT improved scores on the Dyadic Adjustment Scale compared with a wait-list and compared with pretreatment. In most of those studies, over half of treated couples met recovery criteria from Jacobson and Truax, 1991, meaning they were no longer in the distressed range. Deterioration was infrequent. "Over half" and "in most of those studies" are the phrases the paper supports. They do not mean "everyone," and they do not mean "your couple."

You might be thinking a large effect size is a promise with a decimal point attached. It is an average difference on the measures those trials pooled, in samples the authors themselves call small. It does not assign you an outcome. Reading it as a forecast skips the next section, which the authors wrote for a reason.

Limits they state, including the clinics you might actually call

Johnson or Greenberg were investigators on almost every one of the studies, and the authors warn that loyalty to a method can push effects upward. The samples are small, which you can see for yourself in the counts next to the four effects. They also warn that a busy clinic, with less training and more complicated cases, might not see the same effect. Studies that take a treatment apart to find which piece does the work had not been done yet, so that test simply is not in the sources cited here. The 1.31 figure rests on only four studies, and the authors say it might not hold steady. Next to that warning they report a fail-safe n of 49. You don't need the formula to hear what they are telling you: four trials make a thin shelf, and they say so themselves.

Now picture the clinic sentence in your own town. If the people who built the model also ran the trials, a community office with less training and harder cases may not get the same effect. The sources do not show it failing there. They show that the 1999 authors would not promise it holds there. If you keep only one limit in mind, make it that one.

Nothing on this page tells you to start this model, to stop another treatment, or to expect recovery. A decision about therapy belongs with a licensed clinician who can hear your history, including whether anyone is being harmed. This article is education about a published review. It cannot assess you.

If you sat down with that clinician, which of the four assumptions would you want held up against your actual week?

Papers and pages for the trials

Books we recommend

As an Amazon Associate I earn from qualifying purchases.

Frequently Asked Questions

What four ideas does the 1999 review assume?
Feeling and the interaction loop shape each other, so both need attention. People are caught in patterns that block responsiveness, rather than merely short on skill. Emotion defines the bond, so a fresh emotional experience and a fresh interaction are both required. Adult closeness is best understood as attachment.
Which pooled figures should you keep?
The average across outcomes is 1.28. Restricted to the adjustment scale, the weighted average is 1.31 and the reported Z is 6.42.
What sizes sit next to the four scale effects?
Johnson and Greenberg 1985a is 2.19 with 15 against 15. James 1991 is 0.70 with 14 against 14. Goldman and Greenberg 1992 is 1.52 with 15 against 15. Walker and colleagues 1996 is 1.27 with 16 against 16.
How many treated couples left the distressed range?
In most of those marital studies, more than half met Jacobson and Truax recovery criteria. Getting worse was uncommon. That record is not a forecast for the couple reading this.
Why might a community office see something smaller?
Originators helped run nearly every trial, allegiance can puff results up, the groups were small, work that pulls the therapy into ingredients had not been done, and the authors say the average might not appear where training is lighter and cases are messier. A fail-safe n of 49 is part of their stability warning.

emotionally focused therapy · EFT · attachment theory · Sue Johnson · relationship framework

← All articles