Stroke Pattern in Wheelchair Propulsion

Stroke pattern refers to the trajectory of the hand during the recovery phase of manual wheelchair propulsion. During the propulsive or push phase, the hand follows the path of the handrim. However, during the recovery phase the user can choose any trajectory to prepare for the next push. Four stroke patterns have been identified based on the pattern used during the recovery phase (Shimada et al. 1998; Boninger et al. 2002; Koontz et al. 2009):

  • Semicircular (SC): the hands fall below the hand rim during recovery phase.
  • Single looping over propulsion (SLOP): the hands rise above the hand rim during recovery phase.
  • Double looping over propulsion (DLOP): the hands rise above the hand rim, then cross over and drop below the hand rim during the recovery phase.
  • Arcing (ARC): The third metacarpophalangeal (MP) follows an arc along the path of the hand rim during the recovery phase.

The following articles examine the stroke patterns as well as the kinetics and kinematics of the different stroke patterns in relation to the potential for upper extremity injury due to suboptimal biomechanics and/or chronic overuse.

Author Year

Research Design

Total Sample Size

Methods Outcome
Kwarciak et al. 2012




Population: Mean age: 35.7 yr; Gender: males=23, females=2; Level of injury: paraplegia (T3-L1)=17, spina bifida(T10-L1)=6, tetraplegia(C6-7)=1, spinal lipoma=1; Mean use of w/c:16.9 yr.Intervention: Four propulsion patterns (single loop (SL), arcing (ARC), double loop (DL) and semi-circular (SC)) were compared to the participants’ normal pattern. Parameters measured were cadence, peak force, contact angle, braking moment, and impact, as well as EMG muscle activity in specific upper extremity muscles or muscle groups.

Data collection was completed for each participant’s normal pattern after an acclimation period. Subsequent stroke patterns were randomly assigned with a period of instruction and practice prior to data collection. Each data collection period lasted 60 sec with 30 sec warm up prior and rest times between to avoid fatigue.

Outcome Measures: Surface electrodes were used at to measure muscle activation at the shoulder (upper and middle trapezius, pectoralis major, anterior, middle and posterior deltoid), elbow (long head of triceps and biceps), and wrist (wrist extensors and flexors). Data for stroke pattern were collected on the right hand (MCP joint) and wheel (3 points on the hub of wheel). Propulsion variables were measured by an instrumented rear wheel while the participant propelled on a wheelchair treadmill that was normalized to the individual’s parameters on low pile carpet as determined at the start of the study.

1.     Normal propulsion patterns: DL=15, SL=6, ARC=2, SC=2.

2.     Comparisons across patterns were based on average of normal (across low pile carpet and self-selected speed) and experimental propulsion trials.

3.     Hand rim biomechanics: DL=smallest cadence, largest contact angle, smallest braking moment compared to ARC pattern (all p<0.05). The latter 2 were also significantly different than the SL pattern (p<0.05). Though not significant, DL had highest peak force value and SC the lowest peak force as well as lowest impact.

4.     Contact angle of SC was significantly larger compared to arching pattern (p<0.05).

5.     Muscle activity: No significant differences were found in muscle activity between stroke patterns.

Raina et al. 2012b




Population: Mean age: 74.5 yr; Gender: males=31, females=3; Level of injury: paraplegia=16(T6-L1), tetraplegia=18(C6-7), all AIS A or B motor complete; Mean height: 1.75 m.

Intervention: Participants propelled their own manual w/c on a stationary ergometric normalized to propelling on tile floor for a 30 sec period to achieve steady state propulsion followed by 10 sec of data collection for each of four propulsion patterns (arcing (ARC), single–loop-over propulsion (SLOP), semi-circular (SC), double–loop-over propulsion (DLOP)).

Outcome Measures: Push pattern analysis included velocity prior to contact, peak impact force, and the effectiveness of the force at impact. Force was measured at the contact point with the hand rim for the period when force was more than 5 N as measured using the Smart Wheel (3 strain force transducers). Propulsion patterns were tracked using a 6-camera system with 16 reflective markers placed on the manubrium, xiphoid process, spinous processes of T3&T10, greater tubercle of the humerus, medial and lateral epicondyles, deltoid tuberosity, mid forearm, radial and ulnar styloids, and head of 3rd and 5th metacarpals, three markers on the wheel.

1.     Velocity of wrist prior to contact was significantly correlated (r=0.74, p<0.05) with the magnitude of impact force for all participants; tetraplegia=0.81±0.24 m/second, 0.062±0.02 N/kg; paraplegia=0.95±0.37 m/second, 0.061±0.03 N/kg.

2.     Correlation between wrist velocity prior to contact and magnitude of impact force normalized to body weight was stronger for participants with paraplegia (r=0.92) than tetraplegia (r=0.45).

3.     No significant differences in magnitude of impact force between participants with paraplegia and tetraplegia (p>0.05).

4.     Participants with tetraplegia had significantly higher (p=0.02) radial component of impact force than participants with paraplegia (9.2% & 4% respectively).

5.     Percent of impact force applied in tangential direction (effective force) was significantly higher (p=0.005) in paraplegia group (94%) than in tetraplegia group (88%) – suggest lower effectiveness of force application at impact for tetraplegia group.

6.     ARC, SC and SLOP patterns were preferred by both participant groups.

7.     The most common propulsive pattern in the combined sample population was the SLOP.

8.     DLOP not used by participants with tetraplegia; the SC pattern was observed in only one participant with paraplegia.

9.     Impact force between hand movement patterns was not significantly different between patterns (p>0.05) (force normalized to arm weight to account for between subject body mass differences).

10.   Force effectiveness was not significantly different between propulsion patterns.

11.   Percent of effective force at contact varied between 0-25% and 25-95% for participants with tetraplegia and paraplegia, respectively.

12.   The same pattern showed different percentages of force effectiveness in the two participant groups (paraplegia versus tetraplegia).

Feng et al. 2010




Population: SCI (n=9); Mean age: 28.9 yr; Gender: NR; Level of injury: Lumbar=2.2, Thoracic=7.8; Mean time since injury: 11.3 yr; Experience using manual w/c range 2-18 yr.

Intervention: To investigate the glenohumeral kinematic difference between circular and pumping stroke wheelchair propulsion in glenohumeral joint (GHJ) excursion related to shoulder impingement (defined as internal or external rotation beyond 30° of forward flexion or 30° of abduction).

Participants used a study w/c set up to standardize arm position in an optimal position in relation to wheel. Testing done on a roller system, following a protocol of 5 min warm up and three tests of 10 cycles of propulsion for each propulsion pattern; patterns randomly assigned.

Outcome measures: Zebris Motion analysis system with six markers (acromion process, lateral epicondyles, ulnar styloids, and a rigid cross placed on sternum to capture three planes) to measure temporal parameters [push time(s); recovery time (s); push phase (% of cycle); recovery cycle (% of cycle)] and kinematic parameters [Initial and end position flexion-extension, abduction-adduction, and internal-external rotation (degrees)] of each propulsion technique, in addition to impingement excursion.

1.     There were not significant differences in the temporal variables between the two stroke techniques (similar time spent in the pushing and recovery movements).

2.     Circular and pumping strokes showed a ratio of 4:6 between push and recovery times.

3.     In the sagittal plane the starting and ending positions were similar between the two stroke techniques with both starting and ending with approximately 40° of shoulder extension.

4.     There were significant differences between stroke patterns in the frontal and transverse planes;1) on average pumping stroke compared to circular started in larger abduction (56.6°+9.5° versus 44.7°+7.4°, p=0.001), and internal rotation (3.6°+10.3° versus-10.3°+6.7°, p=-.020). 2). End position for pumping was larger than circular for abduction (57.6°+5.1° versus 45.4°+6.2°. p=0.001) and internal rotation (34.1°+11.8° versus-13.4°+7.3°, p=0.001).

5.     The pumping stroke also had a significantly greater excursion in the sagittal, (71.4°+11.4° versus 55.9°+ 11.8°, p=0.001), frontal, (57.6°+5.1° versus 45.4°+6.2°) and transverse planes (42.4°+11.8° versus 25.7°+7.3°) compared to the circular stroke.

6.     A greater percentage of the GHJ movement met impingement excursion (almost three times) during the pumping stroke compared to circular stroke (11.6+11.2% versus 30.9+6.0%, t=-4.670, p<0.001).

Koontz et al. 2009




Population: Mean age: 47.0 yr; Gender: males=28, females=1; Injury etiology: SCI=24 (cervical=5, thoracic=14, lumbar=5), amputation=3, neuropathy=1, spina bifida=1; Length of time using w/c: 14.2 yr.

Intervention: Patients propelled their manual wheelchairs on randomly selected test surfaces consisting of linoleum (1.20 m by 4.50 m), high-pile carpet (1.50 m by 4.50m) and a plywood ramp (1.20 m by 3.60 m, 5° grade) for three test trials.

Outcome Measures: 2 SMARTWheels and a camera set up to collect data for stroke pattern and propulsion variables of applied force, velocity, distance per stroke, contact angle and moment.

1.     The single looping (SL) over propulsion pattern was most commonly used for the initiation of motion (44.9%), followed by arc (35.9%), double looping (DL) over propulsion (14.1%) and semicircular (SC) pattern, (5.1%).

2.     The number of strokes used and the type of surface had no significant effect on the pattern used.

3.     Body weight, body and wheelchair weight combined, and age were not significantly different between patterns

4.     Duration of wheelchair use was significantly different between patterns types onlinoleum for the 1st and 2nd strokes. (p=0.036 and p=0.008 respectively) Participants in the DL and SC pattern group had been using wheelchairs longer (stroke 1: DL/SC=28.0±12.5 yr, SL=11.8±9.7 yr, arc=13.7±8.0 yr; stroke 2: DL/SC=22.0±11.5yr, SL=10.3±6.7 yr, arc=10.5±6.7 yr).

5.     On linoleum:

·  Between group differences approached significance in regard to contact angle with DL/SC having a larger contact angle at stroke 1 (p=0.069) (DL/SC=56.70±11.10 °, SL=45.00±5.55 °, arc=31.30±5.1 °).

·  Between group differences approached significance in regard to average velocity with DL/SC having a faster average velocity (p=0.075) (DL/SC: 0.92±0.06 m/s, SL=0.75±0.06 m/s, arc=0.73±0.07 m/s)

·  DL/SC covered significantly more distance per stroke at stroke 2 compared to arc (p=0.016) (DL/SC=0.53±0.08 m, arc=0.44±0.10 m).

6.     On carpet:

·  Between group differences were significant in regard to peak moment at stroke 3 (p=0.009) (DL/SC=0.26±0.02 m, SL=0.23±0.01 m, arc=0.18±0.02m), average velocity at stroke 3 (DL/SC=1.07±0.08 m/s, SL=0.82±0.06 m/s, arc=0.70±0.09 m/s) and distance per stroke at stroke 3 (p=0.036) (DL/SC=0.53±0.12 m, SL=0.45±0.08 m, arc=0.42±0.13 m).

·  Compared to arc, DL/SC had a significantly greater peak moment (p=0.07), average velocity (p=0.019) and distance per stroke (p=0.043) at stroke 3.

7.     On the ramp:

·  Between group differences were significant in regard to peak resultant force at stroke 3 (p=0.049) (DL/SC=1.64±0.20, SL=1.37±0.11, arc=1.07±0.13).

·  Compared to arc, DL/SC had a non-significantly greater peak resultant force at stroke 3 (p=0.066).

Richter et al. 2007a




Population: Mean age: 36.0 yr; Gender: males=19, females=7; Mean wheelchair use=17 yr; Mean weight: 69.8 kg; Level of injury: paraplegia; Chronicity=chronic.

Intervention: Self propulsion in personal wheelchair on a treadmill set to level, 3° and 6° grades.

Outcome Measures: Stoke pattern – semicircular (SC), single looping (SLOP), double looping (DLOP), arcing (ARC), Speed, Peak force, Push angle, Push frequency, Power output.

1.      Level stroke pattern: 42% ARC; 30% SLOP; 27% DLOP; 0% SC.

2.      3° slope stroke pattern: 69% ARC; 19% SLOP; 12% DLOP; 0% SC.

3.      6° slop stroke pattern: 73% ARC; 23%SLOP; 4% DLOP; 0% SC.

4.      From level to 6° slope: 63% decrease in speed (p=0.000); 218% increase in peak force (p=0.000); 25.5% decrease in push angle (p=0.002); 21.6% decrease in push frequency (p=0.042).

5.      Power output at 3° slope and 6° slope were 2.8 and 3.1 times higher than those at level (p=0.000).

Boninger et al. 2002




Population: Mean age: 35.1 yr; Gender: males=27, females=11; Mean weight=167.2 lbs; Mean height=69 in; Handedness: left handed=5, right handed=33; Level of injury: paraplegia=38; Mean time since injury: 11.1 yr.

Intervention: Self propulsion of personal wheelchair on a dynamometer at 0.9m/sec and 1.8m/sec.

Outcome Measures: Stroke pattern – semicircular (SC), single looping (SLOP), double looping (DLOP), arcing (ARC); Axle position; Beginning stroke angle; Total stroke angle; Cadence; Mean velocity; Push time; Recovery time; Total time in propulsion.

1.      Stroke patterns observed: 45% SLOP; 25% DLOP; 16% SC; 14% ARC.

2.      58% used similar stroke patterns at both speeds, on both sides; however, the remaining subjects alternated patterns between sides and speeds. Most notably, SC pattern use decreased as the speed increased.

3.      DLOP and SC patterns had lower cadence than ARC (p<0.01) and SLOP (p<0.05).

4.      ARC and SC spent the most time in propulsion (p<0.05).


The above studies have investigated the effectiveness of stroke patterns in wheelchair propulsion in the spinal cord injured population. Boninger et al. (2002) studied the stroke patterns of 38 individuals with paraplegia while propelling their own wheelchair on a dynamometer at two different steady state speeds. The SC and DLOP patterns were found to have significantly lower cadence and the least time spent in each phase of propulsion. The SC and ARC patterns had the greatest amount of time spent in propulsion relative to the recovery phase. A correlation has been found between cadence and the risk of median nerve injury (Boninger et al. 1999). The authors concluded a stroke pattern that minimized cadence may reduce the risk of median nerve injury.

Richter et al. (2007a) studied the stroke patterns of 25 individuals with paraplegia propelling their own wheelchairs at self-selected speeds on a treadmill set to level, 3° and 6° grades. In this study, the SC pattern was not used by any of the subjects. For level propulsion, the number of subjects using the remaining three patterns was fairly evenly distributed. However, once the subjects started going uphill 73% of participants used the ARC pattern. No significant difference was found in the handrim biomechanics between the different stroke patterns. The authors caution against training wheelchair users to adopt a certain pattern until more is known about the consequences.

Kwarciak et al. (2012) investigated the effects of the four different stroke patterns on hand rim biomechanics and upper extremity electromyography (EMG) in people experienced with w/c use. They found variability in the participants’ chosen normal propulsion stroke patterns, with 60% using a double loop pattern, 24% using the single loop pattern, and 8% each for using the ARC pattern and the semi-circular pattern. Despite the few significant values in the study, the authors felt the findings supported the recommendations for upper limb preservation that less frequent, long smooth strokes are required. The DL and SC patterns generated the best combination of biomechanics producing the longest contact angle, lowest cadence values, and smallest braking moments. While there were no significant values, the DL also has the advantage of 35% lower elbow muscle activity. However, the authors recommend that users individual style and comfort drive decision between the two (i.e., imposing changes from one pattern to the other is not needed) The authors did question the viability of the single loop pattern, as it produced the largest contact impact at the hand rim, the largest amount of muscle activation and the second worst values for cadence, peak force contact angle and braking moment. The arching pattern results in this study produced suboptimal handrim biomechanics but the low muscle demand is the most metabolically efficiency, to which the authors suggest may be useful for uphill propulsion.

Raina et al. (2012b) identified the purpose of their study as threefold; 1) to determine whether the stroke propulsion pattern affects the magnitude of hand/forearm velocity prior to hand rim contact, 2) to determine if the hand movements of one of the four typical stroke patterns results in a higher effectiveness of propulsion and 3) if differences in propulsion patterns exist between participants with paraplegia and tetraplegia. No differences were noted between patterns, but significant differences were found between the participant groups of paraplegia and tetraplegia. The differences were primarily in the wrist velocity prior to contact with the participants with paraplegia being more highly correlated to magnitude of force impact compared to the participants with tetraplegia, but both correlations were significant. Similar findings were noted for effectiveness of impact forces, with the participants with paraplegia having significantly greater impact force effectiveness than participants with tetraplegia.

Also noted was a difference in muscle activity particularly for the participants with tetraplegia who had a higher radial force impact. The authors noted that the difference in radial force impact may be related to reduced force effectiveness in this group (i.e., weaker grip strength affecting sustained contact with handrim). Therefore, study authors proposed that radial force may have been used by participants with tetraplegia to increase friction on the hand rim during the push phase. Given that in this study all participants with tetraplegia demonstrated low impact force effectiveness in all stroke patterns for propulsion, improving the effectiveness of the impact force or reducing the magnitude of impact force would require alternate means of increasing friction at the hand pushrim interface (e.g., friction gloves) or alterative mechanisms for propulsion (e.g., power assist wheels). These differences in the initial push phase of propulsion between paraplegic and tetraplegia injury levels hold important considerations for maintenance of upper extremity health.

Koontz et al. (2009) explored propulsion patterns, and kinetic and kinematic variables at start up propulsion over a linoleum floor, a carpeted floor and a 5° incline ramp with 29 people with spinal cord injury who used manual wheelchairs. They defined start up as the first three push strokes from a stopped position based on other larger study results. The authors reported that some patterns were difficult to discern, and some were hybrids of two propulsion patterns, therefore using three raters to gain consensus. They found that on any surface, the most common first stroke pattern was an ARC, however those who switched after the first stroke to an under-rim pattern reached higher velocities and experienced fewer negative forces during start up than those who stayed with an ARC pattern. The only exception to this was the ramp, where many participants continued to use the ARC propulsion pattern. The authors speculate this is related to the tendency of the wheelchair to roll backwards on the ramp during the recovery phase; the ARC pattern has a shorted recovery phase. The impact of the first three stroke patterns on function and upper extremity maintenance is seemingly minimal until the consideration of the frequency of start/stop occurrences throughout the day is considered. The authors suggest greater attention needs to be paid to the start up of propulsion in propulsion training particularly the patterns used.

Feng et al. (2010) examined the kinematic differences between two stroke propulsion patterns (pumping and circular) with a focus on the glenohumeral joint excursion as related to shoulder impingement. Based on the research literature they defined impingement as “…contact between the anterior aspect of the humerus and the acromial arch which creates compressive forces on the glenohumeral joint” (p. 448), with a range of internal or external rotation beyond 30° of forward flexion or 30° of abduction. The study wheelchair was adjusted for each participant for optimal propulsion positioning (i.e., 30° elbow flexion when hand on top of rim, distance between rear corner of seat and axis equaled 15% of participant’s arm length). The authors concluded that the pumping stroke pattern of propulsion traveled more and stayed longer in the impingement range than the circular stroke pattern.

The authors indicated that further study is required to determine if this range of glenohumeral joint excursion is related to shoulder impingement injuries, and if the use of the pumping stroke style contributes to shoulder impingement injuries. There are, however, a few limitations of this study, which make it difficult to generalize the findings to clinical practice. The first is the small study size (n=10). The second is the use of a pre-determined set up for the study wheelchair as opposed to examining the participant in their own w/c set up. The third is the use of only two stroke patterns, it is not clear why the authors identified only two stroke patterns and did not related them to patterns identified in the literature despite referencing articles where the four stroke patterns are identified. The fourth is the limited description of the amount of internal and external rotation that is considered as part of the definition of shoulder impingement.


There is level 4 (from four post-test studies: Boninger et al. 2002; Ritcher et al. 2007; Raina et al. 2012b; Kwarciak et al. 2012) evidence that the typical propulsion stroke patterns used by individuals with spinal cord injury varies across the four stroke patterns regardless of level of injury.

There is level 4 (from one post-test study: Boninger et al. 2002) evidence that the semicircular and double-loop-over propulsion wheelchair stroke patterns reduce cadence and time spent in each phase of propulsion, thus using these patterns may reduce the risk of median nerve injury.

There is level 4 (from two post-test studies: Ritcher et al. 2007; Raina et al. 2012b) evidence that there is no difference in hand rim biomechanics during propulsion between the four stroke patterns. However, there is also level 4 (from two case series studies; Boninger et al. 2002; Kwarciak et al. 2012) evidence that the semicircular and double-loop-over propulsion stroke patterns offer the best combination of biomechanics for propulsion.

There is level 4 (from one post-test study: Raina et al. 2012b) evidence propulsion biomechanics differ between people with paraplegia and tetraplegia with the latter group producing lower wrist velocity prior to contact, less magnitude of force impact, and higher radial force.

There is level 4 (from one post-test study: Feng et al. 2010) evidence that the movements associated with particular patterns may increase the risk of shoulder impingement, with pumping stroke pattern exposing the shoulder to greater risk than the circular pattern.

There is level 4 (from two post-test studies: Kwarciak et al. 2012; Boninger et al. 2002) evidence that the ARC stroke pattern has suboptimal biomechanics, but the lowest muscle demand, therefore holds potential for making it useful for short duration, high force propulsions such during ascending a hill or ramp.

There is level 4 evidence (from two post-test studies: Koontz et al. 2009; Richter et al. 2007a) to suggest that the ARC pattern is the most frequently used propulsion pattern used when ascending a slope greater than 3⁰.

There is level 4 evidence (from one post-test study: Koontz et al. 2009) to suggest that it takes the first three propulsion strokes from a resting positioning to reach steady state velocity and while the ARC pattern is most frequently used for the first stroke, those who change to an under-rim pattern for the subsequent strokes, reach steady state velocities quicker and experience less negative mechanical forces during start up propulsion.