Exam-Anxiety Training Improved Self-Reports, Not Grades

The trial

One hundred and five university students in Kazakhstan, aged seventeen to twenty-eight and averaging twenty, completed a study of mindfulness training. Fifty-four attended a twelve-week course and fifty-one attended nothing at all. Their self-reported exam anxiety diverged over the term, and their grade averages did not.

The work was carried out at Zhetysu University in Taldykorgan by Assylzhan Yessimbekova and four colleagues, and published in 2025 in the Journal of Education and Health Promotion. Students were recruited through in-person announcements and social media; 483 were approached, 157 consented, and 105 remained in the final analysis. Sixty-one were women, and thirty-three of the hundred and five were postgraduates.

Assignment was not random. After a baseline round of testing, the researchers deliberately distributed students with similar anxiety, adaptability and grade scores across the two groups, so that the groups would resemble each other at the start. The comparison group received no programme, and the paper describes no alternative activity offered to them in its place.

The course consisted of twelve weekly sessions of an hour, delivered in a large lecture theatre to groups of twenty-three to twenty-nine by the study’s first author, a psychologist with three years of personal practice and one year of experience leading such a group. Test anxiety and adaptability were measured by questionnaires the students filled in about themselves. Grade point averages came from the university’s registration office.

Anxiety scores in the trained group fell from 3.43 to 3.20 on a five-point scale, while scores in the untrained group rose from 3.51 to 3.68. The gap between the groups at the end was statistically significant, with a partial eta squared of .202, above the .14 the authors had set as their own threshold for a large effect. Adaptability ended higher in the trained group than in the untrained one, but the interaction of time and group returned a p of .164, which leaves the two groups changing at the same rate.

Grade averages rose slightly in both groups and the analysis distinguished nothing. Group reached a p of .050 against a Bonferroni-corrected threshold of .016, time a p of .052, and the interaction of the two a p of .642.

The authors then asked whether exam anxiety predicted grades at all. Pooling both groups and both time points, the coefficient was 0.004 with a p of .937, and adaptability carried no indirect effect between the two either. Measurement stopped in late April 2024, shortly before the final exam period, and no follow-up was conducted. Among these students, exam anxiety and academic results were unrelated.

What the design will bear

Everyone in this study was at university. The youngest was seventeen, a third were postgraduates, and nothing here was measured on a school-age child. Applying a twelve-week course for a twenty-year-old sitting finals in Taldykorgan to a fourteen-year-old preparing for GCSEs is an inference, and this paper supplies no ground for it.

The two outcomes that improved were the two the students scored themselves. They knew which group they were in. They had spent twelve weeks with an instructor who had designed the study and believed in the method, and they were then handed a questionnaire asking how anxious examinations made them. The registrar’s figures, which nobody in the lecture theatre supplied, showed no separation at all.

A student who reports less dread before an exam has gained something worth having, and the study is entitled to say that its trained group reported it. What the design cannot separate is a genuine reduction in anxiety from a change in how students describe themselves after a course whose purpose had been explained to them weekly for three months.

Assignment by hand rather than by chance leaves the comparison further exposed. Matching on baseline scores equalises what was measured and says nothing about motivation or family background, which the authors name among the variables they did not control. Students who volunteered for a mindfulness course and then attended at least eleven of twelve sessions are a particular sort of student.

Where a comparison group receives nothing, the trial cannot distinguish the effect of mindfulness from the effect of twelve hours of structured attention in a room with a psychologist who is interested in the student. This is the recurring finding of the wider intervention literature, where results built on untreated or waiting-list comparison groups shrink, and sometimes vanish, once trials with active controls are examined on their own.

Schools now offer mindfulness programmes routinely, and the case made to parents is usually that calmer students learn better. This trial tested that case against registrar grades and separated nothing, in the one outcome the students did not supply themselves. The reduction in reported anxiety may well be real and may well be worth twelve hours of a term, and it is a different claim from the one the programmes tend to advertise. When a school offers you such a course on the promise of better results, the question worth asking is who supplied the numbers.