Why mathematical self-concept matters

Two children with similar prior knowledge work on the same task, and both initially make the same mistake. The first child thinks: “I have made a mistake.” The second thinks: “I am just no good at maths.” Although their starting points are almost identical, the two will go on learning very differently. The difference lies not in mathematical knowledge but in what research calls mathematical self-concept.

What mathematical self-concept is

Mathematical self-concept is a person's assessment of their own mathematical abilities: do I feel up to difficult tasks? Do I believe I can find solutions? What matters is not whether this assessment is objectively accurate. What counts is how the child perceives their own abilities, because that perception governs behaviour: which tasks a child chooses, how long it persists when things get difficult, and whether it reads a mistake as information or as a verdict on itself.

The importance of this characteristic is well evidenced. According to Marsh's Reciprocal Effects Model, subject-specific self-concept and achievement reinforce each other over time: a higher self-concept predicts later gains in achievement, even when prior achievement is statistically controlled for. A meta-analysis of 55 longitudinal studies (Valentine, DuBois & Cooper, 2004) confirms this effect of self-related beliefs on later achievement, particularly when they are measured for a specific subject. A child who feels up to mathematics learns measurably more mathematics as a result.

It does not develop on its own, quite the opposite

One might assume that self-confidence grows by itself over time. For mathematics, longitudinal studies show the opposite: self-related competence beliefs and positive learning emotions typically decline over the course of primary school. Children increasingly compare themselves with their classmates, experience failures more consciously and begin to classify themselves as “good” or “bad” at maths; these early judgements often shape many years of subsequent learning.

For teaching and learning environments this means that anyone who wants to keep self-concept stable or strengthen it is working against a developmental trend. Even the absence of a decline would be a success.

What actually strengthens self-concept

Self-confidence comes not from praise but from the repeated experience of managing difficult tasks under one's own steam. For that experience to occur, several conditions have to fit together, and research on feedback and mistakes has described them precisely.

Tasks must sit at the individual's performance limit: hard enough that solving them is experienced as an achievement, and attainable enough that they succeed. Feedback works best when it comes immediately, refers to the solution process rather than the person, and is not phrased judgementally (Hattie & Timperley, 2007; Shute, 2008). And mistakes only become learning opportunities when the child can correct them without social judgement; the perceived climate around mistakes determines whether a failed attempt leads to deeper engagement or to avoidance.

An adaptive system can meet these conditions structurally: it keeps difficulty continuously at the performance limit, responds to every failed attempt immediately and individually, and nobody but the child sees the mistake. That is exactly what the interplay of the adaptive Knobelbox and the tutor Mazio aims at in MatheZeit, with Mazio giving graduated hints after failed attempts rather than revealing the solution.

What the study shows

In the accompanying survey for the Flensburg efficacy study, the children's mathematical self-concept was measured before and after the six-week period of use. This survey ran without a control group; the results describe changes during use and are not to be read as causal evidence. With that qualification, the findings are remarkably consistent.

Self-concept rose measurably over the period of use, a considerable finding against the background of an otherwise declining trajectory. The pattern at the level of individual statements is revealing: the clearest rise was in agreement with “I can solve difficult tasks too”, followed by “Maths comes easily to me” and “I learn quickly in maths”, that is, precisely the task-level beliefs that research identifies as the first to respond to successful learning experiences.

The children with the least confidence benefit the most

If the children are grouped by their starting level, a compensatory pattern emerges: self-concept and interest rose most clearly in the third of children who had started with the least confidence, and in classes working with MatheZeit for the first time the gain was considerably higher than average. The platform therefore reaches in particular the children who previously showed little confidence and little interest, exactly the group that should be in focus from the perspective of educational equity.

With a pattern like this, the obvious objection is regression to the mean: those who start particularly low are often higher at the second measurement for purely statistical reasons. The report examines this objection and cites objective process data that do not rest on self-reports: the rate of tasks solved at the first attempt rose systematically across sessions, and the children with the lowest starting rates caught up the most, with difficulty increasing continuously and with tasks whose numerical values are generated afresh each time. A rising rate of first-attempt success under these conditions indicates a genuine gain in competence, not a statistical artefact and not memorisation.

Not an end in itself

A strengthened self-concept is not a soft additional goal alongside the actual learning. According to longitudinal research it is an early indicator of future competence development: children who feel up to tasks choose more demanding problems, stay with a solution for longer and learn more as a result, which in turn supports their confidence. Turning this cycle in the right direction early, especially for the children where it is at risk of tipping the other way, is one of the most important tasks of mathematics teaching. A learning environment should therefore also be measured by what children think about themselves after working with it.