The Self-Esteem Scale: What Rosenberg Measures and How It Is Scored

What the ten items actually ask

The scale is built from plain statements about overall self-regard rather than about competence in any domain. The positively worded half asks whether you feel you are a person of worth, at least equal to others; whether you have a number of good qualities; whether you can do things as well as most people; whether you take a positive attitude toward yourself; and whether you are satisfied with yourself on the whole.

The negatively worded half asks the same construct from the other direction: whether you feel you have little to be proud of, whether you are inclined to feel a failure, whether you wish you had more respect for yourself, whether you feel useless at times, and whether at times you think you are no good at all. These five are reverse-scored, so agreeing with them lowers the total.

The balanced wording is not decoration. Half the items running each direction is what stops the scale from measuring a tendency to agree with statements, which is a real response bias and would otherwise be indistinguishable from self-esteem. It is also why you cannot estimate your own score by counting how often you agreed — the direction of each item has to be accounted for first.

How it is scored, and what the number does not mean

Rosenberg's original scoring uses a four-point scale from strongly disagree to strongly agree, giving a total between 10 and 40, with higher meaning higher global self-worth. A commonly cited convention treats scores below 15 as suggesting low self-esteem, but this is a rule of thumb rather than a validated clinical cutoff, and it varies between published sources. The scale was not built as a diagnostic instrument and has no established threshold that separates a disorder from its absence.

It is also not a percentile. Turning a raw total into a rank against other people requires a reference sample from a defined population, and the norms that do exist differ substantially by country, age and how the sample was recruited. Any test that reports your self-esteem percentile without naming its comparison group is presenting a precision it does not have. Reading the total as a direction — low, middling, high — is the interpretation the instrument actually supports.

Different versions score differently, which is the most common source of confusion when two tests disagree. Some administrations use a five-point scale instead of four, some report a mean per item rather than a sum, and some extend the item count. A total from any of those is not comparable to a published Rosenberg total, even when the underlying construct is the same.

The one thing a global measure cannot separate

Global self-worth and self-criticism are not simply opposite ends of one dimension, and this is where a ten-item total loses information that matters. It is entirely possible to hold a reasonably stable sense of being worthwhile while running a harsh internal monitor — the two coexist often enough that treating a moderate total as the whole story hides the actionable part.

Self-compassion is the second thing worth separating, and the research treats it as a partly independent construct rather than as a component of self-esteem. The distinction is about timing: self-esteem is an evaluation, and evaluations fall exactly when you fail, which is when you need something to hold. Self-compassion is a stance rather than a verdict, so it remains available at that moment.

This has a practical consequence for reading a result. Self-esteem sustained by comparing favorably to others is structurally fragile, because it depends on the comparison continuing to go your way. A profile of moderate self-worth with high self-compassion is more robust than high self-worth with low self-compassion, even though the second produces the higher number on a global scale.

The assessment on this site is eighteen items built on the Rosenberg construct and scored as three separate dimensions — sense of worth, self-criticism, and self-compassion — rather than as one total, for exactly this reason. Because items were added and rescored, its output is not comparable to a published Rosenberg total, and it is not presented as one.

What the research says the scale measures, and what it predicts

The scale's psychometric performance has held up well across decades, languages and populations. An item response theory analysis of the ten items found they reliably discriminate across the full range of self-esteem and cover the construct evenly, with no redundant items (Gray-Little et al., 1997). A coordinated administration across 53 nations involving over 16,000 participants confirmed the single-factor structure and found the internal consistency high in every sampled country, though mean levels differed substantially by culture (Schmitt & Allik, 2005).

On what a self-esteem score predicts, the evidence sits between modest and meaningful depending on the outcome. Longitudinal research finds low self-esteem predicts later depression more strongly than depression predicts later self-esteem, which supports the vulnerability model — that self-worth matters for risk, not just as a consequence of already being depressed (Sowislo & Orth, 2013). Life-span development work tracking people from adolescence into old age shows self-esteem rising through young adulthood and middle age before declining in late life, and links it prospectively to better relationship quality, job satisfaction and physical health (Orth et al., 2012).

What the evidence does not support is the stronger version often repeated in self-help contexts — that high self-esteem itself causes better performance, interpersonal success or happiness. The most comprehensive review concluded the correlations are there but the causal direction is mostly the reverse: doing well raises self-esteem more reliably than having high self-esteem raises performance, and when self-esteem is experimentally manipulated, effects on behaviour are small and inconsistent (Baumeister et al., 2003). The exception is that very low self-esteem does predict worse outcomes. The pattern reads more as a floor effect than as a dose-response curve: being far below the mean matters, and most of the variation above that threshold is non-predictive noise.

What moves a self-esteem score, and what it takes to change it

Self-esteem is more stable than mood and less stable than a personality trait, which puts a single measurement in a specific interpretive position. Scores drift over the life course, with a broad pattern of rising through adulthood and dipping in late life, and they move with circumstance — a period of unemployment, a relationship ending, or a sustained run of things going well all register.

Low mood is the largest short-term confound. Depressed mood lowers self-report self-esteem substantially, and the two constructs overlap enough in questionnaire form that a low total during a difficult period is partly measuring the period. This is not a flaw to correct for so much as a fact to read with: taking the same scale again after the circumstances change is more informative than treating one administration as a fixed reading.

On what changes it, the evidence is less flattering to the obvious approach than most self-help material suggests. Self-esteem responds to accumulated evidence of competence and to being genuinely valued in real relationships, so the changes that hold tend to come from doing things and from connection rather than from repeating affirmations — which can backfire in people with low self-esteem by making the gap between statement and belief salient. Where low self-esteem is severe, persistent and accompanied by low mood, a screening questionnaire and a conversation with a professional is a more useful step than a self-help approach, because the pattern that looks like low self-esteem is sometimes depression, and the two call for different responses.