A forum-proposed lecture, scheduled for the week the capstone is due. What separates a technique that survives contact with a real week from one that only works in a study, and how big these effects actually are.
3 May 2027
Where. Outdoors, venue posted by the Friday before. Opens with the ten-minute detox.
This one came off the forum. Enough people proposed some version of “how do we know if any of this actually works” that it got worked up into a lecture, and it landed here because the capstone is due this week and this is the material you need for it. Like the rest of the back half, it can be displaced if the room votes for something else.
You wrote if-then plans in the crit this morning. Now the research behind them, and then the more useful question, which is how big any of this is.
The structural trick#
Gollwitzer’s distinction is between a goal intention, “I intend to reach outcome X”, and an implementation intention, “if situation Y happens, then I will do X”. The second one specifies when, where and how in advance.
The claim is that this moves the work. A goal intention needs you to decide again, in the moment, while tired and already spiralling. An implementation intention hands the deciding to the situation. You did the thinking in advance, when you had some.
The evidence. Gollwitzer and Sheeran (2006, Advances in Experimental Social Psychology) meta-analysed 94 independent tests drawn from 63 reports and found a medium-to-large effect on goal attainment, d = .65. That is large for this field, and much larger than most things on this syllabus.
Three caveats, because a number that good deserves them.
First, the attribution gets muddled constantly. The figure is routinely credited to Gollwitzer’s 1999 paper, which introduced the idea. The quantitative claim is from the 2006 meta-analysis.
Second, 2006 predates the publication-bias tools the field now runs as standard. A meta-analysis from that era is not automatically wrong, but it was assembled before anyone was routinely checking whether the small null studies had been published at all.
Third, there is a newer synthesis. Sheeran, Listrom and Gollwitzer (2024) aggregated 642 independent tests and, notably, declined to report one headline number, framing the effect as conditional instead. When a literature grows sevenfold and stops quoting a single figure, that is usually the literature telling you something.
Now the part that should change your capstone#
Line up what this course has measured.
| Technique | Roughly |
|---|---|
| If-then plans | d = .65, from a 2006 meta-analysis with the caveats above |
| Worry postponement | d = 0.31 on duration, d = 0.19 on frequency |
| A single bout of exercise on state anxiety | about a quarter of a standard deviation |
| Expressive writing | r = .075, about d = 0.15 |
| Mandatory attendance policies | d = .21, from three studies |
| Sugar on mood | nothing, and worse alertness within the hour |
Most of what works, works a little. That is not a reason for despair, it is a design constraint, and it is the single most useful thing to carry into the capstone brief.
A few things follow from it.
Small effects need cheap interventions. If the lift is d = 0.2, an activity that costs an hour and a lot of nerve is a bad trade. Ten minutes and no disclosure is a good one.
Averages hide people. A small mean effect is equally consistent with a technique that transforms a few people and does nothing for most. The moderators in the writing literature are the visible edge of this. Designing for “who is this for” beats designing for everyone.
Watch what you are crediting. The clearest lesson in the course is week 10’s biscuit. You had a break and credited the sugar. Every intervention you design has a plausible boring ingredient sitting next to the interesting one, usually a stop, a limit, or another person.
Where the honesty runs out#
I cannot tell you the effect size of anything invented in this room, and neither can anybody else. The course’s own central claim, that a loop repeats because it never gets the conditions to finish, has never been tested directly. A capstone that designs something sensible, states plainly what it is betting on, and says what would show it had failed, is doing the same thing this course does. That is the standard, not proof.
Discussion#
- If most effects here are small, is the honest version of this course “these things help slightly”, and would you have enrolled in that?
- The if-then number is the biggest on the syllabus and also the oldest and least re-examined. How much should age and scrutiny discount a finding?
- Pick any technique from the table. What is its boring ingredient?
- What would it take to convince you that something you invented worked, if you could not run a trial?
Teaching team
- Marisol Quaye, convenor