CONCEPT
What self-regard actually is and where it comes from.
Self-esteem is a readout, not a resource. It is the running estimate you hold of your own worth, and the useful way to think about it is as an instrument reporting on something else rather than a tank of fuel that gets spent and refilled. This distinction decides everything downstream. If it is a gauge, the way to change the number is to change what the gauge measures. If you believe it is fuel, you will spend years trying to top it up directly and wonder why the effect never holds.
Mark Leary's sociometer theory gives the cleanest account: self-esteem tracks perceived relational value, meaning how includable you are to people whose inclusion matters to you. On that account, the sting of rejection is not a malfunction. It is the instrument working. Social exclusion was expensive enough for long enough that a monitor for it is exactly what you would expect to find.
William James proposed a second reading, and the two are compatible. In *The Principles of Psychology* he framed self-esteem as a ratio of successes to pretensions — what you achieved over what you staked your self-regard on. That formulation has an underused consequence. There are two levers, not one. You can raise the numerator, or you can stop staking your self-regard on domains you have no intention of competing in.
Self-esteem is downstream of standing and evidence, and it is nearly impossible to move by talking to yourself about it. Affirmation attacks the display, not the input.
The self-esteem movement of the 1980s and 90s treated the gauge as a cause: raise how children feel about themselves and performance, behavior, and outcomes will follow. Roy Baumeister and colleagues reviewed the literature in 2003 and reached a much narrower conclusion — high self-esteem correlates with self-reported outcomes far more reliably than with objectively measured ones, and where it accompanies achievement it looks more like a consequence than a driver. Their review also noted that threatened high self-esteem, not low self-esteem, is the profile associated with aggression.
So the intervention was aimed at the dial. Predictably, the dial moved and little else did.
If you want a self-belief that actually predicts behavior, the one to use is self-efficacy, in Albert Bandura's sense: a domain-specific judgment about whether you can execute a particular thing. "I am a worthwhile person" forecasts almost nothing. "I can hold a hard conversation without going silent" forecasts a lot, and it is built the way any belief with evidence behind it is built — by accumulating instances, starting small enough that the instances are real.
The practical order is therefore inverted from the popular one. Do not repair the estimate and expect behavior to follow. Produce behavior worth estimating, in a domain you have actually chosen, in front of people whose judgment you accept. The number follows, and when it arrives that way it is not fragile, because it is not a claim. It is a measurement.