Mental training in archers: Kim 2021 meta-analysis

A meta-analysis of seventeen Korean studies on psychological skills training for archers, published between 1990 and 2017, with 273 participants in all.

The overall effect, in the random-effects model, is a Hedges' g of 0.469, with a 95% confidence interval from 0.347 to 0.591; in the fixed-effects model it is 0.427, from 0.364 to 0.491. The scale the authors themselves adopt calls anything between 0.2 and 0.5 small, so the overall result sits at the top of the low band: on their percentile reading, an archer in the middle of the trained group would stand at the 68th percentile of the untrained one. Heterogeneity between the studies was substantial, with a Q of 47.678 on 16 degrees of freedom, p below 0.001, and an I squared of 66.4%.

In the subgroups, the effect was larger in middle and high school athletes (0.666, from 0.508 to 0.823) than in adults (0.338, from 0.230 to 0.446), with Q equal to 11.326 and p equal to 0.001; larger in programmes over twelve weeks (0.581) than in shorter ones (0.393), with p equal to 0.025; and larger in dissertations (0.746) than in journal articles (0.384), with p below 0.001.

Nothing was separated by frequency (one or two sessions 0.428, three or more 0.637, p equal to 0.161) or by the length of a single session (under sixty minutes 0.639, over 0.496, p equal to 0.282). The check on publication bias by the trim and fill method added three studies and brought the overall effect from 0.469 down to 0.398.

The search covered four Korean databases, the Academic Research Information Service, Nuri Media, Korean Academic Information and the Korea Research Foundation, for work published up to 2019, with keywords for archery, psychological skills training, goal setting, psychology training, relaxation and routine. From the 3,467 initial results, reading titles and abstracts and then full texts led to seventeen papers: fourteen journal articles and three dissertations, all from Korean national universities. Where a thesis and an article covered the same data, the article was kept.

Eligible studies were those on archers, with an experimental test of the effect of a psychological intervention, with anxiety, psychological skills or performance as the dependent variable, and with statistics sufficient to compute an effect size. Two researchers selected independently and resolved disagreements by discussion.

Coding covered fifteen items, from the type of training to the hours per session; the first author did it and then redid it, comparing against the first pass. Effect sizes were computed as Cohen's d and converted into Hedges' g to correct the overestimation that arises with small samples, using the Comprehensive Meta-Analysis 2.0 programme. Given the heterogeneity, the overall estimate was made with the random-effects model, using the study as the unit of analysis so that several outcomes from the same paper would not be treated as independent. Publication bias was examined with a funnel plot and with trim and fill.

The included studies are small: their samples run from 8 to 40 participants, with a median of 15, and eight of the seventeen have twelve people or fewer.

The abstract does not match the results. It says adult players had bigger effect sizes, while the table and the discussion give 0.338 for adults against 0.666 for middle and high school students, which is the opposite. This record follows the table.

The subgroup counts match neither the seventeen studies nor each other. The table says it is counting studies, and then reports 67 and 36 for the split by level, 99 and 26 for duration, 26 and 34 for frequency, 28 and 45 for session length and 100 and 25 for publication type: five different totals. They are evidently counts of effect sizes rather than of studies, and that undoes the very independence the overall estimate had taken care to preserve.

What is pooled is also mixed: anxiety, questionnaires on mental skills and shooting performance all go into a single number, so the 0.469 does not say what psychological training does to a score. There is no assessment of risk of bias in the included studies, no information on blinding and no distinction between controlled and uncontrolled designs.

There is nothing more recent than 2017, and the search queried Korean databases only: the authors list this as a limitation, along with the impossibility of separating the types of training, since almost all the studies applied two or more at once. Two of the seventeen rows in the table carry the same author and year, Han 1998, with different sample sizes.

Finally, the reading of bias: the authors take the fact that dissertations give a larger effect than journal articles as a sign that there is no excessive distortion. That is one possible reading, but the trim and fill in the same paper lowers the overall estimate by about 15%.

For an archer, the honest summary is that in Korea, across thirty years, mental preparation of various kinds has been followed by small improvements in a mixture of anxiety scores, questionnaires on mental skills and shooting results, and that the improvement was clearest in young athletes and in programmes running beyond three months.

It does not say which kind of mental work matters, because the studies almost always applied several at once and the review could not separate them: imagery, relaxation, goal setting, routines and biofeedback all sit inside the same number. Nor does it say how much of the 0.469 is score and how much is questionnaire, which is the question an archer would ask first.

The figure for adults deserves a cautious reading. The effect size in adults is half that in school students, and the authors explain it by saying mature athletes have already acquired these skills. That is an interpretation of a difference between two groups of studies rather than a result about adults, and it comes with the authors' own warning about the variation in the number of cases across subgroups.

Duration, by contrast, is the firmest indication to come out of here, and the reason it circulates: beyond twelve weeks the effect was larger than below, and the frequency of sessions changed nothing. If these seventeen studies together suggest one thing, it is that time matters more than tempo. It remains a suggestion from a comparison between subgroups, not from a study built to answer that question.