The Affirmation Mistake That Backfires (and the Simple Fix)

Positive affirmations make some people feel worse because a statement too far from what you already believe about yourself triggers an argument, and you lose that argument. The finding comes from a 2009 experiment in which participants repeated "I am a lovable person." People with low self-esteem who repeated it ended up in a worse mood and with lower state self-esteem than people who did not repeat it at all. It is the most important result in the affirmation literature and the one least likely to appear in an article about affirmations.

The study

Joanne Wood, W. Q. Elaine Perunovic and John Lee, working at the University of Waterloo, published "Positive Self-Statements: Power for Some, Peril for Others" in Psychological Science in 2009. They started from an unglamorous observation: positive self-statements are recommended everywhere, and almost nobody had tested whether they work.

Participants were sorted by trait self-esteem, then asked to repeat "I'm a lovable person" at intervals, prompted by a bell, while a control group did not. Afterward they reported their mood and their current self-esteem.

The results split cleanly by starting point. Participants with high self-esteem who repeated the statement felt slightly better than controls. Participants with low self-esteem who repeated it felt worse than the low self-esteem participants who had not repeated anything. The intervention did the most damage to the group it is most often prescribed for.

A later condition in the same paper is the part that gets cited least and matters most. Low self-esteem participants who were asked to consider ways in which the statement was both true and not true ended up feeling better than those instructed to focus only on how it was true. Permission to disagree with the affirmation improved the outcome. Insisting on it made things worse.

The mechanism: self-verification

The best-supported explanation is self-verification theory, developed by William Swann at the University of Texas beginning in the early 1980s. The core claim is counterintuitive: people are motivated to confirm their existing self-views, including unflattering ones. A coherent self-concept, even a negative one, makes the world predictable. Feedback that contradicts it is destabilizing, so the mind works to neutralize it.

That neutralizing work is the problem. When you assert something about yourself that your self-concept rejects, you do not simply fail to believe it. You actively retrieve counterevidence. "I am a lovable person" prompts a search, and for someone with low self-esteem the search returns the friendship that ended, the message that went unanswered, the room they felt invisible in. The affirmation functions as a query against a database of your worst memories, and it runs successfully.

A second mechanism from persuasion research reinforces this. Social judgment theory, developed by Sherif and Hovland, describes a latitude of acceptance around any position a person currently holds. Messages inside that latitude get assimilated and move you slightly. Messages far outside it get contrasted: they are rejected, and the rejection often pushes you further from the message than you were before. This is well established for persuasion by other people. Wood and colleagues argued it applies when the persuader is you.

The 5 conditions that make an affirmation likely to backfire

1. A large gap between the statement and your current self-view. This is the primary risk factor. The further the claim sits from what you presently believe, the more counterevidence gets retrieved. A person who feels unlovable saying "I am deeply loved" is asking for a rebuttal.
2. Trait claims rather than value or process claims. "I am confident" asserts a fixed attribute that can be checked against memory and found wanting. "I care about doing this well" asserts a value, which cannot be falsified by yesterday's failure.
3. An instruction to suppress doubt. The Wood study is explicit here. Being told to focus only on why the statement is true produced worse outcomes for low self-esteem participants than being allowed to hold both sides.
4. Repetition as the mechanism. Repeating a statement you do not believe means running the counterargument repeatedly. Frequency is not neutral. It compounds whatever process the statement started, which is why the way repetition shapes belief cuts in both directions.
5. Low baseline mood at the moment of practice. Negative mood biases memory retrieval toward mood-congruent material, which loads the counterargument search before it starts. The morning practice most people are told to adopt happens at a time of day when this varies a great deal between individuals.

What this finding does not prove

It does not prove affirmations are harmful in general. This is a single paper from a single lab, using undergraduate participants and measuring mood immediately after a short laboratory task. It has been influential and widely cited, but it has not been subjected to the kind of large multi-site replication that would settle it. Treating one 2009 study as the final verdict is the same error as treating a poster in a gym as the final verdict, just pointed the other direction.

It also does not indict self-affirmation, and this distinction is almost universally lost. Wood and colleagues tested repeating a flattering statement about the self. That is a different intervention from Steele-style self-affirmation, which asks people to reflect briefly on a value they hold and has a substantially better evidence base, including a 2025 meta-analysis in American Psychologist that pooled 129 independent tests from 67 articles. Headlines that read "science says affirmations don't work" are collapsing two literatures that used different methods and got different answers. The difference between the two, and what separates an affirmation that holds up from one that does nothing, is laid out in the breakdown of which affirmations survive testing.

Finally, the effects measured were immediate and modest. Nobody demonstrated lasting damage. What was demonstrated is that the intervention has a direction of effect that depends on who is doing it, which is enough to make blanket advice irresponsible.

What to do instead

The practical implication is not to abandon self-talk. It is to shrink the gap between what you say and what you can currently accept.

• Move from trait to value. Replace "I am strong" with "I care about handling this well." Values are chosen, so they cannot be contradicted by evidence.
• Move from assertion to question. Asking "how do I want to handle today?" invites the mind to generate an answer rather than to attack a claim. Interrogative self-talk has some experimental support and does not trigger the same refutation process.
• Allow the counterargument. The Wood study found that people who could hold both sides of the statement did better than people who could not. A statement you are permitted to doubt is more durable than one you are required to believe.
• Change the grammatical person. Referring to yourself by name creates distance from the emotional content, and there is EEG and fMRI evidence that this reduces reactivity without extra effort. It sidesteps the self-verification conflict rather than fighting it.
• Choose the statement yourself. A sentence that arrived from someone else's list has no particular relationship to your self-concept. One you selected sits inside your latitude of acceptance by construction, which is most of the battle.

The uncomfortable summary is that affirmation advice has been dispensed for decades with almost no attention to who it works for. The one study that looked found the effect reverses in the people most likely to be reaching for it. That is not a reason to stop talking to yourself. It is a reason to be more careful about what you say.