Self-Affirmation Theory
What writing about your core values does under threat, and what it does not say about 'I am' statements.
Do affirmations actually do anything?
The best-known research is not about repeating 'I am' statements. Cohen and Sherman's 2014 review covers values-affirmation: short writing exercises about a personal value. In studies, these reduced defensive responses to threatening information and, in some school settings, improved performance under stereotype threat.
Source: Cohen & Sherman, 2014
Affirmations get a bad rap in serious circles - and rightly so when they're generic, outcome-focused, or implausible. But there is a legitimate science behind them when they're done correctly.
Geoffrey Cohen and David Sherman's 2014 review ("The Psychology of Change: Self-Affirmation and Social Psychological Intervention") covers values-affirmation: short writing exercises in which people write about a value that matters to them. In the studies it reviews, these exercises:
- Reduced defensive responses to threatening information about the self
- Improved performance under stereotype threat in some school settings
- Varied by context - some studies found lasting effects, others small effects or none
What it does not test
This research did not test repeated "I am" statements. It tested writing about a value you already hold.
If you use spoken affirmations, the closest lesson is to keep them about who you are and what you value, and to keep them believable.
This is also why our free Mantra Generator targets who you are rather than an outcome.
Frequently asked
Does this research support daily 'I am' statements?
Not directly. The studies tested brief written exercises about personal values, usually done once or a few times. Repeating 'I am' statements is a different practice, and it is not what this research measured.
How large are the effects?
They vary a lot by setting. Some school studies found lasting effects from a few brief exercises; other studies found smaller effects or none. Treat it as a promising, context-dependent result, not a guarantee.