Three Sessions of Imagery Rescripting: The Protocol Is Ready, the Evidence Is Not
- In 66 outpatients with major depression already receiving routine CBT, three added sessions of imagery rescripting did not outperform three sessions of guided relaxation on the primary outcome: the condition x time interaction on the BDI-II was not significant, F(3, 182.87) = 1.88, p = .135.
- Both arms improved and held their gains. Within-group effect sizes for rescripting were d = 0.87 at post-intervention and d = 0.94 at 8-week follow-up, against d = 0.55 and d = 0.67 for relaxation. The trial could not establish that the gap between those two trajectories is real.
- Two memory-linked measures did separate: distress attached to the target memory, F(6, 336.58) = 2.21, p = .042, and perceived memory burden, F(6, 355.16) = 2.59, p = .018, both fell further under rescripting. Negative core beliefs about self, others and world did not, all ps >= .051.
- The authors report that their power calculation was accidentally built on a group difference of d = 0.75 rather than the intended d = 0.25, call it "(perhaps over-)optimistic", and conclude the trial "may, in retrospect, be considered a pilot study, although it was not designed as a formal pilot trial".
Imagery rescripting is the clearest clinical expression of the memory-updating idea: reactivate an autobiographical memory, drive it to its emotional peak, then insert an outcome that the original event never delivered. This trial from Freiburg, Lübeck and Amsterdam – Arnoud Arntz, who built the modern protocol, is a co-author – asked whether three such sessions bolted onto ongoing CBT beat three sessions of beach-and-forest relaxation for depression. The answer was no, and the way the authors report that no is more instructive than the result itself.
Where the trial actually landed
Sixty-six adults meeting DSM-5 criteria for current major depression on the SCID-5-CV, all already in routine CBT at a university outpatient clinic, were randomised to three sessions of imagery rescripting (n = 33) or three sessions of guided imagery relaxation (n = 33). Baseline BDI-II was 23.03 (SD 7.53) and 24.58 (SD 9.42). The trial was prospectively registered (DRKS00031495), sessions were video-recorded and independently rated for adherence, competence and treatment differentiation, and assessments ran to eight weeks after the last session.
Everyone got better. The main effect of time on the BDI-II was large and unambiguous, F(3, 182.87) = 25.26, p < .001, with BDI-II falling roughly six points and staying down (b = -6.06 at post, -6.85 at FU2, both p < .001). The pre-registered hypothesis – that rescripting would bend that curve further – failed. Model-based BDI-II means at 8 weeks were 15.04 for rescripting and 18.86 for relaxation, a between-group d = 0.45 with a confidence interval of [-0.07, 0.96] that comfortably includes zero. The one contrast that reached significance was at the 4-week follow-up (d = 0.70, 95% CI [0.17, 1.24], p = .017), an exploratory pairwise comparison sitting under a null omnibus interaction, which is exactly the kind of finding that does not survive replication. Response rates tell the same story: 41.9% versus 14.3% at 4 weeks looks decisive until you read the authors' own note that the confidence intervals around every proportion were wide and overlapping.
The most useful paragraph in the paper is the one about power. The team specified a group x time interaction of 0.25, intending to power for a small-to-medium difference. But a linear interaction of that size accumulates across four timepoints, implying a cumulative separation of 0.75 SD by the final follow-up – a large effect, and not one that psychotherapy comparisons against active controls produce. They caught the error, printed it, and drew the correct conclusion: the study was not adequately powered to detect the small between-group differences that this literature actually yields. Non-significance here is not evidence of equivalence. It is evidence that the question was asked with an instrument too blunt to answer it.
What you can take into the room
The protocol is the deliverable part, and it is unusually well specified. A 60-minute introductory session uses a semi-structured interview to surface three distressing autobiographical memories, weighted toward childhood, along with the core beliefs, emotions and perceived burden attached to each. In sessions 1 and 2 the patient re-enters one memory in imagery with full sensory and contextual detail and names the emotions and needs of the younger self; at the emotional peak the therapist enters the scene and changes the outcome – protection, comfort, confrontation where warranted – until the younger self is safe. In session 3 the patient enters the remaining memory as their present adult self, intervenes on the younger self's behalf, then re-experiences the scene from the younger self's point of view. Five graduate psychologists in advanced training delivered this after 15 hours of instruction plus supervision, which is a real signal: this is learnable inside a training rotation, not a decade of apprenticeship.
Tolerability was a non-issue. Patients rated the sessions as highly pleasant in both arms, 8.21 versus 8.44 out of 10, t(61.02) = -0.71, p = .483, with no hint that walking into a childhood memory at emotional peak was aversive relative to guided relaxation. Perceived capability to cope with distressing memories trended higher after rescripting (7.73 versus 7.01, p = .097, d = 0.42). The exclusion criterion was high suicide risk, defined as acute ideation with plan or intent; comorbidity was otherwise permitted, and routine CBT ran unstandardised alongside.
Here is the discipline this paper demands. What is established: three sessions of rescripting are feasible, well tolerated, deliverable by supervised trainees, and they reduce distress and burden tied to the specific memory worked on. What is not established: that any of this lowers depression more than a competent relaxation exercise delivered by the same warm therapist for the same amount of time. Note where the largest between-group effect in the entire trial landed – patient satisfaction on the ZUF-8, administered here in a truncated seven-item version (the authors limited it to items 1–7), 24.62 versus 21.97, t(56.59) = 3.46, p = .001, d = 0.90. Rescripting made a stronger impression on patients than on their BDI-II scores. That is worth something clinically, and it is not the same claim as efficacy. Use the protocol for what it demonstrably does: work a specific, identified memory that a patient is carrying. Do not tell them, or yourself, that it is the ingredient that will lift the depression.
The largest between-group effect this trial produced was on patient satisfaction, not on depression – rescripting made a stronger impression on patients than on their symptom scores.
With 33 patients per arm and a power calculation the authors themselves flag as built on an effect roughly three times larger than intended, this trial cannot distinguish a genuinely absent advantage from one it was never equipped to see. Routine CBT ran unstandardised in the background, the arms differed at baseline in age (32.5 versus 39.2 years), and follow-up stopped at eight weeks.