Couples therapy can improve relationship satisfaction and communication on average, according to research reviews. Results vary, and an average benefit is not a promise of reconciliation. The strongest question is whether an appropriate, safely delivered approach helps your particular goals, with progress reviewed over time.
What research shows
| Research | What the study found | Important limit |
|---|---|---|
| Roddy and colleagues, 202058 studies; 40 unique samples; 2,092 couples | Relationship satisfaction improved substantially within treated groups from before to after therapy: Hedges’ g = 1.12. Waitlisted groups showed little average change. | The sample consisted exclusively of opposite-sex couples. A before-and-after effect does not isolate treatment from every other influence. |
| Rathgeber and colleagues, published online 201833 randomised studies; 2,730 participants | Behavioral and emotionally focused approaches produced an overall controlled effect of g = 0.60 after treatment. Their difference was not statistically significant. | Publication bias may inflate the estimate. Longer-term evidence was limited, with no EFT studies in the 12-month comparison. |
An effect size describes the size of a measured difference; it is not the percentage of relationships “saved.” The two figures above are not a contest between reviews: one describes change within treatment groups, while the other adjusts change against control conditions. Different populations, comparators and follow-up periods also matter.
Roddy and colleagues found improvements in several relationship domains and generally maintained gains at follow-up. Their stated sample limitation means the findings should not be assumed to represent every relationship structure or identity. Read the study abstract and publication details.
The Rathgeber review found a smaller controlled effect at six months. Its 12-month result concerned behavioral therapy only and did not show a significant advantage over the included alternatives. That narrow result does not prove that every couple loses all benefit after a year. The authors also cautioned about publication bias. Read the research paper (PDF).
Does this establish a best therapy method?
No. These selected reviews support a potential benefit from couples therapy, but they do not establish one brand or clinic as universally superior. The behavioral category in a research review is not interchangeable with every service marketed under a method name. A certificate, satisfaction survey or provider testimonial also answers a different question from a controlled trial.
Ask the practitioner which approach they propose, why it fits the concern and what would lead them to adapt it or refer you elsewhere. For example, a couple seeking a clearer separation decision needs a different immediate goal from a couple jointly working on repairing trust. Neither should be promised a particular relationship outcome.
Distinguish different kinds of claims
“Clients liked the service” concerns experience. “Relationship distress decreased” concerns a measured outcome. “This treatment outperformed a comparison” requires a suitable study design. “This will save your marriage” is a promise that none of those findings establishes.
AAMFT’s evidence update describes research on couple and family interventions across different problems. Its scope is broader than ordinary relationship counseling, and findings should not be applied to every method, provider or couple. Read the professional evidence overview.
Questions behind a success percentage
Ask which clients were counted, how many supplied follow-up data, whether people who stopped early were included and what counted as success. A survey immediately after a pleasant residential stay is different from a later assessment of everyday functioning.
Look for the comparison group, follow-up duration and relevant limitations. A published paper may study a particular population or an additional clinical condition. If a provider cannot explain the source of a percentage, treat it as unverified rather than as a basis for choosing care.
Compare approaches without choosing by brand alone
EFT and the Gottman Method describe different frameworks, but the name does not verify the clinician’s underlying professional standing, supervision or fit for your concern. Ask what assessment would lead them to recommend another service.
The therapeutic relationship also deserves practical attention: can both people ask questions, understand the plan and report a problem? A powerful session is not automatically a useful outcome. Nor is a difficult conversation automatically evidence that treatment is harmful. Look at purpose, consent and what changes afterward.
Choose meaningful goals for your circumstances
A goal might be a clearer decision about the future, less escalation, a workable parenting agreement or better follow-through. Remaining together is not the only possible useful outcome. Write goals in terms that can be discussed without turning one partner into the other’s assessor.
Review the plan with the clinician if the goals remain unclear or progress is absent. Seek individual specialist support where fear or coercion prevents honest participation. This guide is an evidence-reading framework, not an independent assessment of a particular provider’s results.
Start by asking what outcome was measured
Research can examine relationship satisfaction, observed interaction, individual symptoms or whether partners remain together. These are different outcomes. A service may improve one without changing another. Before interpreting a positive result, identify what the researchers measured and whether it matches the concern you want help with.
Also ask who reported the outcome. A partner’s own account, a therapist’s judgment and an observer’s rating provide different perspectives. Disagreement between them may be informative. For a practical decision, look for whether both partners’ experiences were considered and whether the measure captures something meaningful in ordinary life.
For an illustrative example, a couple may report fewer arguments because they can resolve a planning problem more easily. Another may report fewer arguments because they have stopped discussing it. The same headline description can represent different experiences. A useful assessment considers context and consequences as well as a numerical change.
Understand what the comparison condition means
A before-and-after study asks whether people changed over time. A controlled study also asks how that change compares with another condition. The comparator could be a waiting list, usual care or a different treatment. Each answers a different question, so two effect sizes should not be placed in a league table without examining the designs.
A waiting-list comparison may help show benefit beyond waiting during the study period. A comparison with another active treatment addresses a more demanding question about relative effects. Neither design automatically tells you which local clinician to choose or whether an intensive retreat delivers the same service as the treatment studied.
When authors report no statistically significant difference, that is not automatically proof that the treatments are identical. The study may have limited precision or may not have been designed to establish equivalence. Read the actual conclusion and the uncertainty around the estimate rather than translating “not significant” into a sweeping judgment.
Notice who is missing from the final result
People may leave treatment, decline a later survey or be excluded from a study before it begins. Ask how missing data and early departures were handled. Results based only on those who completed every appointment can answer a different question from results considering everyone originally allocated to treatment.
Here is an invented example of why the denominator matters. A program enrolls 40 couples and receives positive follow-up reports from 18 of the 20 couples who respond. Saying that 90% of respondents reported benefit is different from claiming that 90% of all 40 couples benefited. The experiences of the nonresponders remain unknown; they should not automatically be counted as either successes or failures.
Check eligibility too. If a study excludes particular safety concerns or clinical conditions, its findings should not be stretched to establish suitability in those circumstances. A research sample may also differ from you in age, relationship structure, culture or access to ongoing support. That affects confidence in applying the result, not the legitimacy of your need for help.
Size, certainty and practical importance are separate
A large estimate from uncertain evidence may be less dependable than a more modest, consistently supported effect. The Cochrane Handbook’s discussion of evidence certainty considers issues including bias, inconsistency, indirectness, imprecision and publication bias. These are reasons to examine how much confidence a finding deserves, rather than treating publication itself as a quality guarantee.
A statistically detectable change also needs interpretation in the context of daily life. Ask whether the study explains what the measured difference means for participants. For your own care, the relevant question might concern whether a difficult discussion is manageable, whether an agreement is kept or whether a decision is clearer.
The Cochrane guidance on interpreting results emphasizes considering the evidence and its limitations when drawing conclusions. A careful account can acknowledge benefit and uncertainty at the same time. It does not have to reduce the answer to either “therapy always works” or “the evidence proves nothing”.
Keep evidence for a method separate from a provider’s results
A clinic may cite research on an established approach while offering a different combination of sessions, staffing, hospitality and optional activities. Ask which parts of its actual program resemble what was studied. A method’s evidence does not automatically validate every package using its name.
If a service publishes its own outcomes, ask for the measurement method, dates, number of participants and follow-up. Independent analysis can strengthen confidence, but a named evaluator alone is not enough; the underlying question and data still matter. Testimonials may help you understand what some clients liked, while remaining a selective account of experience.
Use research to ask better questions at assessment. Why does this approach fit the concern? What alternatives are reasonable? How will the clinician notice limited benefit or harm? When will the plan be reconsidered? Combining those questions with a clear account of your own goals provides a more useful decision than selecting whichever provider advertises the largest percentage.
Read the date and scope of a review together
The date a review was published is not necessarily the date of the newest study it included. Check its search period and eligibility criteria. A review restricted to one method, population or outcome may be useful while leaving other questions unanswered. A newer article is not automatically more reliable merely because it appeared later.
For a claim about a new service format, ask whether the supporting evidence actually studied that format. If the evidence is indirect or limited, the provider should explain that clearly. Uncertainty can inform a cautious, reviewable decision without becoming a reason to promise benefit or dismiss a potentially helpful option outright.