Despite the large volume of research on the biomechanics of running, the upper extremities have been virtually ignored by researchers. Although some speculation has been made concerning the various purposes of swinging the arms (e.g., they help in maintaining one's balance;​ they set the tempo for the legs;​ they help keep the trunk square;​ etc.), the function of the arm swing has never been investigated thoroughly. The purpose of this dissertation was to investigate the nature of the arm action in running.

Ten male recreational runners ranging in age from 20-32 years were filmed while running on a treadmill at three different speeds:​ (1) "slow" — 3.8 m/s, (2) "medium" — A.5 m/s, and (3) "fast" — 5.4 m/s. These represented running paces of 7, 6, and 5 minutes per mile, respectively. A four-camera three-dimensional (3D) cinematographic technique was used (Williams, 1980) with each camera operating at a nominal rate of 100 frames per second.

Simultaneous electromyograms were obtained from eight upper extremity and trunk muscles believed to be involved in the arm swing. These muscles were

1. Pectoralis major — clavicular portion (PEC),
2. Anterior deltoid (AD),
3. Middle deltoid (MD),
4. Posterior deltoid (PD),
5. Latissimus dorsi — upper part (LAT),
6. Biceps brachii — long head (BIC),
7. Brachioradialis (BR), and
8. Triceps brachii — long head (TRI).

The EMG and film data were synchronized by means of a digital pulse elicited at each left foot-strike (LFS) from the output of an insole footswitch device placed inside each subject's shoe.

A 14-segment mathematical model of the body was used and consisted of the head, trunk, two upper arms, two forearms, two hands, two thighs, two calves, and two feet. The 3D estimates of joint centers digitized from the films represented the endpoints of each of the segments. Body segment parameters (BSPs) were obtained from various sources in the literature. Several different combinations of BSPs were used, and a certain set was deemed most appropriate for a given subject if it minimized the fluctuations in the computed linear and angular momentum values during the airborne phases of the running cycle.

The raw 3D segment endpoint data were smoothed using a digital filter with cutoff frequencies ranging from 2-8 H_z. Different cutoff frequencies were chosen objectively for each coordinate of each end-point using an algorithm similar to the one described by Jackson (1979) adapted for use with a digital filter.

The segment endpoint data served as input to a computer program which calculated the following quantities:​

1. the paths of the centers of mass (CMs) of the arms, the body-minus-arms, arn the whole body over the running cycle,
2. the contributions of the arms, head+trunk, and legs to the changes in total-body momentum over each contact phase ("lift" and "drive"),
3. the angular momentum of each body segment and various systems of segments about the body CM, and
4. the net moments at the shoulder and elbow joints.

The results of the EMG and net joint moment analyses showed each of the eight muscles to have bursts of phasic activity at different times throughout the running cycle. The PEC and BIC showed the least activity with mean peak integrated values ranging from 10-24 %MAX. The PD and LAT showed the most activity with mean peak integrated values ranging from 30-60 %MAX. In each muscle the EMG activity increased with running speed.

EMG activity in the PEC and AD correlated well with the occurrence of a net flexor moment at the shoulder which halted the backward swing and initiated the forward swing of the arm. Activity in the MD, PD, and LAT produced a net extensor moment at the shoulder throughout most of the forward swing and into the first half of the backward swing of the arm. The eccentric activity during the early part of the forward swing appears to have been used to control the forward swing of the arm, which was aided by gravity.

Instead of a single phase each of flexion and extension per cycle as for the shoulder joint, the elbow showed two phases of flexion and two of extension per cycle. The primary extension phase (PEP) occurred at contralateral foot-strike (CFS) followed by a much smaller secondary extension phase (SEP) at ipsilateral foot-strike (IFS). The EMG activity in the BIC and BR was closely associated with flexor moments at the elbow which arrested each extension phase and initiated elbow flexion. The PEP was initiated by a peak extensor moment at the elbow occurring at the time of peak activity in the TRI. The SEP (absent in some subjects) occurred without either an extensor moment or TRI activity and was probably due to gravity.

The results of the center of mass analysis revealed that the arms did not reduce the magnitude of the vertical oscillations of the body CM in running. Just the opposite was found - the arms increased the vertical range of motion of the body CM. This effect increased as speed of running increased.

In the horizontal directions, however, the arm swing was generally found to reduce the excursions of the body CM from side to side and front to back. The addition of the arm swing would therefore tend to keep a runner’s horizontal velocity more constant.

The results of the lift and drive analyses agreed closely with the center of mass results. In the vertical direction, the arms were found to make a positive contribution to lift, roughly 5-10% of the total. This contribution increased with running speed.

In the AP direction, on the other hand, the arms were generally found to make a small negative contribution to drive, although considerable variability existed between subjects. A negative drive value would be consistent with the idea that the arms tend to reduce the changes in the forward velocity of the runner, rather than increase them.

The results of the angular momentum analysis showed that while the body possessed varying angular momentum about all three coordinate axes, the arms made a meaningful contribution to only the vertical component (H_z). It was in the arms' contribution to the total-body H_z that the main function of the arm swing became apparent.

The arms were found to generate an alternating positive and then negative H_z pattern during the running cycle. This tended to cancel out an opposite pattern H_z of the legs. The trunk was found to be an active participant in this balance of angular momentum with the upper trunk rotating back and forth with the arms and the lower trunk with the legs. The result was a relatively small total-body H_z throughout the running cycle.

The balance was not complete, however, as the lower body (legs and lower trunk) possessed roughly 15% more H_z than the upper body (arms and upper trunk). This difference was reduced to only 7% at the fast running speed, however, suggesting that the arms may play an even larger role at faster running speeds.

The nature of this balancing role that the arms and upper trunk play appears to be in the generation of internal torques between upper and lower body about the z axis. The resulting "action-reaction twist" about the trunk long axis gives the legs the needed H_z to move through alternating strides leaving only a minimal amount of H_z to be received from the ground.

The results of case studies of two individual subjects demonstrating asymmetrical arm actions at each speed showed the potential of the arms to compensate for each other and for asymmetries elsewhere in the body.

In one xubject whose right arm motion apparently was restricted from wearing the EMG electrodes, the left arm was swung more vigorously to make up for the loss in the H_z generated in the right arm. This asymmetrical arm action, while creating a symmetrical H_z pattern, manifested itself in asymmetrical contributions of the arms to lift and drive over the left and right foot contact phases. Maintaining symmetry in lift and drive patterns was, perhaps, of secondary importance to maintaining the H_z balance.

The second subject showed an arm asymmetry which could be linked back to an asymmetrical pattern of leg motion. At each running speed the subject demonstrated a net decrease in his forward velocity over the left foot contact phase and then a net increase over the right foot contact phase. Accompanying this asymmetrical drive pattern were increases in both the legs' and arms' H_z levels during the second half of the cycle compared to the first half. The arms generated their increase through a more forceful backswing of the left arm which, in addition to partially balancing the increased H_z of the legs, helped to provide a portion of the increased drive the body experienced over the second half of the cycle.

Based on the findings of this study the following conclusions are drawn:​

1. The main functions of the upper extremities in running involve translation along and rotation about the vertical (z) axis. The former involves the ability of the arms to provide a small but important contribution to the total propulsive lift on the body. The latter involves the ability of the arms and upper trunk to balance the H_z of the lower body by providing the majority of the angular impulse about the z axis needed to put the legs through their alternating strides in running.
2. Secondary functions of the arm swing include the reduction of the excursions of the body CM from side to side and front to back.
3. The arms generally do not make any direct contribution to the forward propulsion of the body ("drive") during normal distance running.
4. There is no apparent advantage of the "classic style" of swinging the arms directly forward and backward over the style that most distance runners adopt of letting the arms cross over slightly in front. The cross-over, in fact, contributes to the reduction of the side-to-side excursions of the body CM mentioned in (2) above.