Two conditions are necessary and sufficient to demonstrate the cause of impingement syndrome for a given shoulder configuration: (a) the greater tuberosity and the acromion and/or coraco-acromial ligament must impinge the subacromial structures; and (b) the compressive force generated by such an impingement must be sufficient to cause microtrauma in one or more of the subacromial structures. The present study was concerned with determining those shoulder configurations in front-crawl stroke that satisfied the first condition. Two steps were taken to determine such shoulder configurations. In the first step, shoulder configurations known to induce shoulder impingement were measured for each shoulder in a laboratory setting. This measurement determined the range of those shoulder configurations that are potentially harmful because of impingement. In the second step, the shoulder configurations that were exhibited in front-crawl swimming and classified as potentially harmful because of impingement were identified.
The shoulder movements that cause impingement were described by Neer (20,21) and Neer and Welsh (22). These consisted of: (a) a forceful elevation of the arm at or beyond the maximum active elevation of the arm, which configuration of the shoulder is used as an “impingement test” (4,5,11,16,17,19); and (b) voluntary arm elevation in planes within or lateral to the “functional arc” (Fig. 1) when the arm is internally rotated. The magnitude of the internal rotation angle is a key factor when impingement is to be determined. These verbal descriptions found in the literature indicate that impingement occurs not only at certain shoulder configurations but also within a narrow range of combinations of elevation and internal rotation angles for a given vertical plane of arm elevation. One mechanical manifestation of impingement might be that the increased compressive force present under the acromion and/or coraco-acromial ligament inhibits one or more of the three rotational degrees of freedom at the joint. For example, assume that 90° of abduction at the shoulder with the forearm parallel to the frontal axis of the body (90° internal rotation) is the maximum abduction attainable at that internal rotation angle (Fig. 2A). The shoulder allows a further abduction with a reduced internal rotation angle (Fig. 2B) until it reaches the maximum abduction for that reduced internal rotation angle (Fig. 2C). In this case (a) the maximum abduction = 90° when internal rotation = 90°, (b) the maximum internal rotation = 90° when abduction = 90°, and (c) the maximum abduction >90° when internal rotation <90°. In the present study, such combinations of elevation and internal rotation angles were measured throughout the functional range of the shoulder—thus defining an individual’s “boundary range of motion (ROM)”—and used to distinguish those shoulder configurations indicative of impingement.
A member of a college men’s swimming team volunteered to participate in this study. The subject was asked to undergo a boundary ROM testing procedure and to swim in an indoor swimming pool. The subject did not have shoulder problems at the times of the experiments and was able to perform the required activities without restriction. This study was approved by the Review Committee of the University of Iowa, and the subject provided written informed consent.
The performances for the active ROM and the swimming were recorded with three and two cameras, respectively, and the three-dimensional (3D) coordinates of selected body landmarks of the subject were determined using the Direct Linear Transformation algorithm (1). The determined 3D coordinates of the body landmarks were then used to compute the shoulder configurations at each instant by means of three Cardan angles.
The subject was asked to sit on a chair and perform complete elevations of each arm in a series of vertical planes, with the humerus in maximum voluntary internal rotation. The subject was instructed to maintain the hips and trunk against the back support of the chair and to elevate the arm only at the shoulder joint. This prevented hyper-extension of the trunk and assured that the longitudinal axis of the trunk remained vertical. The examiner observed the position of the trunk throughout each trial, and if any motion of the trunk was evident, the trial was repeated.
A light metal rod with two infrared emitting diodes (IRED) at known positions was attached to the lateral end of each acromion using a ball-and-socket joint (Fig. 3). The two IRED on each metal rod were used to determine the position of the center of the ball-and-socket joint, which was used to define the shoulder girdle-embedded reference system (Fig. 4). The orientation of the rod was adjusted for the subject so that the moving arm of the subject did not obscure the camera views of the diodes on the rods at any instant. A humerus marker system (Fig. 3) was used to provide a firm base for attaching IRED on the upper arm. This system consisted of a light-weight aluminum cuff and a hinge. The cuff was fully adjustable to fit securely around the upper arm. Three IRED were attached on each of the four sides of the cuff. The axis of the hinge was placed over the medial epicondyle of the humerus, and the free end was fixed over the medial side of the forearm with athletic tape. With this arrangement, the internal and external rotation of the upper arm could be monitored while still allowing the subject to move the elbow joint freely. A laser pen was attached to the cuff and the emitted light was used to visualize and control movements of the longitudinal axis of the humerus. Strips of paper were hung from the ceiling in an arc centered at the shoulder joint of the subject (Fig. 5). Each strip of paper served as a target for the light ray emitted by the laser pen when elevating the arm.
A WATSMART motion analysis system (Northern Digital, Waterloo, Canada) was used to determine the 3D coordinates of the selected body landmarks. Three time-synchronized cameras operated at 50Hz were fixed on a frame that was firmly attached to the ceiling (Fig. 5). Each camera of the WATSMART system contained a silicon photodiode that generated a voltage in response to infrared light emitted from the diodes attached to the subject. The magnitude of the generated voltage was converted into the 2D coordinates of the point at which the infrared light was located in the optical field. The 2D coordinates of each point measured by the three cameras were transformed into the 3D coordinates of that point.
A reference system was defined for each segment as shown in Figure 4. Shoulder configurations were expressed as a series of instantaneous angular orientations of the right- and left-arm-embedded reference systems with respect to the shoulder girdle-embedded reference system. Three sequential Cardan rotation angles—horizontal abduction angle (HA), elevation angle (EL), and internal rotation angle (IR)—were used to define the angular orientations (Fig. 6). The shoulder girdle-embedded reference system was expected to tilt sidewards in the frontal plane of the trunk as an arm was elevated. This sideward tilt was caused by elevation/depression of the shoulder girdles that would occur naturally with an arm elevation. The sideward tilt (Tilt) was, therefore, measured as the rotation of the shoulder girdle-embedded reference system about the frontal axis of the trunk.
A moving, right-handed, orthogonal reference system (R0: r0s0t0), whose orientation coincided initially with S: xsyszs, was rotated through the angles θ, Φ, and ψ about the s0-, t1-, and r2-axes to form right-handed orthogonal R1: r1s1t1, R2: r2s2t2, and R3: r3s3t3 reference systems, respectively. The initial position of the moving reference frame was one in which the arm was abducted to 90° from the anatomical position, the wrist positioned vertically above the elbow, and the heights of the shoulders were the same. In this initial position, the r0-axis and the s0-axis of the moving reference frame (R0: r0s0t0) coincided with the longitudinal axis of the upper arm and the longitudinal axis of the forearm, respectively. The angles θ, Φ, and Ψ were thus all equal to 0° in the initial position.
The matrices to quantify the three rotations were MATH 1 MATH 2 MATH 3 The rotation matrix to quantify the orientation of the final reference frame (R3: r3s3t3) with respect to the initial (R0: r0s0t0) was MATH 4 MATH 5 This rotation matrix could also be derived from the two matrices AS/I and AR/I for the right shoulder or AS/I and AL/I for the left shoulder. Since these two matrices were expressed with respect to the global reference system (I), the matrix that expressed the orientation of R: xRyRzR with respect to S: xSySzS, or AR/S, was MATH 6 Similarly, the matrix that expressed the orientation of L: xLyLzL with respect to S: xSySzS, or AL/S, was MATH 7 In both cases, MATH 8 Finally, the rotation angles θ, Φ, and Ψ were determined in terms of the elements aij as MATH 9 MATH 10 MATH 11 The angles were converted to describe the anatomical definitions of the shoulder motions as follows:MATH 12
A computation of three Cardan angles is subject to a singularity problem in which no unique solution can be determined. This problem arises when the second rotation angle (EL) is an integer-multiple of ± 90°. In such cases, the first (HA) and the third (IR) rotations take place about the same axis so that no separation of the two angles is possible. In addition, there was a potential inaccuracy in determining the HA and IR angles when the EL angle was close to 90°, and also in determining the IR angle when the elbow angle is close to 0°. This inaccuracy was expected because a given error in determining 3D coordinates of body landmarks could be magnified in the process of defining reference systems and also of computing the three Cardan rotational angles. Reasonable ranges of EL and elbow angles for the computation of HA and IR angles were determined experimentally and will be discussed later.
The determined series of internal rotation angles were expressed as a function of the horizontal abduction and elevation angles. This function was smoothed using a cubic spline function and interpolated to define the maximum internal rotation angle for given combinations of horizontal abduction and elevation angles.
A three-dimensional videography technique with panning periscope systems (28) was used for the data collection. Two periscope systems (Fig. 7) were fixed on the pool deck so that oblique front and back views of the swimmer were recorded. A Panasonic SVHS camcorder mounted on each periscope system was used to record the performances at 60 Hz. A large test section (1.5 × 8.4 × 2.0 m) was defined to cover two complete stroke cycles of a swimmer. The subject was asked to perform one trial of front-crawl swimming at a self-determined stroking speed for a long-distance event and another trial using hand paddles at a self-determined stroking speed for a sprint event. Each trial consisted of two lengths of a 22.9-m pool.
The videotapes from the two camcorders were digitized manually using a PEAK Motion Measurement System (Peak Performance Technologies, Denver, CO) and the resulting sets of two-dimensional coordinate data were used as input to custom-made software that generated the corresponding three-dimensional coordinates. Two stroke cycles from the nonpaddle trial were selected for analysis and one stroke cycle from the paddle trial was selected for accuracy testing. Eight body landmarks (joint centers of the shoulders, hips, elbows, and wrists) were digitized to define the orientations of the trunk, shoulder girdle, and the upper arms.
Three reference systems, S, R, and L, were defined in the same manner as previously described with reference to boundary ROM, except that the longitudinal (yT) axis of the trunk was directed from the middle of the hip joint centers to the middle of the shoulder joint centers. The computational procedure used to determine the instantaneous shoulder configuration was also as described previously.
Shoulder configurations observed during the front-crawl stroke were considered indicative of impingement if one or both of the following criteria were met: (a) the IR angle obtained in the stroke exceeded the maximum active IR angle obtained for the given combination of HA and EL angles when the arm was near or above shoulder height (EL > −10°); and (b) the EL angle observed in the stroke exceeded the maximum active EL angle. If the shoulder configuration lay beyond the reasonable range of EL and elbow angles for an acceptable accuracy in determining HA and IR angles, the criterion (b) was used to identify impingement. The exclusion of such shoulder configurations from the analysis with respect to the criterion (a) was not expected to alter the results substantially for two reasons. First, the elbow was almost completely extended only when the arm was stretching forward, shortly after the arm entry into the water. The EL angle was expected to be at or beyond the maximum active EL angle in this period, and it was thus still possible to determine impingement using the second criterion. Second, the maximum active EL angle was expected to be substantially smaller than 90° because the shoulder girdle-embedded reference system tilted in the same direction as arm elevation. An apparent arm elevation of 90° involved a certain ratio of the measured EL and TILT angles for each subject; and it was expected that this ratio would be similar to the scapulohumeral rhythm. (Inman et al. (12) reported the ratio of the glenohumeral and the scapulo-thoracic movements is generally 2:1 during active arm elevation and “remarkably constant” for each individual.)
RESULTS AND DISCUSSION
Accuracy in measurements.
The validity of the measurement procedure used to define the boundary ROM was tested. The three Cardan angles were determined from precision potentiometers mounted over the three rotation axes on a physical model, and the corresponding angles were determined from the position data of the model segments recorded using the WATSMART motion analysis system. The measurements were taken at various combinations of the three angles. A high correlation (r >0.98) was found for all angles calculated. The test-retest reliability of the determination of the subject’s shoulder ROM was also tested. Several angles representative of the subject’s shoulder ROM, such as the maximum EL angles for various HA angles, the maximum IR angles for various HA angles, and the EL angle at which maximum IR angle was observed, were measured for collegiate swimmers. A high correlation (>0.77) was found for all angles representative of the subject’s shoulder ROM. These results testify, respectively, to the validity and reliability of the data collection technique used to define boundary ROM.
Reliability in determining the shoulder motion exhibited during swimming trials was evaluated using a fast-speed trial of the swimmer using hand paddles. It was expected that a trial under these conditions would provide the least reliable data that might be obtained for swimming trials because the bubbles created by the stroke with hand paddles obscured the body landmarks of the subject more severely than was the case for trials performed under any of the other conditions. A complete stroke cycle recorded on the videotape from each camcorder was digitized twice and four sets of the shoulder motion data were obtained from the four possible combinations of the 2D coordinate data. Assuming that the best available estimate of the true shoulder motion was represented by the mean values for the four sets of shoulder motion data, the difference between the mean and the obtained shoulder motion for a given combination of 2D coordinate data set was considered as the error for that combination.
The results of this test are shown in Tables 1 and 2. The estimated errors were the mean errors determined over the four sets of shoulder motion data; correlation coefficients were determined for every pair of the four sets of shoulder motion data. The accuracy of the HA and the IR angles decreased as the EL angle became close to ± 90° as described before. It is clear from Table 1 that the accuracy of the HA and IR angles were inferior to the accuracy of the other angles. To attain a reasonable accuracy in determining these two angles, appropriate ranges for the EL and elbow angle were defined to limit the computation of the HA and IR. Table 2 shows the results of the test within the ranges of >15° for elbow angle, and −70° to 70° for EL. Within these ranges, the mean errors were less than 2.4° for each angle, and the correlation coefficients for each pair of data were >0.89, indicating a reasonable accuracy for the present study.
The boundary ROM of the subject is graphically displayed in Figure 8. The volume below the surface of the graph indicates those shoulder configurations that were anatomically permissible for the subject. In other words, the shoulder joint was anatomically constrained to move within the ROM indicated by the volume below the surface of the graph. A part of the constraint force that limited the ROM was generated presumably by the compressive force resulting from the impingement. Characteristics of the pattern of the boundary ROM were: (a) the IR angle decreased sharply as the EL angle increased above about 20° (when HA angle was any given value > −30°); (b) the EL angle, at which there was a sharp decrease in IR angle, increased as the HA angle decreased; and (c) the IR angle decreased as the HA angle increased when the EL angle was any given value >20°.
The first characteristic was consistent with the observation first reported by Codman (6) that external rotation is critical to the attainment of full abduction. This characteristic appeared clearly in the range from −10° to 25° of HA angle and 20° to 50° of EL angle. The second and third characteristics appear to be consistent with a statement by Neer and Welsh (22) that elevation of the arm lateral to the functional arc “invites further impingement and recurrence of symptoms.” This statement would mean that the greater tuberosity reaches the acromion easily when an arm is elevated in a plane lateral to the functional arc. The statement implies and the data show that the maximum EL angle attained for a given IR angle or the maximum IR angle attained for a given EL angle was smaller when the arm was elevated in a plane lateral to the functional arc than in a plane within the functional arc. This agreement between the widely-accepted verbal descriptions of the shoulder configurations at which impingement occurs and the corresponding characteristics of the boundary ROM measured in the present study supports the notion that the determined boundary ROM is an acceptable indicator of impingement.
Impingement during front-crawl stroke.
The shoulder configurations exhibited in front-crawl swimming and classified as potentially harmful because of impingement were identified. Figure 9 shows the shoulder configurations of the subject throughout two complete stroke cycles. The graph on the top displays the HA and EL angles of the shoulder exhibited during the stroke cycles. The subject attained the maximum EL angle shortly after the arm entry and exceeded his maximum active EL angle (66.7°), indicating that the shoulder was impinged at and shortly after the arm entry. The values of HA angle remained negative for most of the subject’s stroke cycles, indicating that his stroking action took place in planes anterior to the frontal plane of his body. The graph on the bottom displays the IR angle of the shoulder exhibited during the stroke cycle and the boundary ROM computed for the combination of the HA and EL angles exhibited at every given instant during the stroke cycle. The boundary ROM was computed when −10° < EL <70° and elbow angle >15° (0° was displayed otherwise). The thick horizontal lines shown between the two graphs indicate instants during which the shoulder was impinged because of: (a) the EL angle exceeding the maximum voluntary EL angle; or (b) the IR angle equaling or exceeding the boundary ROM (Fig. 9).
This subject impinged his right shoulder 7% of the stroke time (%ST) at, and shortly after, the arm entry into the water and 5%ST in the middle of the recovery phase. The observed impingement at, and shortly after, the arm entry would suggest that the hydrodynamic force exerted on the hand forcibly elevated the arm beyond the maximum active EL angle. This is the period of the stroke cycle in which the subject had his arm elevated so as to reach well forward. The hydrodynamic force exerted on the hand has a long moment arm about the shoulder joint so that it causes a large elevation torque at the shoulder joint. Such a large torque at the shoulder seems to have driven the arm beyond the maximum elevation angle. The observed impingement in the middle of the recovery phase agrees with the belief that the shoulder motions used in the recovery phase of front-crawl swimming cause the impingement syndrome (5,15,22,24).
There are two major limitations associated with the present method. First, the identification of shoulder configurations (boundary ROM) indicative of impingement was based entirely upon the mechanism of impingement described in the literature. No attempt was made in this study to observe the evidence of impingement within the shoulder joint. A further study is needed to demonstrate the evidence of impingement throughout the entire shoulder range of motion by means of roentgenogram, MRI or ultrasonogram. A use of either of the latter two techniques would allow quantification of the magnitude of deformation made on the subacromial structures. If impingement is to be visualized by means of the above techniques while measuring the shoulder configurations, the measured magnitudes of the IR angle for the boundary ROM are likely to be reduced to a certain degree. This is because a submaximum deformation of the subacromial structure would be observed at submaximum IR angle at a given combination of HA and EL angles. In comparison, the maximum IR angles were used to define the boundary ROM in the present study, so that the subacromial structures must have been deformed maximally at these IR angles and caused a considerable stress on the structures. This suggests that the extent of impingement (%ST) determined in this study is likely to be an underestimation of the “true” extent of impingement that causes damages on the subacromial structures.
The second limitation was associated with the modeling of the body segments. Three body segments (represented by two humerus-embedded reference systems and one shoulder girdle-embedded reference system) were defined and used for the computation of shoulder configuration. In this modeling two shoulder girdles were treated as a single segment and defined the shoulder girdle-embedded reference system. The determined shoulder configuration, therefore, excluded the influence of side-bending and twisting of the torso but included the influences of the configurations at both the glenohumeral and the scapulo-thoracic articulations in the measurements. Ideally, the configurations at the glenohumeral joint should have been isolated so that the boundary ROM could have been determined exclusively at that joint. The boundary ROM measured in this study was, however, considered to manifest the characteristics of the “true” boundary ROM of the glenohumeral joint because the humerus and scapula were known to move synchronously in a consistent manner (7,9,12,23,25). The ratio of the movements at the glenohumeral and the scapulo-thoracic articulations, called scapulohumeral rhythm, is generally about 2: 1, except in the first sixth of the complete arc for abduction (<−60° of EL). The elevation angle at the glenohumeral joint should, therefore, exhibit a consistent relationship with the apparent arm elevation angle, that is, the angle between the upper arm and the torso. This interpretation should be granted in the present study because the computation of the shoulder configurations did not include the influence of the side-bending and twisting of the torso. The high test-retest reliability of the data collection method used to measure boundary ROM would also support the consistent relationship. If the movements at the two articulations failed to maintain a consistent relationship, a considerable amount of variability should have been found in the testing. Hence the boundary ROM measured in this study was considered to represent the characteristics of the “true” boundary ROM at the glenohumeral joint.
There are, however, some conditions in which this interpretation would not apply. Abnormal scapula motion might take place when an individual experiences shoulder pain (23), when muscles are fatigued or weakened (14,18), and when an individual has a pathological condition (3,6,17,23). In the present study none of the above conditions applied to the subject, and it was therefore assumed that the measured boundary ROM was representative of the boundary ROM at the glenohumeral joint.
The procedure described here was developed to identify instances at which impingement occurs during front-crawl swimming. This procedure can also be used to assess impingement in other athletic activities. The results of such studies will help in understanding the mechanism or technical cause of impingement injuries and in improving the techniques to reduce the potential risk to the shoulders. Further studies are needed to confirm the occurrence of impingement by means of advanced visualization techniques such as MRI and ultrasonogram.
This study was supported by grants from the American College of Sports Medicine Foundation and the University of Iowa Student Government. The authors thank James Andrews, Thomas Brown, Thomas Cook, and Warren Darling for their valuable comments which improved the quality of this study. The authors also thank Birgitte van Don, Christel Kippenhan, Ellen Thorson, and Victor Wang for their technical support.
Addrress for correspondence: Toshimasa Yanai, School of Physical Education, University of Otago, P. O. Box 56, Dunedin, New Zealand. E-mail: email@example.com.
1. Allard, P., J. P. Blanchi, and R. Aissaoui. Bases of three-dimensional reconstruction. In: Three-Dimensional Analysis of Human Movement. P. Allard, I. A. F. Stokes, and J. P. Blanchi (Eds.) Champaign, IL: Human Kinetics, 1995, pp. 19–40.
2. Allegrucci, M., S. L. Whitney, and J. J. Irrgang. Clinical implications of secondary impingement of the shoulder in freestyle swimmers. J. Orthop. Sports Phys. Ther. 20:307–318, 1994.
3. Arnheim, D. D. Modern Principles of Athletic Training. St. Louis, MO: Times Mirror/Mosby College Publishing, 1985, pp. 683–725.
4. Ciullo, J. V. Swimmer’s shoulder. Clin. Sports Med. 5:115–137, 1986.
5. Ciullo, J. V. and G. G. Stevens. The prevention and treatment of injuries to the shoulder in swimming. Sports Med. 7:182–204, 1989.
6. Codman, E. A. On stiff and painful shoulders: the anatomy of the subdeltoid or subacromial bursa and its clinical importance. Boston Med. Surg. J. 154:613–620, 1906.
7. Doody, S. G., L. Freedman, and J. C. Waterland. Shoulder movements during abduction in the scapular plane. Arch. Phys. Med. Rehabil. 51:595–604, 1970.
8. Dominguez, R. H. Shoulder pain in age group swimmers. In: Swimming Medicine IV. B. Eridson and B. Furberg (Eds.). Baltimore: University Park Press, 1978, pp. 110–114.
9. Freedman, L. and R. R. Munro. Abduction of the arm in the scapular plane: scapular glenohumeral movements. J. Bone Joint Surg. 48A:1503–1510, 1966.
10. Hall, G. Hand paddles may cause shoulder pain. Swimming World 21:9–11, 1980.
11. Hawkins, R. J. and J. C. Kennedy. Impingement syndrome in athletes. Am. J. Sports Med. 8:151–158, 1980.
12. Inman, V. T., J. B. DeC. M. Saunders, and L. C. Abbott. Observations on the function of the shoulder joint. J. Bone Joint Surg. 26A:1–30, 1944.
13. Jackson, D. W. Chronic rotator cuff impingement in the throwing athlete. Am. J. Sports Med. 4:231–240, 1976.
14. Kamkar, A., J. J. Irrgang, and S. L. Whitney. Nonoperative management of secondary shoulder impingement syndrome. J. Orthop. Sports Phys. Ther. 17:212–224, 1993.
15. Kennedy, J. C. and R. J. Hawkins. Swimmer’s shoulder. Physician Sportsmed. 2:34–38, 1974.
16. Kennedy, J. C., R. J. Hawkins, and W. B. Krissoff. Orthopaedic manifestations of swimming. Am. J. Sports Med. 6:309–322, 1978.
17. Magee, D. J. Orthopedic Physical Assessment. Philadelphia, PA: W. B. Saunders, 1987, pp. 62–91.
18. Matsen, F. A. and C. T. Arntz. Subacromial impingement. In: The Shoulder. C. A. Rockwood and F. A. Matsen (Eds.). Philadelphia: W. B. Saunders, 1990, pp. 623–646.
19. McMaster, W. C. Painful shoulder in swimmers: a diagnostic challenge. Physician Sportsmed. 14:108–122, 1986.
20. Neer, C. S. II. Anterior acromioplasty for the chronic impingement syndrome in the shoulder. J. Bone Joint Surg. 54A:41–50, 1972.
21. Neer, C. S. II. Impingement lesions. Clin. Orthop. Rel. Res. 173:70–77, 1983.
22. Neer, C. S. II and R. P. Welsh. The shoulder in sports. Orthop. Clin. North Am. 8:583–591, 1977.
23. Poppen, N. K. and P. S. Walker. Normal and abnormal motion of the shoulder. J. Bone Joint Surg. 58A:195–201, 1976.
24. Richardson, A. B., F. W. Jobe, and H. R. Collins. The shoulder in competitive swimming. Am. J. Sports Med. 8:159–163, 1980.
25. Saha, A. K. Theory of the Shoulder Mechanism: Descriptive and Applied. Springfield, IL: Charles C. Thomas, 1961, pp. 28–53.
26. Simon, E. R. and J. A. Hill. Rotator cuff injuries: an update. J. Orthop. Sports Phys. Ther.
27. Ticker, J. B., S. Fealy, and F. H. Fu. Instability and impingement in the athlete’s shoulder. Sports Med. 19:418–425, 1995.
28. Yanai, T., J. G. Hay, and J. T. Gerot. Three-dimensional videography of swimming with panning periscopes. J. Biomech. 29:673–678, 1996.