Choking in archery: Diotaiuti 2021 study

A hundred and fifteen competitive archers licensed in Lazio, seventy-two men and forty-three women, aged ten to seventy-four with a mean of thirty-nine, filled in a questionnaire in June and July 2019. No arrows, no scores: standard instruments measure self-consciousness, sport anxiety in its three components, coping style and decentering, the ability to step back from one's own thoughts, and the archers are asked how often they choked in the past six months.

The model that comes out has a clear shape. Somatic anxiety, social anxiety and trouble concentrating together feed rumination, which the article calls pre-occupation, and they account for it well, 63.8 per cent. Coping style mediates that step: those who meet stress by returning to the task ruminate far less, those using emotional control strategies also do, to a smaller degree. Rumination then predicts how often choking happens, while decentering and the perceived support of the coach run the other way.

On the final result, though, the numbers are modest and should be read for what they are. The model predicting choking frequency from rumination, coach support and decentering accounts for 19.3 per cent of the variance, 17.2 adjusted: four fifths of that frequency stay outside. Decentering, which the abstract says can strongly limit choking, is worth 4.7 per cent of the variance.

There is then something to know about the main variable. The declared frequency of choking runs from a minimum of one to a maximum of seven, with a mean of 3.08: there is no archer with zero. So the model does not account for who chokes and who does not, it accounts for how often a group in which everybody declares choking does so.

The article, in its title, calls itself an exploratory pilot study, and that is worth keeping in mind while reading the abstract, which instead speaks of the factor directly influencing the frequency of choking.

A cross-sectional study, one administration. The sample is declared to have been drawn at random among the Italian federation's licensees in Lazio, 2,325 athletes in all, and a few lines later one reads that participants were recruited directly at official competitions: those are two different descriptions of the same sampling, and the response rate is absent. An anonymous questionnaire online on a platform, with informed consent.

The general section collects personal data, class and category, bow type, a preference between indoor and outdoor, training alone or in a group, weekly sessions and their duration, perceived coach support in training and in competition, the highest level of competition in the past year and the frequency of choking in the past six months. The sample shoots 3.22 sessions a week for 2.32 hours, with eleven years of practice since finishing the basic course. Olympic bow 59 per cent, compound 23, barebow 16, longbow 3.

The psychometric section uses four validated instruments, with Cronbach's alpha reported for every scale: the self-consciousness questionnaire in its three subscales of private and public self-awareness and social anxiety, from 0.70 to 0.79; the Sport Anxiety Scale with pre-occupation at 0.90, somatic anxiety at 0.92 and concentration disturbance at 0.73; the coping style inventory with approach at 0.68 and avoidance at 0.70; and the decentering scale for sport at 0.78.

The analysis is elaborate: Pearson and Spearman correlations, t-tests, analysis of variance, three hierarchical regressions, mediation analysis with the PROCESS macro and ten thousand bootstrap samples, and a structural equation model. The assumption checks are declared and reported, which is not common: Mardia's multivariate kurtosis below the critical value, variance inflation factors under two, tolerances above 0.60, residuals with a mean of zero and a Durbin-Watson statistic of 1.58.

The final structural model has good fit indices, with a chi-square of 21.974 on seventeen degrees of freedom and p equal to 0.186, a comparative index of 0.979 and a root-mean-square error of approximation of 0.051. The ninety per cent interval for the last of those, which the authors report, runs from zero to 0.105, that is it also takes in poor fit, and the two parsimony indices stand at 0.557 and 0.594.

The first thing to say concerns what was measured, because it settles everything else. The dependent variable is the frequency of choking, recalled over the past six months. It is not a performance: the definition of choking the article itself quotes from the literature is a drop in performance under pressure, and a drop in performance is nowhere observed here, because there are no scores, no competition results and, as the authors write among their limits, individual results in competition were not verified. What is accounted for is how often a person applies a label to themselves, not how often that person shoots worse.

And that variable has no zero: the declared minimum is one. So in the sample there is nobody who did not choke in six months, or the question did not allow saying so. The descriptive statistics on that variable, moreover, do not hold: one reads that the mean duration of choking, expressed in training sessions, is 6.02 with a standard deviation of 9.75, and that the maximum duration recorded is 15.57 with a standard deviation of 33.73, that is a maximum duration which has a mean and a standard deviation, and which is expressed in episodes where the preceding sentence spoke of sessions.

Three regressions out of three have the denominator degrees of freedom printed wrongly. With a hundred and fifteen cases and five predictors the residual degrees of freedom are 109, not 114; with four predictors 110; with three, 111. Redoing the calculations from the R squared values with the correct degrees of freedom gives F equal to 38.42, 13.18 and 8.85, that is the three published values to within six hundredths, while with the printed 114 one would get 40.18, 13.66 and 9.09. So the F values are computed correctly and the number printed beside them is an error repeated three times: it changes no conclusion, but it is the kind of thing a reader uses to judge how closely the manuscript was read.

In the correlation table two cells do not add up. Coach support in competition correlates with choking at minus 0.151 and is marked as significant at five per cent, while with a hundred and fifteen cases that coefficient gives a probability of about 0.11. The table's note says Spearman's coefficient was used for choking, which moves the figure a little, not enough. This matters because that is the raw correlation between coach support and choking, the bare datum behind the article's most quoted conclusion: in the multivariate models the effect comes out significant, with a weight of minus 0.248, but on its own the link is worth minus 0.15 and does not reach the threshold. At the opposite end, the correlation between rumination and decentering, minus 0.497, is the only one of that size without asterisks, while with this sample it sits below a thousandth.

In the limitations section one reads that most participants compete in the first category, the most competitive, with elite athletes of the federation. The table says 33.9 per cent in the first category: not most, a third. And looking at the level of competition, 59 per cent are at regional or interregional level and 11 per cent at international. It is a sample of competitive archers at every level, not an elite sample.

On causal language: the design is cross-sectional and the authors acknowledge it, asking for caution about conclusions on causality. The abstract, though, writes that rumination is the factor directly influencing the frequency of choking and that decentering has been shown able strongly to limit that frequency. The arrows of a structural model are hypotheses put in by the researcher, not directions found in the data, and with a questionnaire taken once the direction can be reversed: someone who chokes more often has more reason to ruminate. As for strongly, decentering adds 4.7 per cent of accounted-for variance to a choking frequency of which 19.3 per cent is accounted for in all.

Two smaller things. The statistical section announces eight indices for evaluating the model and then lists twelve. And thirty-four participants out of a hundred and fifteen, 29.6 per cent, declared that the questionnaire period fell inside a choking phase of their own: nearly a third of the answers were given during the very episode they were asked to recall, and this is used neither as a variable nor discussed.

Finally, for the register, a point that needs care. The introduction describes target panic and its signs, the body stiffening, the bow arm locking, trouble anchoring and getting through the clicker, the release anticipated or delayed, and links it to choking under pressure citing other authors. None of this is measured in this study: the questionnaire asks about choking, not about target panic, and there is no measure of automaticity, of movement control or of performance. Anyone looking here for support for a mechanism of target panic will not find it, for or against.

What this work offers, to those who shoot and those who coach, is a map of how the mental components hold together, not a measure of how much they weigh.

The map is this. The anxiety that counts is not one thing: somatic anxiety, the heartbeat and the muscular tension, social anxiety, the fear of judgement, and trouble concentrating add up and feed rumination, and together they account for almost two thirds of it. Rumination is the intermediate step: it is there, and not directly in anxiety, that the model sees the link with how often choking happens.

In between sits the style with which stress is met, and this is the most useful point in practice. Those who respond to stress by returning to the task report far less rumination, with a correlation of minus 0.58, and less choking, minus 0.33. Those using emotional control and rationalising strategies get the same effect to a smaller degree. These are correlations, not proof of effectiveness, but they say where to look: the target of an intervention is not anxiety itself, it is what is done with it.

Decentering, managing to look at one's own thoughts without being carried off by them, runs the right way: the more of it, the less choking is reported. It is a capacity that can be trained, and this work gives one more reason to try. But the weight, here, is small: 4.7 per cent of the variance of a variable accounted for to 19.3 per cent. Anyone presenting it as the solution goes beyond the data.

Coach support is the article's new element, and the most fragile. In the models it comes out with a weight of the same order as the others, but the raw correlation with choking frequency is minus 0.15 and does not reach the threshold. To be read as a pointer: an archer going through a choking phase should not go through it alone. Not to be read as a quantity.

Two things this study does not say, and that it is important not to take from here. The first: it says nothing about scores. Nobody shot, no competition result was verified, and choking is a self-assessment. The second: it says nothing about target panic, although the introduction describes it. The release is not measured, automaticity is not measured, the block is not measured.

One figure remains that counts for a coach: out of a hundred and fifteen competitive archers, every one declares at least one choking episode in six months, and nearly a third answer the questionnaire while going through one. Whatever the models say, the phenomenon in that group is widespread rather than exceptional.