What has been tested in randomised trials, and at what effect size

Descriptive research on delay runs to hundreds of studies. Randomised controlled trials of interventions run to a few dozen. That imbalance is the first thing to note about this part of the field: a great deal is recorded about what correlates with delay, and comparatively little has been tested under experimental conditions.

Internet-delivered cognitive behavioural therapy

The most substantial trial programme comes from Alexander Rozental and Per Carlbring at Stockholm University. Their 2015 trial in the Journal of Consulting and Clinical Psychology randomised adults reporting severe delay to guided internet-delivered CBT, unguided internet-delivered CBT, or a waiting-list control across a ten-week protocol. Both active conditions recorded reductions on self-reported delay measures relative to the waiting list, with the guided condition performing somewhat better than the unguided one. Follow-up assessments reported that gains were largely maintained.

Rozental and colleagues subsequently published a systematic review and meta-analysis of randomised trials in this area. The pooled estimate was in the small-to-moderate range, drawn from a small number of eligible trials with substantial heterogeneity between them and a non-trivial risk of bias in several.

Implementation intentions

A separate line derives from Peter Gollwitzer's work on implementation intentions — specifying in advance the situation in which an intended action will be carried out. Gollwitzer and Paschal Sheeran's 2006 meta-analysis pooled ninety-four independent tests and reported a medium-to-large effect on goal attainment.

Two qualifications matter for this catalogue. The meta-analysis concerned goal attainment broadly, not delay specifically; extending it to procrastination is an inference the original paper did not make. And it was published before the replication debates that reshaped much of social psychology, in a period when preregistration was uncommon and effect sizes in the pooled literature were, across many areas, subsequently revised downward. Studies applying the technique to delay directly, such as work by Elizabeth Owens and colleagues in 2008, used small student samples.

The self-forgiveness study

One frequently repeated finding deserves separate treatment. Michael Wohl, Timothy Pychyl and Shannon Bennett published a paper in 2010 titled "I forgive myself, now I can study", reporting that students who reported forgiving themselves for delaying before a first examination reported less delay before a second.

The paper is cited widely, often as though it demonstrated that self-forgiveness reduces delay. Its design does not support that reading. The sample was around 120 undergraduates at a single institution, the measures were self-reported at two time points, and the design was longitudinal-correlational rather than randomised. Nothing was manipulated. The result is consistent with the proposed explanation and equally consistent with the reverse — that students who happened to delay less afterwards reported feeling better about the first occasion. As far as the published record shows, it has not been replicated in a randomised design.

What the trial literature supports

Taken together: structured cognitive-behavioural protocols delivered over several weeks have been tested in randomised trials and record modest reductions on self-report measures against waiting-list controls. Most other techniques circulating in general writing have been tested in small samples, tested for a different outcome, or not tested experimentally at all.

The gaps are specific and known. Few trials use behavioural outcome measures. Few use active control conditions. Few follow participants beyond a year. Few recruit outside student and self-selected online populations. Researchers in the field have named these gaps repeatedly; closing them requires funding that a research area without a clinical diagnostic category attracts only intermittently.