The science

Goal setting and if-then plans: why vague intentions fail

Why "I will stop" fails, what if-then plans do, and the meta-analytic catch when the habit you are fighting is already strong.

By the WINNR team · 17 October 2026 · 11 min read

You have already decided. Probably several times: in the morning, after a bad night, with real feeling. Then at eleven at night, in a specific room, holding a specific phone, the decision was simply not there.

That is not a motivation problem. You are not short of reasons. You are short of a mechanism that converts the decision into a behaviour at the moment it is needed, when you are tired and nobody is watching. This post is about that mechanism: what it does, how large the effect is, and the one finding that should make you less confident than most articles will.

The gap between deciding and doing is measurable

Psychologists call it the intention-behaviour gap. Across 422 studies summarised in ten meta-analyses, the average correlation between what people intended and what they later did was r = 0.53 [1]. Intentions get acted on roughly half the time.

The sharper test is experimental. If intentions cause behaviour, raising someone's intention should raise their behaviour by a similar amount. Webb and Sheeran found 47 experimental tests meeting that standard. A medium to large change in intention, d = 0.66, produced only a small to medium change in behaviour, d = 0.36 [2]. Most of the gap is created by people who did intend to change and then did not act [1]: a very well populated category.

So the lever is not more conviction. It is the handover between conviction and action.

What an if-then plan is, and what it does

An implementation intention is a plan written to a fixed grammar: if situation Y happens, then I will do X. Peter Gollwitzer formulated it, and it is deliberately narrower than a goal. A goal names an outcome. An implementation intention names a trigger and a single response.

The headline evidence is the meta-analysis WINNR already cites on its science page: across 94 independent tests covering 8,461 participants, forming an if-then plan produced a medium to large effect on goal attainment, d = 0.65 [3]. In the health subset the effect was d = 0.59 across 23 tests and 2,861 participants [3].

Two things do the work, according to the same review. The cue becomes more mentally accessible, so you notice the moment instead of realising afterwards that it happened, and the response becomes partly automatic, so starting it costs less deliberation. You are not building willpower. You are delegating the decision to the situation while you are clear-headed.

A plan is only a delivery system, though. In two studies, implementation intentions improved goal attainment only when the underlying goal intention was strong [4].

The catch that most articles leave out

Here is the finding that matters most for you. If-then plans work less well against strong habits.

Webb, Sheeran and Luszczynska tested this directly. In a laboratory task with twelve analysable participants, where a habitual response was interfering with performance, forming an if-then plan reduced the interference, but the benefit was smaller the stronger the habit was. In a one-month field study of 172 adolescent smokers aged 17 to 21, the plans reduced smoking among people with weak or moderate habits and did not reduce it among people with strong habits [5]. Both are small studies, in a different behaviour and a younger group than this post is written for. Their conclusion is in the paper's title: habit strength moderates implementation intention effects.

The pattern repeats in diet research, which covers both adding and subtracting a behaviour. A meta-analysis of 23 studies found if-then plans worked better for promoting healthy eating, d = 0.51, than for reducing unhealthy eating, d = 0.29 [6]. Stopping is harder than starting. More generally, the more automatic a behaviour has become, the less your intentions predict whether you do it, because the context triggers it before intention is consulted [1].

Put that together honestly. The d = 0.65 headline is not your number. Yours is lower, possibly much lower, and the better established the pattern is, the lower it goes. No trial has tested implementation intentions on pornography use at all, so treating porn as the kind of cue-driven habit these studies measured is itself an extrapolation.

What follows is not "do not bother", but: do not expect the plan to carry the whole load. Making the behaviour slower to reach is the other half of the job, covered in our post on whether porn blockers work.

Why "I will not watch porn" is the wrong shape

The natural plan for a behaviour you want to stop is a negation: if the urge comes, then I will not open the site. There is experimental evidence that this specific form backfires.

Adriaanse and colleagues compared negation plans, "if situation, then not habitual response", against replacement plans and against the intention alone. The negation plans produced ironic rebound effects on both cognitive and behavioural measures, meaning they strengthened the very habit they were written to break. The rebound was strongest when the habit was strong [7].

So the plan has to name something you will do. "If I am in bed with my phone after midnight, then I will put it on the kitchen counter" is a plan. "Then I will not open the browser" is a plan that may work against you.

Approach framing, on thinner evidence

Goal framing points the same way, on weaker evidence. In a one-year Swedish experiment, 730 people's resolutions could be classified as approach or avoidance. Those with approach-oriented goals, framed as something to do, reported success more often than those with avoidance-oriented goals, framed as something to stop: 58.9 per cent against 47.1 per cent, a small effect, V = 0.11 [8]. Participants were randomised to different levels of support, and one of those groups was asked to phrase its goals as approaching something rather than avoiding it, so goal wording is partly confounded with condition. Success was also self-reported and the sample self-selected.

Treat it as a nudge in the same direction as the negation finding, not as proof: "I am building evenings that do not need it" beats "I am quitting porn".

Does goal-setting theory transfer to a stopping goal?

Locke and Latham summarised nearly four decades of goal-setting research: specific, difficult goals beat telling yourself to do your best, tested on well over 100 tasks with more than 40,000 participants, with meta-analytic effect sizes between d = 0.42 and d = 0.80 [9].

Now the fine print. That evidence is almost entirely about performance on tasks, in workplaces and laboratories, where output is measured and feedback is available. It is not about stopping a private behaviour at midnight. The same paper reports that goal effects shrink as tasks get more complex, d = 0.48 for the most complex against d = 0.67 for the least, and that on novel tasks a learning goal, aimed at working out how to do the thing, can beat a performance goal [9].

Changing an established private behaviour is a complex, novel task, so the transfer is partial. What carries over is specificity and feedback: "fewer than three times this week, logged" is usable, "cut down" is not. What does not carry over cleanly is difficulty. A hard target plus a scoreboard that resets to zero after one slip is a recipe for quitting the project rather than the behaviour, the subject of our post on why streak counting backfires. For a first month the learning goal is the better bet: find out exactly what your cues are rather than hit zero by Friday.

Mental contrasting and MCII

Gabriele Oettingen's addition is to put something in front of the plan. You name the wish, imagine the best outcome vividly, identify the main inner obstacle, and only then write the if-then plan. The combination is called mental contrasting with implementation intentions, or MCII, and the claim is that it filters out wishes you are not committed to.

The evidence is real and distinctly smaller than for plain if-then plans. A meta-analysis of 21 articles, 24 effect sizes and 15,907 participants found g = 0.336, 95% CI 0.229 to 0.443, with significant publication bias, a trim-and-fill correction bringing the estimate down to 0.242, and two included studies from a single report making up about 87 per cent of the pooled sample, which the authors excluded from their moderator analyses [10]. A second meta-analysis, restricted to health behaviour, found g = 0.28 across a total sample of 1,528, and in the seven studies that added implementation intentions to mental contrasting, adding them did not improve on mental contrasting alone [11].

The most relevant test is the least encouraging. In an online randomised trial of 346 smokers who wanted to cut down or quit, MCII produced no overall effect on cigarettes per day against a motivational control. A marginal interaction suggested a benefit among more dependent smokers, but it rested on a one-tailed test and the authors call their findings preliminary [12].

Worth the ten minutes, because naming the inner obstacle is how you find the real cue. Not worth confidence.

The closest thing to a direct test

One randomised trial has tested a structured self-help programme for problematic pornography use (PPU) that included goal setting: participants set a target for the coming week, and the final module covered planning for difficult moments [13].

In that trial 264 people were randomised, 123 to the programme and 141 to a waiting list. At six weeks the programme group scored lower on PPU, d = 1.32, and on use frequency, d = 1.65 [14]. Those numbers need a large caveat: 89.4 per cent of the intervention group dropped out, overall follow-up was 34.5 per cent, and the analysis used complete cases only, and because dropout was 89.4 per cent in the intervention arm against 44.7 per cent in the control arm, the effect sizes rest on 13 intervention participants and 73 controls. The programme also used goal setting and relapse planning rather than if-then plans in Gollwitzer's format [13].

The direct evidence: one trial, mostly men, six weeks, heavy attrition. Promising and unfinished.

Plans you could actually use

Everything below is a reasoned application of the research above, not a tested protocol. No study has validated this wording for this behaviour. What the research constrains is the shape: a cue you can genuinely notice, and a single action you will do rather than one you will not.

• If I get into bed with my phone, then I will put it on the kitchen counter and read until I fall asleep.
• If I am still awake and scrolling after midnight, then I will plug the phone in across the room and put the light out.
• If I shut the laptop for the night and feel restless, then I will put my shoes on and walk to the end of the street.
• If I am alone in the house on a Sunday afternoon with nothing planned, then I will text one person and arrange something for the evening.

Four rules the evidence supports. Write two plans, not ten, aimed at your two most frequent situations. Name a cue that is a time, a place or a posture rather than a feeling, because you have to notice it in time. State a replacement action, never a negation [7]. And expect the plan to do less work the more automatic the pattern is [5][6], so pair it with changing the situation.

Three things sit outside this post. What the urge feels like while you hold it belongs to ACT for pornography use. Working out the thought and situation that precede use, if you cannot name a cue at all, belongs to functional analysis in CBT for pornography use. And if plans work for a fortnight and then stop, the question may be self-concept rather than planning: see identity change. The starting sequence is in our guide on how to stop watching porn.

Common questions

do if-then plans work for porn?

No study has tested them on pornography use. They have a medium to large effect across 94 tests of other behaviours [3] and a smaller effect when the competing habit is strong [5]. A reasonable bet with a discount applied, not a proven method.

what is an implementation intention?

A plan in the form "if situation Y happens, then I will do X", decided in advance. It names a trigger and one response instead of an outcome [3].

should my goal be zero or just less?

The research cannot tell you where to set the number. Goal-setting theory says only that specific goals beat vague ones [9], so a goal you can measure beats "cut down", and a target you abandon after one bad night is worse than an easier one you keep.

how many if-then plans should I make?

Start with two. The effect depends on noticing the cue, and the more cues you monitor, the less likely you are to catch any of them. Add more once the first two fire reliably.

does writing the plan down matter?

Every study we could find had participants write the plan, so nobody has tested writing it against holding it in your head. Write it down because that is the form that was tested.

WINNR is a science-based app for men who want to regain control over their porn use. Its planning layer works the way this post describes: your own cues, two plans at a time, a replacement action rather than a prohibition. The app has not launched and is expected in mid-November 2026, and a layer connecting users with psychologists and coaches is in development. What the evidence does and does not support is set out on our science page.

A note on what this is

WINNR is not a treatment or therapy. If porn use is seriously affecting your life, speak to a licensed professional; if you are in crisis, contact your local emergency services.

Want a plan for your own pattern?

Take the private 4-minute check

Fighting an urge right now?

Open the urge page

#

Want a plan for your own pattern? Take the private 4-minute check. Fighting an urge right now? Open the urge page.