Digital video footage was analysed using SIMI Motion 7.2 software (SIMI Reality Motion Systems GmbH, Unterschleißheim, Germany). Three consecutive, non-breathing stroke cycles, for each participant, were selected for the analysis. A stroke cycle was defined as the period from water entry of the hand of the unaffected arm, to the next entry of that hand. The estimated locations of the glenohumeral joint centre and elbow joint centre of both the affected and unaffected arms were manually digitised at 50 Hz to obtain the angular position of the upper arms, as a function of time. Upper arm angle was defined as the angle between the upper arm and a horizontal reference, established using the water surface. To obtain the vertical displacement of the feet during the kick, the estimated locations of the ankle joint centre and the big toe were manually digitised at 50 Hz. The midpoint of the line intersecting these two landmarks was used to determine the position of each foot, as a function of time. Before testing, the skin overlaying the joint centres was marked with black pen to help estimate their location.
The intra-tester and inter-tester reliability of the two-dimensional digitising protocol were established by performing repeat digitisations of a randomly selected trial. The coefficient of repeatability (Bland & Altman, 1986) was obtained for arm angle on the affected side. This involved calculating the 95% limits of agreement (± 1.96 standard deviations of the differences) between each data set.
Low intra-tester and inter-tester repeatability coefficients demonstrated that the digitised data were both reliable and objective. The intra-tester and inter-tester repeatability coefficients for arm angle were 1.4° and 1.3°, respectively.