Forearm EMG in archers: Soylu 2006 study

The study asks whether stronger archers repeat the electrical signal of the forearm muscles more closely from one arrow to the next. Twenty-three participants were compared: seven high-level archers (mean FITA score 1,303.4 with a standard deviation of 26.2), six beginners (1,152 with a standard deviation of 9.0) and ten people who did not shoot, who were simply assigned a score of 250.

The measure of similarity is the variance ratio, a number between 0 and 1 where 0 means identical traces and 1 means no repeatability at all. The variance ratio correlates with the FITA score, with a coefficient of 0.72 for both the extensor digitorum and the flexor digitorum superficialis. The medians of the three groups differ (p below 0.002 for both muscles), and in the pairwise comparisons it is the non-archers who stand apart: the extensor median of the high-level archers is lower than that of the non-archers, and for the flexor both archer groups sit below the non-archers. Between high-level archers and beginners, by contrast, there is no statistically appreciable difference.

The authors checked for fatigue: the median frequency of the spectrum, which falls as a muscle tires, was 115.6 hertz in the first shots and 106.2 in the last, with no statistical difference. Their conclusion is that the variance ratio might serve to assess technique, progress and talent selection, and they point out themselves that a coefficient of 0.72 covers about 52% of the variation in scores.

The recordings are not new: they are the same ones used in an earlier paper by the same group, from 2005, re-analysed here with a different index of repeatability. Silver / silver chloride electrodes were placed longitudinally over the belly of the flexor digitorum superficialis and the extensor digitorum, 2 centimetres apart, with the reference electrode on the olecranon of the ulna, chosen during piloting among several sites because it does not disturb the archer. The acquisition system worked between 8 and 500 hertz, sampling at 1,000 hertz, with a maximum inter-electrode impedance of 6 kiloohms and no mains filter.

Each participant did a single session of 15 shots, the first three for warm-up. From the remaining twelve shots, two seconds of signal were taken, with the fall of the clicker set at the first second, and the trace was full-wave rectified and smoothed with a 60 millisecond moving average. Repeatability was computed over the window from 900 to 1,500 milliseconds, which covers the end of the draw, the aim and the first part of the follow-through.

The variance ratio compares, interval by interval, the spread of the twelve traces around their mean with the overall spread of the signal. The FITA score comes from earlier competitions, not from the session. The analyses use Spearman correlation, because the scores are not normally distributed, the Kruskal-Wallis test for the three groups, and Dunn's multiple comparisons as a post-test. Fatigue was estimated with a Fourier transform over four half-second windows per shot, comparing each archer's first and last shot.

The weak point is the group of non-archers. Ten people who do not shoot were assigned a FITA score of 250, the same for all of them, with a standard deviation of zero: a made-up value so that everyone could be placed on one scale. The correlation of 0.72 between score and repeatability rests largely on this separation between those who shoot and those who do not, and the pairwise comparisons bear that out, since the only difference that emerges is the one against the non-archers. Between the seven high-level archers and the six beginners, that is, between archers shooting 1,303 and archers shooting 1,152, no difference appears. The study therefore documents that shooting a bow makes the movement repeatable, not that among archers the repeatability of the signal separates those who score higher.

The sign of the correlation is not printed. Since a low variance ratio means more similar traces, a higher score has to go with a lower ratio, so the relationship is inverse: it has to be inferred from the group medians.

The authors point out two further limits themselves: the coefficient accounts for about half the variation in scores, so forearm activity is not enough to say why an arrow lands off centre; and the index holds for comparing groups, not for judging an individual, as long as there are only twelve shots per archer.

The limits of the material remain. Twenty-three people, a single session, twelve arrows each, only two muscles, and FITA scores taken from earlier competitions rather than from the session measured, with everything that differs between a competition and a laboratory. Distance and shooting conditions are not described in this paper, which refers back to the 2005 work. The authors, finally, accept for simplicity that the signal recorded over each of the two muscles comes from that muscle, while acknowledging that crosstalk between neighbouring forearm muscles exists.

What this work puts a number on is consistency: not how hard a muscle works, but how closely its trace resembles itself from one arrow to the next. People who shoot regularly have far more similar traces than people picking up a bow for the first time, and that shows clearly; between an archer shooting 1,150 and one shooting 1,300, with these data, it does not.

For a coach the interesting result is the outline. The repeatability of the forearm signal covers about half the variation in scores and says nothing about why an arrow landed off centre, so as an indicator it serves for looking at a group rather than judging a single archer on twelve arrows; the authors put that limit in writing themselves, before proposing the method for talent selection. Anyone using electromyography as a working tool will find here a simple way of quantifying the similarity between shots, with the measurement window and the filter settings already specified.