Skip to main content
Category: Effects
Type: Overestimating one’s capacity to control a visceral impulse, such as fatigue, hunger, or a cigarette craving, and then allowing more of that temptation within reach
Origin: Loran F. Nordgren, Frenk van Harreveld, and Joop van der Pligt, “The Restraint Bias: How the Illusion of Self-Restraint Promotes Impulsive Behavior,” Psychological Science 20(12), December 2009, pages 1523–1528, DOI 10.1111/j.1467-9280.2009.02468.x. A six-page proof PDF was read. That file prints placeholder volume marks, so the page range is the journal record, not a number printed in the proof. The counted case is Study 4 of that paper
Also known as: Illusion of self-restraint, the phrase in the paper’s subtitle. It is not the Hot–Cold Empathy Gap as one score
Quick Answer — Restraint Bias is the tendency to overestimate one’s capacity for impulse control. Nordgren, van Harreveld, and van der Pligt named it in 2009. In a cold state, with the craving quiet, the capacity rating runs high, and people leave the cue closer. Among 55 smokers who had already quit for three weeks, that rating tracked less avoidance. It did not, by itself, track how much they smoked four months later.

​
What is Restraint Bias?

Restraint Bias is the tendency to overestimate one’s capacity for impulse control.
We argue that people generally, unlike Odysseus, exhibit a restraint bias: a tendency to overestimate one’s capacity for impulse control.
That sentence is Nordgren, van Harreveld, and van der Pligt (2009). Impulse control here means governing a bodily urge: fatigue, hunger, drug craving, sexual arousal. The overestimate is a belief about capacity. It is not a measurement of how strong the urge is. The paper’s claim is that the belief changes a strategy. People who rate their capacity higher leave more temptation within reach. The urge then has less distance to cross. A useful picture is a cake and a clock. Just after a meal, the cake looks easy to refuse, so it stays on the counter. Mid-afternoon, the same cake is a different problem, and the counter was chosen hours earlier. Study 2 is that picture with a numbered rank and a €4 stake. The Hot–Cold Empathy Gap is the difficulty of simulating a state you are not in. Restraint Bias is the capacity rating that follows, and the distance you then set. George Loewenstein’s 1996 chapter is the paper’s starting point for the gap. That chapter was not opened. Projection Bias carries today’s preference onto a later day. This page asks a narrower question: do you think you can refuse. The Overconfidence Effect is too much trust in the accuracy of a judgment. The Illusion of Control is too much trust that you can steer a chance outcome. The Planning Fallacy is a time estimate. A student who leaves reading until the last week can be using more than one of these. Study 1 counts only the fatigue rating and the share of work left for that week.

​
Restraint Bias in 3 Depths

  • Beginner: When a cue feels easy to refuse, notice whether the urge is quiet right now. The paper treats the quiet rating as the inflated one. The hungry, tired, or craving rating was lower, and it came with more distance from the cue.
  • Practitioner: Before a diet week, a quit week, or a heavy study week, write where the cue will sit: another room, the desk, the hand. Study 3 paid smokers on that ladder. The payment is the experiment. The ladder is the decision the paper actually varied.
  • Advanced: A high capacity rating is not, in these studies, the same number as the later lapse. In Study 4 the rating tracked avoidance, and avoidance tracked smoking status four months later. The direct correlation between the rating and that status was not significant. A rule that says “confidence causes relapse” is already wider than the table.

​
Origin

The name and the four studies are one paper. Nordgren was at the Kellogg School of Management, Northwestern University. Van Harreveld and van der Pligt were at the University of Amsterdam. The proof prints a received date of 28 February 2009 and an acceptance date of 21 April 2009. PubMed records the article as published online 30 October 2009, then in the December 2009 issue.
Recovering smokers with more inflated impulse-control beliefs exposed themselves to more temptation, which led to higher rates of relapse 4 months later.
That sentence is the abstract. It states a chain: belief, then exposure, then relapse. The results section of Study 4 prints the two links separately. Impulse-control beliefs were not related to smoking status four months later, r(55) = .16, not significant. The case below keeps both sentences. The abstract’s “led to” is the authors’ chain. It is not a second test in which the belief score itself predicted the later category. The paper starts from what it calls the empathy-gap effect. It attributes to Loewenstein (1996) a cold-to-hot empathy gap: a person who is not hungry, angry, or aroused underestimates how that state will change preference and behavior. The reason he gives, as this paper reports it, is that memory can store the occasion and the rank of the urge, and still cannot bring back the sensation. The 1996 chapter was not opened. The citations for sexual arousal, hunger, fear, and drug craving are Ariely and Loewenstein (2006), Nordgren, van der Pligt, and van Harreveld (2007), Van Boven, Loewenstein, and Dunning (2005), and Sayette, Loewenstein, Griffin, and Black (2008). Those four were not opened. One earlier study is described in enough detail to see the parent finding. Nordgren, van der Pligt, and van Harreveld (2006) had pain from ice water interfere with a memory test. People who later explained their score while free of pain underestimated the pain. People who explained it while in the ice water did not. That 2006 PDF was not opened. The sentence here is the 2009 summary. From that finding the authors predict three steps. In a cold state, capacity is overestimated. In a hot state, the same rating is more realistic. The higher rating leads to more exposure, and more exposure leads to more of the impulsive act. They add that people are usually cold. If the prediction holds, the overestimate would be the ordinary case, and caution would be what gets skipped. The introduction uses Odysseus and the sirens as the contrast. He expected he could not refuse the song, so he put wax in the crew’s ears and had himself tied to the mast. The authors’ point is that a higher expectation of refusal would have meant a weaker precaution. They do not cite a Greek edition. The story is their illustration.

​
Key Points

Four distinctions keep the capacity rating, the distance to the cue, and the later act from collapsing into one slogan.
1

The cold rating runs higher

Study 1 randomly assigned 72 students to a tiring memory task of nearly 20 minutes or a 2-minute version of it. The long task had been used to induce fatigue in the 2006 paper. Fatigue ratings were higher after the long task, M = 5.17 versus M = 3.87, F(1, 70) = 41.84, p < .001. On the paper’s impulse-control score, where a higher number means more perceived control, the fatigued group scored M = 5.09 and the rested group M = 5.60, F(1, 70) = 8.17, p = .005. Study 2 asked 79 people, a mix of students and employees, either as they entered a cafeteria or as they left. Entry was hungrier, M = 4.88 versus M = 2.43, F(1, 77) = 100.00, p = .001. The satiated group had the higher control score, M = 4.91 versus M = 4.32, F(1, 77) = 6.86, p = .01. The paper treats the hot rating as the more realistic one. It does not print a separate true score of capacity against which both groups could be graded.
2

The rating moves the cue

Study 1 asked what percentage of next semester’s studying would wait until the last week. Fatigued students said 52.68 percent. Rested students said 59.38 percent, F(1, 70) = 4.51, p = .04. Fatigue correlated with less intention to cram, r(72) = −.34, p = .005, and impulse-control beliefs fully mediated that link, z = −2.70, p = .005. Both means are still near half the semester. The paper’s prediction said “the majority of work.” The printed means are a slim majority, and the gap is about seven points. Study 2 had people rank seven snacks and keep one. Returning it uneaten in a week earned €4 plus the snack. Satiated people chose a more favored snack, M = 6.21 versus M = 5.47 on the rank from least to most favorite, F(1, 77) = 10.76, p = .002. The paper calls 6.21 a first or second favorite and 5.47 a second or third. Greater perceived control correlated with the more tempting choice, r(79) = .35, p = .002.
3

Distance is what the later act followed

In Study 2, 39 people returned the snack. The return rate was 60.5 percent in the hungry condition and 39.0 percent in the satiated condition, chi-square(1, N = 79) = 3.65, p = .06. The paper says that difference was not significant. People who returned the snack had chosen a less favored one, M = 5.51 versus M = 6.20, F(1, 77) = 8.95, p = .004. Snack choice fully mediated the link between hunger and returning, z = −2.05, p = .04. Study 3 assigned 53 student smokers to false feedback from a bogus implicit-measures test, which the paper says was an Implicit Association Test. High-control feedback produced a higher craving-control score, M = 4.90 versus M = 4.28, F(1, 51) = 4.50, p = .04. During Coffee and Cigarettes (Jarmusch, 2003), they chose a cigarette in another room (€2), on the desk (€4), in the hand (€6), or in the mouth (€8), and were paid only if they did not smoke. High-control participants chose more temptation, M = 3.00 versus M = 2.38, F(1, 51) = 5.54, p = .02. The paper describes those averages as a cigarette in the hand and a cigarette on the table. Failure to abstain was 33.33 percent versus 11.52 percent, chi-square(1, N = 51) = 3.59, p = .06. The paper calls that difference significant and prints p = .06. The test’s N is 51. The assignment line says 53. The proof does not say where the other two went.
4

A direct arrow is a further claim

Study 4, the case below, is the field version of the same chain. Higher impulse-control beliefs went with less avoidance of smoking cues, r(55) = −.32, p = .02. Those beliefs were not related to smoking status four months later. Avoidance was. A sentence that skips the middle step is the abstract’s wording, not the correlation table. The discussion also refuses a simple war on confidence. It cites a view, Bandura (1982) and Baer, Holt, and Lichtenstein (1986), that self-efficacy helps self-control. It then says unrealistic control can hinder. It leaves open, citing Rothman, Baldwin, and Hertel (2004), that a higher sense of control may help a person start, while a realistic belief may help the person continue. Those four works were not opened. This page does not turn that sentence into a program.

​
Applications

These four moves use splits the 2009 studies actually make. They are not a claim that a worksheet changes a clinic’s quit rate, a class’s grades, or a household’s snacks by a measured amount.

Set the distance while the urge is quiet

The cold moment is when the paper finds the high capacity rating, and it is also when the location gets chosen. Write the location before the craving, the hunger, or the late-night fatigue. Study 3’s ladder was another room, the desk, the hand, and the mouth. Use the ladder as a list of distances. Do not copy the euro payments into a budget.

Ask for the rating twice

If a plan is made just after a meal, a cigarette, or a rest, add a second line for the same plan during the urge. Study 1 and Study 2 are that contrast. The hot rating was lower, and the hot group left itself less temptation. The paper does not show that the hot rating is a person’s permanent trait.

Score avoidance apart from confidence

Study 4’s belief score did not have a significant direct link to later smoking. The avoidance items did. Those items were staying away from people who smoke, refusing the idea of an occasional cigarette, keeping cigarettes out of the house, car, and office, avoiding them when stressed, and asking others not to smoke nearby. A review that records only “I can handle it” has skipped the row the follow-up tracked.

Leave the slogan off the meeting

The discussion says unrealistic control can hinder ongoing self-control, and it also says confidence may help people begin. Alcoholics Anonymous is their illustration of a program that repeats powerlessness, not a measured result. A meeting that only lowers confidence, or only raises it, has left that paragraph.

​
Case Study

Study 4 asked whether the same chain appears after the early withdrawal of quitting. The authors call it one of the biggest puzzles in addiction research that relapse often occurs after physical withdrawal symptoms cease. Their account is that once those cravings fade, a recovering smoker overestimates the capacity to handle cues such as old places, and then seeks those cues out. They contacted 55 people through a large smoking-cessation program, three weeks into the program. Eligibility required abstinence for all three weeks. At that point each person gave the smoking version of the impulse-control score, the same sort of score as in Study 3, and rated five avoidance statements. Four months later they were asked which category fit their smoking: not at all, once or twice a week, once or twice a day, half a pack a day, or a pack a day or more. The belief score did what the exposure half of the prediction said. Smokers who rated their capacity higher reported less avoidance, r(55) = −.32, p = .02. The same score did not relate to smoking status at the follow-up, r(55) = .16, not significant. Avoidance did. Less avoidance predicted a higher category on the five-point smoking measure, beta = −.76, SE = .29, t(55) = −2.61, p = .01. The follow-up for the whole sample was 32.71 percent still abstinent, 18.2 percent smoking once or twice a week, 12.73 percent once or twice a day, 23.62 percent half a pack a day, and 12.74 percent a pack a day or more. Those five shares are everyone who was followed. They are not the relapse rate of a high-belief half and a low-belief half. The paper does not print that split. The lesson the authors draw is that natural differences in the capacity belief changed how people arranged temptation, and that the arrangement tracked later smoking in this group. The boundary is in the same discussion. Only people who had already lasted three weeks were eligible. Many people who try to quit never last that long, so the authors say the finding may fit smokers who already have some ability to refrain, and may not fit every attempt. Avoidance and later smoking were both reports. The study does not randomly assign a belief and then count cigarettes. It does not test a clinic protocol. The percentages and the coefficients belong to these 55 people, this program, and this four-month window.

​
Boundaries and Failure Modes

Restraint Bias does not say that a confident person will relapse, or that a cold rating is false in every domain. Study 4’s direct correlation with later smoking was not significant. The path the table supports runs through avoidance. Using the abstract’s “led to” as if the belief column itself predicted the five categories adds a test the results section does not print. Two other contrasts sit on p = .06. The cafeteria return rates, 60.5 percent versus 39.0 percent, are the ones the paper calls not significant. The film failure rates, 33.33 percent versus 11.52 percent, are the ones it calls significant, with the same printed threshold. A summary that promotes both to p < .05, or that drops the film result because it missed .05, is editing the proof. Study 2 was not randomly assigned. People entering a cafeteria can differ from people leaving it in ways hunger does not cover. The authors say so. The discussion’s heroin sentence is a citation, not a fifth study of this bias. Badger and colleagues (2007) are reported as finding that heroin users valued an extra dose of the substitute buprenorphine more while craving than while satiated. That PDF was not opened. No price from it appears here. The Alcoholics Anonymous sentences are the authors’ example of a practice that keeps repeating powerlessness, because they suspect the cold state pulls people back toward the inflated belief. No meeting was counted. A further misuse is to export any of these figures onto another year, another clinic, a class, a hiring meeting, or a household rule. The student gap is about seven points on a 0-to-100 plan for one next semester. The snack stake was €4. The film stakes were €2, €4, €6, and €8. None of those is a recommended price, a quit rate, or a share of people who will fail a diet.

​
Common Misconceptions

Three readings add a result, or a merge, that the 2009 paper does not print.
The paper uses the empathy gap as the reason a cold person misjudges an urge. Restraint Bias is the next belief: capacity to control that urge. The studies then measure a behavior the gap papers, as this article describes them, were not built around. Students set a study share. Cafeteria visitors picked a snack and either brought it back. Smokers chose how close a cigarette sat, and some of them smoked it. The Hot–Cold Empathy Gap page was not edited. The two names stay on two pages because the questions differ. One asks whether the quiet mind can simulate the urge. The other asks whether the capacity rating moves the cue, and whether the cue’s distance shows up in the later act.
The shares 32.71, 18.2, 12.73, 23.62, and 12.74 percent describe all 55 people. The paper does not split them by high and low impulse-control beliefs. The belief score’s correlation with smoking status was r(55) = .16 and was not significant. What differed with the belief was avoidance. A split of those five shares into a high-belief relapse rate and a low-belief relapse rate is not printed.
The film failure contrast and the cafeteria return contrast are each printed at p = .06. The authors use different words for them. They say the return-rate difference was not significant. They say the smoking-failure rates differed significantly. The snack-choice mediation of returning is a different test, z = −2.05, p = .04, and that one is what they treat as the path from hunger to the week-later act. Quoting only the percentages, without the p values and without that mediation, makes the condition look like a settled return gap. The proof does not say that.
These six pages sit beside Restraint Bias. Each one asks a different question.

Hot–Cold Empathy Gap

Underestimating a visceral state while you are outside it. This paper uses that gap as the reason the capacity rating inflates. The rating and the exposure are the added claims.

Projection Bias

Letting today’s preference stand in for a later day’s preference. A cold forecast of hunger is in that neighborhood. The question here is the refusal rating, and where the snack then sits.

Optimism Bias

A tilt toward expecting the better outcome. A high restraint rating can feed that tilt. Study 4’s optimism, if one wants that word, did not itself correlate with the later smoking category.

Overconfidence Effect

Too much trust that a judgment is accurate. Restraint Bias is trust in a specific capacity, the capacity to govern an urge, followed by a change in exposure.

Illusion of Control

Too much trust that you can steer a chance outcome. A cigarette in the hand is not a roulette wheel. The paper’s outcome is whether the cigarette was smoked.

Planning Fallacy

Underestimating how long a task will take. Study 1’s last-week share can look like that mistake. The mediator the paper prints is the fatigue-control belief, not a general time bias.

​
One-Line Takeaway

When the urge is quiet and the refusal feels easy, choose the cue’s distance first, and keep that distance on the record you later check.