Fig. 3: Performance oscillates within the stride cycle.

a An example trial from one participant displaying the position-tracked vertical head height over time whose peaks and troughs are used to define the steps of a stride cycle. Target onsets are displayed with black arrows, and red markers indicate the behavioural outcome of the task (Hit, Miss). b To allow pooling of data over participants, each stride (two steps) was resampled to a normalised range of 1–100%. The approximate swing and stance phases of the stride cycle are displayed above the x-axis. c Target presentation density was approximately uniform over the stride cycle, validating the random-probing and stride epoching procedure. d–f Group level data (N = 36) show clear oscillations over the stride cycle for detection accuracy (hit rate), reaction time, and the likelihood of manual responses. The best-fitting first-order Fourier model is shown in each figure and approximated 2 cycles per stride (i.e. the step rate). Individual data points (grey) are shown for 10% increments over the stride cycle. g Permutation testing of the best-fitting Fourier models. The observed data in c–f were fitted with a single-component Fourier model at all frequencies between 0.2 and 10 cycles per stride (in steps of 0.2 cps), and goodness-of-fit (R²) was calculated. The prominent peaks at approximately 2 cps show that the best fits, whether for accuracy (blue), reaction times (red), or responses (magenta), occur at approximately 2 cycles per stride and far exceed the 95% permutation confidence interval calculated using the maximum test statistic across all frequencies per permutation (dotted lines and grey shading). These peaks did not co-occur with the above-chance fluctuations in the likelihood of target presentation, which is shown by the grey line.