OUP user menu

Stride length regulation in Parkinson's disease: the use of extrinsic, visual cues

Gwyn N. Lewis, Winston D. Byblow, Sharon E. Walt
DOI: http://dx.doi.org/10.1093/brain/123.10.2077 2077-2090 First published online: 1 October 2000


It has been well documented that marked improvements in the hypokinetic gait pattern of Parkinson's disease patients are possible with the use of appropriate visual cues. This project served to evaluate Parkinson's disease gait performance as well as residual processing capacity while using fixed or gait-regulated visual cues. Three-dimensional kinematic, kinetic and electromyographic gait analysis was carried out on 14 patients and 14 matched controls in baseline conditions and with two types of visual cues: taped step length (SL) markers and an individualized subject-mounted light device (SMLD). A probe reaction time paradigm was invoked to assess residual processing capacity. Ratings of perceived task load were also made using the NASA-Task Load Index. Stride length and gait velocity were reduced in patients in baseline conditions. Both of these parameters increased to control levels with the use of visual cues. These alterations were generally accompanied by modifications of lower limb kinematics and kinetics towards control subjects. Perceived task load was higher in all conditions and was further elevated by the use of the SMLD for both groups. Patients produced larger overall reaction times, although reaction time was not different between baseline and SL marker conditions. Reaction time was increased in both groups when using the SMLD. The overarching finding is that stride length can be regulated in Parkinson's disease using stationary visual cues without increased central processing capacity or perceived effort. This may occur via utilization of visual feedback, reducing the patients' reliance on kinaesthetic feedback for the regulation of movement amplitude.

  • gait
  • Parkinson's disease
  • basal ganglia
  • electromyography
  • CV = coefficient of variation
  • FSR = force sensitive resistor
  • r.m.s. = root mean square
  • SL = step length
  • SMA = supplementary motor area
  • SMLD = subject-mounted light device


Disorders of gait are one of the most common symptoms of Parkinson's disease (Hoehn and Yahr, 1967). Specifically, patients tend to demonstrate a shuffling gait pattern with a shortened stride length and a reduced overall velocity. Cadence may be normal in comparison with controls, but is often elevated in relation to the individual's own velocity. Previous 2D kinematic studies have also revealed reductions in ankle, knee and hip joint range of motion and substantial increases in trunk flexion angle (Knuttson, 1972; Peterson et al., 1972; Murray et al., 1978; Dietz et al., 1981; Blin et al., 1990; Pedersen et al., 1997). These gait disorders progressively worsen as the disease advances, severely limiting the patient's quality of life.

Several studies have demonstrated that Parkinson's disease patients can increase stride length if appropriate cues are provided or indicated. Although improvements in the spatiotemporal gait pattern of patients have been demonstrated with the use of instructional (Behrman et al., 1998) and auditory cues (Thaut et al., 1996; McIntosh et al., 1997), the most successful type of external cues appear to be visual (Morris et al., 1994b). In a classic study by Martin, patients were asked to walk along a runway with a variety of visual cues placed upon it (Martin, 1967). From this, it was revealed that when patients walked over certain types of cues improvements in stride length were evident. The most effective cues were those perpendicular to the walking path and spaced about one step length (SL) apart. Since this time, a number of subsequent studies (e.g. Morris et al., 1994b, 1996) have demonstrated that with appropriate visual cues Parkinson's disease patients can produce a gait pattern of normal velocity, cadence and stride length, highlighting that the specific deficit in Parkinson's disease is the internal regulation of stride length.

What is less certain in Parkinson's disease are the precise mechanisms behind the reductions in stride length and the mechanisms behind the improvements in stride length that arise with the use of visual cues. Parkinson's disease arises through a reduction in dopaminergic neurons in the substantia nigra of the basal ganglia, a group of nuclei implicated in the control of automatic, well-learned movements (Roland et al., 1980). In particular, these are sequential movements such as eating, writing, talking and walking. From previous research on normal and pathological basal ganglia functioning, it has been proposed that the basal ganglia are implicated in two main roles in the control of these movements (Morris et al., 1995), primarily through their interaction with the supplementary motor area (SMA). The first role is as an internal cue or trigger to enable movement sequences to be carried out without attention. During a movement sequence, phasic output from the basal ganglia inhibits the thalamus which projects to the SMA, triggering the release and completion of the forthcoming sub-component of the sequence (Brotchie et al., 1991). This affords a means for well-learned movement sequences to be internally regulated and carried out automatically. The second role is its contribution to cortical `motor set'. The basal ganglia aid in the preparation and maintenance of motor plans in a state of readiness for action, enabling motor functions to be carried out functionally and appropriately (Robertson and Flowers, 1990). Although dysfunctions in both of these roles of the basal ganglia have been implicated in the stride length deficiencies evident in Parkinson's disease, recent studies have proposed that they may be more likely related to a disorder of motor set (Morris et al., 1994b).

The mechanisms underlying improvements in gait with the use of visual cues are less well understood. From the results of a dual-task study on Parkinson's disease gait, Morris and colleagues suggested that attention was an important factor in the enhancements in stride length (Morris et al., 1996). In this study it was found that the improvements in stride length seen after the use of visual and attentional cues were reduced when patients were asked to carry out a secondary task while walking. The reduction in stride length evident was proportional to the complexity of the secondary task. The authors proposed that the visual cues draw the attention of the patients, thereby invoking alternative, more conscious, motor control pathways in the regulation of gait, enabling the faulty basal ganglia to be bypassed. As an alternative explanation, Azulay and colleagues suggested that the impression of dynamic visual cues achieved when patients walk over a striped runway may improve motor performance (Azulay et al., 1999). They found that when stroboscopic lighting was implemented to conceal the dynamic visual cues, the improvements in gait with a striped runway were suppressed in Parkinson's disease patients. It was hypothesized that facilitation of a motion-responsive visual-motor pathway may lead to the gait improvements in Parkinson's disease.

Accurate descriptions of the biomechanical alterations associated with stride length reductions in Parkinson's disease are difficult as to date there is a distinct lack of 3D gait analysis data on patients. A case study on a 71-year-old female patient by Morris and colleagues revealed reductions in ankle power generation at toe-off that may be a substantial contributor to the shortened stride length seen (Morris et al., 1999). In visually cued walking conditions, where stride length was increased, there were improvements in ankle power; however, it was still significantly less than that of control subjects.

Studies on Parkinson's disease gait utilizing EMG are also limited in number. However, a consistent finding of these has been the maintenance of the phasing of muscle activation patterns during the stride cycle, particularly in the distal lower limb muscles (Dietz et al., 1981; Cioni et al., 1997). Another common feature evident in Parkinson's disease patients is a reduction in the amplitude of gastrocnemius activity during the stance phase (Dietz et al., 1981, 1997; Cioni et al., 1997; Dietz, 1997). Cioni and colleagues found evidence of this in patients both ON and OFF medication, although to a lesser extent during the ON phase (Cioni et al., 1997). The reduction in plantar flexion activation provides a likely explanation for the diminished ankle power and shortened stride length seen in Parkinson's disease patients, although studies combining EMG and kinetic/kinematic video analysis to investigate this are lacking. Also lacking are EMG studies of patients in augmented walking conditions, such as when utilizing visual cues.

In light of the need for further research established above, the general aim of the current study was to investigate the biomechanical and possible neurological mechanisms behind the gait disorders in Parkinson's disease, and behind the improvements seen with the use of visual cues. Specifically, we intended to (i) conduct 3D gait analysis, including EMG, of Parkinson's disease patients while in normal and visually cued walking conditions, (ii) assess residual processing capacity and the perceived effort of walking of patients in normal and visually cued walking conditions. One of the two types of visual cues implemented was a laser cueing device developed specifically for use in the study. This device was designed as a practical alternative to SL markers and also provided a means for specifying stride length that did not alter the dynamic visual flow in comparison baseline (non-cued) conditions.



Fourteen idiopathic Parkinson's disease patients and 14 age-, sex- and height-matched controls volunteered for the study. Details of the subjects are shown in Tables 1 and 2. The patient group ranged in age from 58 to 84 years (mean 71.3 ± 7.6 years) and had an average height of 1.69 ± 0.09 m. The average age and height of the control group were 70.5 ± 6.5 years and 1.71 ± 0.07 m, respectively. The patients had an average estimated Hoehn and Yahr rating (Hoehn and Yahr, 1967) of 2.8 ± 0.8 and an average disease duration of 9.1 ± 5.7 years. All testing was carried out while the patients were in a self-diagnosed ON state of medication, no less than 20 min post-medication intake.

View this table:
Table 1

Details of patient group characteristics

PatientAge (years)SexHeight (m)Weight (kg)H and YYears diagnosedMedicationDosage (mg/day)
F = female; M = male; H and Y = Hoehn and Yahr scale; NA = not available.
184M1.76662.5 4Sinemet400/100
280M1.82783 5Sinemet600/150
378M1.6468310Sinemet CR800/200
476F1.64702.517Sinemet CR400/100
574F1.52604 6SinemetNA
672M1.79792 5Sinemet CR600/150
772M1.83612 3..
872M1.6471314Sinemet CR1100/275
1069M1.7575313Madopar HBS500
1167F1.65663 7Sinemet CR600/150
1264F1.62591.5 2Madopar750
1457M1.63641 6Sinemet CR600/150
View this table:
Table 2

Details of control group characteristics

ControlAge (years)SexHeight (m)Weight (kg)
NA = not available.

To be included in the patient group, subjects were required to be without postural instability, have no other neurological conditions and be able to walk along a 10 m walkway at least 60 times, with rest breaks. It was also desired that the patients demonstrated a shuffling gait pattern or a shortened stride length while walking, ON medication. All subjects were required to have normal or corrected-to-normal hearing and vision, no cardiovascular limitations impairing walking ability/efficiency, or any orthopaedic limitations or dyskinesias affecting gait.


Spatiotemporal gait parameters

Force sensitive resistor (FSR) foot switches of 1–2 cm diameter were inserted into the insoles of the subjects' shoes corresponding to the positions of the heel and the base of the first metatarsal of the foot. Foot switch data were used to define the gait cycle (the time of activation of a heel foot switch of one leg to its subsequent activation) and to determine cadence. Two sets of timing lights (On Spot LTX 2000 Sprint Performance System) set up 1 m in from the beginning and end of the runway enabled the determination of average gait velocity and, in combination with the foot switches, provided a mean value of stride length along the runway. The between-trial coefficient of variation (CV) of velocity, cadence and stride length was also determined using the following formula:Math

Gait analysis

Three-dimensional kinematic and kinetic gait analyses were conducted using motion analysis equipment and a full body marker set. Seventeen retroflective markers were placed on the lower limbs and trunk of the subjects, enabling 3D analysis of foot, shank, thigh, pelvis and trunk body segments. Five Falcon cameras (sampling speed 60 Hz) were placed around the laboratory to capture subjects walking in the middle two-thirds of the runway. Two Bertec force plates (40 × 60 cm and 60 × 90 cm) were incorporated into the runway at this position to obtain ground reaction force, which was used in a mathematical model with the video data to estimate joint moments and powers. Video data were collected and processed using EVa and Orthotrak (Motion Analysis CorporationTM, San Diego, Calif., USA) collection and analysis programs.


EMG activity of tibialis anterior and soleus muscles was recorded using silver/silver chloride Pellet electrodes with a 0.5 cm active surface. Standard skin preparation techniques were used to reduce impedance. Two electrodes were placed 2 cm apart on the belly of each muscle in line with fibre direction. Soleus electrodes were placed 2.5 cm distal to the medial head of the gastrocnemius muscle (Winter and Yack, 1987). Tibialis anterior electrodes were placed over the area of greatest bulk just lateral to the crest of the tibia on the proximal half of the leg. A ground electrode was attached to the subject's non-dominant elbow over the lateral epicondyle.

EMG signals were recorded at 1200 Hz with a bandwidth of 30–600 Hz. The raw signals were pre-amplified, full-wave rectified, time normalized to stride cycles and averaged over each muscle, leg and walking condition. The averaged stride cycles were then divided into eight even intervals of time. For each interval the root mean square (r.m.s.) amplitude of the signal was determined and for each subject was expressed as a percentage of the maximum r.m.s. amplitude for that muscle and leg.


In accordance with the declaration of Helsinki, the University of Auckland Human Subjects Ethics Committee approved the procedure and informed consent was obtained from all subjects prior to testing. Upon arrival at the laboratory the subjects were prepared for data collection. This process included the placement of the 3D marker set, EMG electrodes and foot switches and took ~45 min. The subjects then completed the simple reaction time tasks and walking trials. All testing was completed in 1 day and the same protocol was used for both patient and control subjects. The total duration of the test sessions ranged from 1.5 to 2.5 h (controls) and 2.5 to 4 h (patients).

Walking conditions

Subjects completed walking trials in three different conditions. The first of these was a baseline condition, the remaining two were visually cued conditions [step length (SL) marker and subject-mounted light device (SMLD)]. In each condition the subjects walked up and down a 10 m runway ~10 times (20 trials), with rest breaks provided as necessary. All subjects completed the baseline condition first while the order of the following two visually cued conditions was randomized between subjects.

In the baseline condition the runway was featureless and instructions to the subjects were simply to `walk to the end of the runway at your normal speed'. For the SL marker condition, strips of white tape (5 × 50 cm) were placed along the runway, perpendicular to the walking path, at intervals corresponding to a normal SL for each subject, as determined by their height (Winter et al., 1990). Instructions in this condition were to `walk to the end of the runway by stepping over the lines'. The SMLD condition involved the use of a laser device that was attached to the subject's chest (Fig. 1). The device projected two laser lines on to the floor in front of the subject that were approximately 50 cm wide and spaced the same SL apart as that indicated by the SL markers. In this condition the subjects were instructed to `step up to the line as you walk along the runway'. The lines were projected continuously while the subjects were walking.

Fig. 1

SMLD (A) and SMLD mounted on a subject (B).

Residual processing capacity

To assess residual processing capacity while walking, a dual-task paradigm involving a probe reaction time task was invoked. This methodology has been previously utilized to assess the attentional demands of voluntary movement, including gait (Lajoie et al., 1993). In 16 (80%) of the walking trials in each of the three conditions, a light vibratory stimulus was delivered to the back of the subject's dominant hand via a stimulus band. The subjects were asked to respond to this stimulus as quickly as possible by pressing a button switch that was carried in the same hand. The stimulus was randomly preceded by an auditory warning signal to draw the subject's attention to the impending task. Simple reaction time was also assessed using the same apparatus. In each of two conditions, sitting and standing, 30 simple reaction times were collected in three blocks of 10 trials, completed at the beginning, middle and end of the test session.

Simple and probe reaction time tasks were repeated if the response time fell outside the range of 100–1000 ms. To remove outliers, response times in each condition for each subject were ordered and extreme values were removed from either end. For the simple reaction time data the fastest four and slowest six values were removed, while for the probe reaction time data the fastest two and slowest three values were removed.

Perceived task load

The perceived effort of walking was assessed using a NASA-Task Load Index (Hart and Staveland, 1988), which was completed at the conclusion of all trials in each cueing condition. The NASA-Task Load Index is a weighted rating scale that assesses six aspects of task load, including the mental, physical and temporal demands, as well as effort, frustration and perceived performance.

Statistical analysis

Dependent variables were analysed using a group (patient, control) × cueing condition (baseline, SL markers, SMLD) repeated measures mixed ANOVA (analysis of variance) with planned comparisons between the three cueing conditions. The level of significance, using a Hunyh–Feldt correction factor, was set at 0.05. A family of paired t-tests were implemented to investigate interactions. The level of significance of the t-tests was adjusted using a Bonferroni correction. Results are reported as mean ± standard deviation.


Spatiotemporal data

Velocity, cadence and stride length

Gait velocity, cadence and stride length results are shown in Table 3 and Fig. 2. Interactions between group and cueing condition were evident for gait velocity [F(2,50) = 3.767, P = 0.029], cadence [F(2,52) = 9.389, P < 0.001] and stride length [F(2,50) = 13.899, P < 0.001]. In baseline conditions, average gait velocity was significantly slower in the Parkinson's disease patients than in the control group (paired t-test; n = 14, P = 0.005). This reduction in velocity was due to a shortened stride length in the patient group (P = 0.004). There was no difference in cadence between the two groups in baseline conditions (P = 0.8).

View this table:
Table 3

Means and standard deviations of control and patient velocity, cadence and SL data in the three walking conditions

BaselineSL markersSMLD
n 141414141314
Velocity (m/s)1.06 ± 0.211.39 ± 0.221.17 ± 0.181.36 ± 0.171.22 ± 0.191.41 ± 0.24
Cadence (steps/min)120 ± 11.0117 ± 8.0105 ± 14.1120 ± 11.7112 ± 11.7113 ± 12.9
Stride length (m)1.10 ± 0.251.42 ± 0.181.34 ± 0.091.36 ± 0.071.29 ± 0.201.49 ± 0.17
Fig. 2

Patient velocity, cadence and stride length means, with standard deviation bars, expressed as a percentage of baseline control values for three walking conditions. 100% is indicated by the dotted line.

With the addition of visual SL markers, patient stride length increased to 1.34 ± 0.09 m, which was not significantly different from baseline control stride length (P = 0.14). A reduction in cadence that approached significance was evident (P = 0.008), but overall gait velocity was increased from baseline conditions (P = 0.003). These two parameters were also not significantly different from baseline control results (cadence: P = 0.02; velocity: P = 0.03). Control group stride length (P = 0.08), cadence (P = 0.2) and gait velocity (P = 0.3) were not altered from baseline conditions.

Similar results were seen for the patients when using the SMLD. Stride length increased (n = 13, P < 0.001), cadence was similar to baseline (n = 14, P = 0.04) and overall gait velocity was increased (n = 13, P = 0.001). These values were not different from the corresponding baseline control results (velocity: n = 13, P = 0.1; cadence: n = 14, P = 0.2; stride length: n = 13, P = 0.2). Gait velocity in the patients was similar to that obtained using the SL markers (n = 13, P = 0.7); however, cadence was faster in the SMLD condition (n = 14, P = 0.006). Stride length was also slightly, but insignificantly, shorter (n = 13, P = 0.3). Control subjects exhibited an increase in stride length when using the SMLD that approached significance (n = 14, P = 0.05), although cadence (P = 0.4) and overall gait velocity (P = 0.7) were unchanged from baseline.

CV of velocity, cadence, stride length

CV results are shown in Table 4. There were no differences between patients and controls for either the CV of velocity or cadence [F(1,23) < 1]. The introduction of visual cues resulted in a decrease in CV of velocity from baseline across both groups [F(2,48) = 23.379, P < 0.001]. This was evident for both the SL marker (planned contrast: n = 28, P < 0.001) and SMLD conditions (n = 27, P = 0.02). CV of cadence was largely unchanged by the addition of visual cues, although there was a slight increase in CV of cadence above baseline when using the SMLD (planned contrast: n = 27, P = 0.02). In contrast, CV of stride length had main effects of both cueing condition [F(2,48) = 54.205, P < 0.001] and group [F(1,24) = 9.428, P = 0.005]. Control subjects had a significantly lower overall CV of stride length in comparison with the Parkinson's disease patients. The use of SL markers markedly reduced CV of stride length across both groups (planned contrast: n = 28, P < 0.001); however, it was unchanged when using the SMLD (n = 27, P = 0.2).

View this table:
Table 4

Means and standard deviations of gait velocity, cadence and SL CV

BaselineSL markersSMLD
n 141414141314
CV of velocity (%)7.12 ± 2.985.64 ± 3.053.53 ± 1.753.08 ± 1.825.20 ± 1.284.40 ± 2.08
CV of cadence (%)2.94 ± 1.792.73 ± 1.593.52 ± 2.082.69 ± 1.734.05 ± 1.853.18 ± 1.67
CV of SL (%)5.32 ± 2.553.42 ± 1.511.76 ± 1.660.77 ± 0.244.53 ± 2.042.59 ± 0.86

Kinematics and kinetics

As one of the aims of this study was to determine the biomechanical mechanisms behind the reductions in stride length in Parkinson's disease and behind the improvements that are seen with the use of visual cues, 3D gait analysis was selectively undertaken on four patients (P1, P3, P7, P9) who demonstrated a stride length in baseline conditions that was >15% shorter than their height-predicted stride length, and who significantly increased this stride length with the use of SL markers. In addition to this, sagittal plane kinematic analysis of the ankle, knee and hip joints was undertaken on all subjects from whom appropriate video data were obtained (11 patients, 11 controls). This subjective analysis focused on the ankle and hip joints, which are the main contributors to alterations in stride length (Judge et al., 1996), and on the baseline and SL marker conditions.

A striking feature of the kinematic and kinetic results was the large variability associated with the patients, making consistent trends difficult to detect. However, there were a number of characteristics that appeared to be common in the patient group, particularly for the sagittal plane profiles. In baseline conditions, the heel rocker at foot strike was often reduced or lacking in comparison with controls, indicating a foot flat ground contact, and ankle plantar flexion at toe-off was also often reduced (Fig. 3A). Commonly associated with these was a reduction in ankle power generation at the toe-off phase (Fig. 3B). Reductions in stance phase knee extension were prevalent, and in the more severely affected patients this was sometimes accompanied by a reduction in peak knee flexion during the swing phase (Fig. 3C). Most of the patients demonstrated a lack of hip extension during mid-stance in at least one of the limbs (Fig. 3D); reductions in peak hip flexion at heel strike were also seen but were less common. Reductions in pelvic rotation were evident in some patients and one patient appeared to demonstrate an out of phase pelvic rotation (Fig. 3E). Trunk flexion angle also varied considerably between individuals, but was substantially increased in the more severely affected patients (Fig. 3F).

Fig. 3

Examples of patient kinematic and kinetic profiles in baseline conditions. Solid line = subject mean; dotted line = subject standard deviation; shaded area = 1 SD of control subjects in baseline conditions; n = number of strides analysed. (A) Ankle flexion (Patient 3); (B) ankle power (Patient 7); (C) knee flexion (Patient 9); (D) hip flexion (Patient 9); (E) pelvic rotation (Patient 9); (F) trunk flexion (Patient 9).

In the SL marker condition, alterations in the kinematic and kinetic profiles of the patients towards that of the controls were evident, especially for the less severely affected patients. Specifically, these alterations included increases above baseline conditions in the heel rocker (Fig. 4A), ankle dorsiflexion during late stance (Fig. 4A), ankle plantar flexion and ankle power generation at toe-off (Fig. 4B), peak hip extension angle (Fig. 4C) and the magnitude of pelvic rotation (Fig. 4D). In the more severely affected patients, however, there were also some instances of abnormalities, or differences from the control subjects, in the 3D profiles in the SL marker condition. Principally, these were alterations in ankle moment (Fig. 5A) during mid-stance that were accompanied by an increase in the magnitude of trunk flexion movement (Fig. 5B), possibly indicating that these patients were employing upper body momentum to increase stride length to that required. This is an example of an energetically inefficient gait pattern. Similar alterations in the kinematic and kinetic profiles of the patients were evident in the SMLD condition, although generally to a lesser extent. Also lacking was the occurrence of the inefficient gait pattern evident in the SL marker condition.

Fig. 4

Examples of patient kinematic and kinetic profiles in the SL marker condition. Solid line = subject mean; dotted line = subject standard deviation; shaded area = 1 SD of control subjects in baseline conditions; n = number of strides analysed. (A) Ankle flexion (Patient 3); (B) ankle power (Patient 7); (C) hip flexion (Patient 9); (D) pelvic rotation (Patient 9).

Fig. 5

Examples of patient kinematic and kinetic profiles in the SL marker condition of Patient 1. Solid line = subject mean; dotted line = subject standard deviation; shaded area = 1 SD of control subjects in baseline conditions; n = number of strides analysed. (A) Ankle moment; (B) trunk flexion.


Percentage r.m.s. amplitude data were arcsine transformed before statistical analysis. Non-dominant limb results are demonstrated in Fig. 6. Interval 4 was the only interval statistically analysed for the soleus muscle as alterations in stride length involving plantarflexor activation mainly arise during terminal stance (approximately Interval 4). For tibialis anterior, Interval 1 was analysed in order to examine alterations in activation at heel strike.

Fig. 6

Means and standard deviations of percentage r.m.s. amplitude EMG for tibialis anterior (TA) and soleus muscles in the non-dominant leg. Stride cycles were divided into eight even intervals of time. Patient TA average: open circles = baseline; filled squares = SL markers; filled triangles = SMLD.

In the soleus muscle a main effect of group for the non-dominant leg [F(1,24) = 8.59, P = 0.007] revealed that the patient group (43.7 ± 28.5%) demonstrated a higher average percentage r.m.s. amplitude than the control group (38.2 ± 29.4%). For both groups it was found that the EMG level at Interval 4 was significantly higher when using the SMLD (88.4 ± 14.4%) in comparison with baseline (82.2 ± 15.8%) (planned contrast: n = 25, P < 0.001). Conversely, there were no differences between the SL marker (80.4 ± 14.7%) and baseline conditions for this interval (P = 0.2). Three-way interactions between group, condition and interval were also present for the soleus muscle in both the non-dominant [F(14,336) = 2.853, P = 0.004] and dominant limbs [F(14,336) = 2.138, P = 0.04], although the results of the paired t-tests implemented did not reach conventional levels of significance at the interval of interest (P > 0.05).

In the tibialis anterior muscle a main effect of cueing condition was present in the non-dominant leg [F(2,44) = 3.976, P = 0.03]. Planned contrasts revealed greater EMG activity in the SMLD condition (50.0 ± 28.8%) in comparison with baseline (48.6 ± 28.8%) (planned contrast: n = 26, P = 0.03). A condition by interval interaction [F(14,308) = 2.634, P = 0.02] revealed a specific increase in EMG activity from baseline (75.1 ± 7.0%) at Interval 1 when using the SMLD (78.3 ± 18.7%) (paired t-test: n = 26, P < 0.001). In the dominant limb there were no significant results evident for the tibialis anterior muscle (all P > 0.05, corrected α = 0.008).

Reaction time

Analysis of median reaction time was carried out after the removal of outliers. Results are shown in Table 5. It was revealed that the control subjects were significantly faster than the Parkinson's disease patient group [F(1,26) = 17.648, P < 0.001]. Simple reaction time was faster than probe reaction time across both groups (planned contrast: n = 28, P < 0.001). There were no differences between sitting and standing reaction time (P = 0.09), although there were differences evident between the three probe reaction time conditions. Reaction time when walking with the SL markers was not different from baseline walking conditions (P = 0.7), but it was significantly increased above baseline when walking with the SMLD (P = 0.01).

View this table:
Table 5

Median reaction times (ms) and standard deviations for patients and controls in the five conditions

SittingStandingBaselineSL markersSMLD
n 1414141414
Patient390 ± 65.6364 ± 70.2456 ± 89.9464 ± 104.3489 ± 91.6
Control285 ± 41.6270 ± 38.4353 ± 84.2340 ± 56.8388 ± 97.0

Perceived task load

Ratings of perceived task load are shown in Table 6. These data were arcsine transformed before undergoing statistical analysis. A main effect of subject revealed that the Parkinson's disease patients had a higher task load rating than the controls in all three walking conditions [F(1,26) = 10.848, P = 0.003]. Cueing condition also demonstrated a main effect for NASA-Task Load Index ratings [F(2,52) = 11.506, P < 0.001]. The addition of visual SL markers resulted in no significant change in perceived task load from baseline (planned contrast: n = 28, P = 0.4); however, across both groups there was a significant increase in perceived task load from baseline when using the SMLD (P < 0.001).

View this table:
Table 6

Means and standard deviations for NASA-Task Load Index ratings

BaselineSL markersSMLD
n 141414
Patient28.2 ± 15.132.8 ± 15.345.3 ± 25.6
Control14.7 ± 11.614.6 ± 13.722.0 ± 14.4


Spatiotemporal data

The characteristic hypokinetic gait pattern evident in the patient group under baseline conditions supports previous findings on parkinsonian gait (Knuttson, 1972; Peterson et al., 1972; Murray et al., 1978; Bagley et al., 1991; Morris et al., 1994b; Thaut et al., 1996; Pedersen et al., 1997). Compared with similar aged control subjects, the patients demonstrated, on average, a 24% reduction in gait velocity and a 23% reduction in stride length, while cadence was relatively comparable. These features are a hallmark of a gait pattern consisting of short, shuffling steps taken at a high frequency.

The improvement in stride length and velocity that occurred with the use of visual SL markers also supports findings from similar studies (Bagley et al., 1991; Morris et al., 1994b, 1996). Clearly, Parkinson's disease patients can enlarge stride length if appropriate cues are provided and can walk with a comparable gait pattern to control subjects. The finding that Parkinson's disease patients also exhibited a larger CV of stride length than the control subjects follows previous research (Blin et al., 1990) and highlights the difficulties that patients have specifically in the internal regulation of stride length. Caution should be taken, however, in the interpretation of the CV as differences in the means of the two groups will affect the results. While taking this into consideration, stride length variability was still markedly reduced with the introduction of SL markers, again providing evidence that with an external specification of required force amplitude and with visual feedback of responses, the regulation of movement amplitude can be enhanced. Possible reasons for deficiencies in the internal regulation of stride length in Parkinson's disease patients include: a disruption in the internal cueing of sequential movements (Morris et al., 1994b); an increased variability of force production (Blin et al., 1990; Hausdorff et al., 1998); alterations in postural reflexes (Blin et al., 1990; Hausdorff et al., 1998); or disturbances in the interpretation of, or responses to, sensory information and feedback.

Similar changes in the spatiotemporal gait pattern of the patients occurred when using the SMLD, suggesting that this may be an effective tool for improving the regulation of stride length for this population. It was apparent that responses to the SMLD varied markedly between individuals. For those with small to moderate reductions in stride length, the SMLD was effective in increasing stride length to, and even above, their height-predicted estimate. However, the SMLD was less successful for those with severe reductions in baseline stride length. This observation may indicate the optimal stage of the disease when cueing interventions such as the SMLD could usefully be implemented. The more severely affected patients were unable to increase stride length to that specified by the cue, suggesting that it may be difficult for Parkinson's disease patients to maximize the benefits of novel visual cues in the later stages of the disease. It is proposed that patients who learn to utilize cueing interventions such as the SMLD in the early stages of disease progression might be able to implement these cues more successfully in the later stages, when they will be of most benefit. It is possible that the regulation of gait via alternative neural pathways, which possibly occurs when utilizing visual cues, may be facilitated if these alternative pathways have previously been utilized.

Similar studies on Parkinson's disease gait have examined patients while ON medication (e.g. Bagley et al., 1991; Morris et al., 1994b, 1996; Thaut et al., 1996; Behrman et al., 1998). It would be of interest to examine the stride length–cadence–gait velocity relationships of the patients in normal and augmented conditions while in the OFF phase of medication. Gordon and Reilmann have reported that l-dopa causes measurable impairments in grasping force in Parkinson's disease patients (Gordon and Reilmann, 1999). They have suggested that the analysis of patients OFF medication is likely to provide more information about motor deficits that are caused by the pathophysiology of the disease itself, without the complications of medication effects. It is possible that the effects of medication on force generation may be obscuring the natural gait parameter relationships, making it difficult to interpret and relate findings of studies utilizing patients ON medication to neurological mechanisms.

Biomechanical considerations

Perhaps the most striking feature of the 3D kinematic and kinetic profiles was the variability of results within the Parkinson's disease group, likely due to the large range of disease severity in the patients analysed. Reductions in peak joint angles in the sagittal plane profiles are consistent with previous studies employing 2D (Knuttson, 1972; Murray et al., 1978; Forssberg et al., 1984) and 3D (Morris et al., 1999) analysis techniques of Parkinson's disease patients. The absence or reduction of the heel rocker at foot strike is characteristic of a foot-flat ground contact and is perhaps the most significant contributor to its description as a `shuffling' gait pattern. Reductions in ankle plantar flexion at toe-off, increased stance phase knee flexion and reductions in hip extension will all also contribute to a shortened stride length. The reduced ankle power generation at toe-off evident in this study was also found in a single-subject study on a Parkinson's disease patient by Morris and colleagues and seems the most likely contributor to the reduced stride length (Morris et al., 1999). The importance of this deficit has been illustrated by Judge and colleagues, who found that ankle plantarflexor power was the strongest predictor of stride length in elderly subjects (Judge et al., 1996). Reductions in power may arise from a diminished activation or weakness of the ankle plantarflexors, co-contraction of the ankle dorsiflexors or increased muscle stiffness (Kerrigan et al., 1998). No conclusive evidence of co-contraction of the antagonist muscles about the ankle joint was seen in the EMG data, suggesting that this is not likely to have contributed to reductions in ankle power.

If increased muscle stiffness, which is a symptom of Parkinson's disease (Dietz et al., 1981), does contribute to the reduced ankle power generation then it is possible that this factor may influence both cadence and stride length parameters in the Parkinson's disease population. Muscle stiffness influences the natural oscillatory properties of the swing limb, with increased stiffness increasing the resonant swing, or step, frequency (Holt et al., 1996). Taking this into consideration, it is possible that the altered spatiotemporal gait pattern adopted by Parkinson's disease patients is the most efficient given the altered mechanical properties of their limbs. Although measurements of muscle stiffness were not made in the current study, this suggestion seems to be supported by the less energetically efficient gait pattern seen in the more severely affected patients when asked to walk with an increased stride length, and raises the issue of whether or not therapeutic interventions should be attempted to normalize the kinematic profiles for individuals with Parkinson's disease or other movement disorders. Therefore, treatments such as physiotherapy, which are able to reduce muscle stiffness, may be beneficial to the gait pattern of Parkinson's disease patients.

The most interesting features of the EMG results were the lack of differences in the patient group between the different cueing conditions, and the similarity of EMG modulation over the phases of the gait cycle between patient and control groups. In the patients there was no direct evidence found of reduced ankle plantarflexor activation leading to the reductions in stride length evident in baseline conditions. It is possible that the lack of significant results here may have been, again, due to the large range of disease severity of the patients. However, there was some evidence of alterations in EMG amplitude in both muscles of the non-dominant limb when using the SMLD. As patient stride length increased in this condition and an increase in control stride length which approached significance was also observed, it is likely that these increases in stride length were achieved, in part, through modulations in the amplitude of EMG activity. These findings suggest that Parkinson's disease patients maintain the ability to modulate and increase muscle activation if appropriate cues are supplied.

It is also evident that the modulation of EMG activity over the gait cycle in the distal muscles of the lower limb is maintained in Parkinson's disease patients. This finding suggests that spinal locomotor circuits are intact in Parkinson's disease and that deficiencies arise in the intensity of the input to these circuits. Similar alterations in triceps surae and tibialis anterior muscle activity have been found previously when patients have moved from the OFF to the ON phase of medication (Cioni et al., 1997). It is possible that the addition of l-dopa may enhance the intensity of the input to the spinal circuits, resulting in increases in EMG activity at critical phases, and that the use of appropriate visual cues may enhance this input further.

Neurological mechanisms

The present results show that cadence variability in Parkinson's disease is comparable with controls. Coupled with previous research that has shown that the modulation of cadence remains intact in Parkinson's disease patients (Morris et al., 1994a), these findings refute the hypothesis that gait disorders are due to difficulties in the internal cueing of gait submovements. Deficits with internal cueing would have been manifest as irregularities in cadence regulation, which were not seen. Instead, the spatiotemporal data support the contention that gait deficiencies in Parkinson's disease arise from a disturbance in the motor set function of the basal ganglia specifically relating to the regulation of movement amplitude.

The demonstration of a slower simple reaction time in Parkinson's disease patients compared with age-matched control subjects supports previous research (e.g. Yanagisawa et al., 1989; Berry et al., 1999; Kutukcu et al., 1999). Of more interest in the present study was how reaction time changed across the different walking conditions, particularly the reaction times of the Parkinson's disease group. The finding that reaction time was unchanged from baseline to SL marker conditions in the patients tends to contradict the idea that visual cues may invoke higher cortical structures that bypass the faulty basal ganglia–SMA system. These may be accessible through conscious/voluntary control via visual pathways. However, if this were the case one might expect an associated cost with respect to residual processing capacity, i.e. an increase in probe reaction time. As an alternative explanation, it is suggested that with visual cues patients are able to (i) more easily specify an adequate stride length, and (ii) utilize intact visual feedback to regulate the success of the movement, reducing the reliance on kinaesthetic feedback. A previous study utilizing an upper limb positioning task has suggested that impairments in kinaesthetic feedback may exist with Parkinson's disease (Demirci et al., 1997).

In the SMLD condition, reaction time increased above baseline. The increase in reaction time was accompanied by an increase in the perceived effort of the task and an increase in EMG activity in both of the muscles analysed in the non-dominant leg. Since these changes occurred in both groups, it is evident that the increased attention required by the patients in this condition is not above that required by the control group. This suggests that the changes reflect the increased demand of the cue rather than a specific increase in attentional requirement of the patients. In contrast, SL markers appear to be an effective visual cue for improving the spatiotemporal gait pattern of most Parkinson's disease patients that do not compromise the attentional demands of the patient and do not result in significant increases in the effort of walking. To provide more conclusive evidence that residual processing capacity was not altered between baseline and SL marker conditions it would be necessary to manipulate the difficulty of the secondary task and compare the effects on reaction time between the two conditions.

The question still remains as to whether, in baseline conditions, patients specify a normal stride length but fail to generate the required amplitude, or they specify a shortened stride length and regulate their movements around a smaller scale. Demirci and colleagues suggested that comparable reductions in kinaesthetic feedback and corollary discharge, or efference copy, were responsible for the diminished movement amplitude seen in Parkinson's disease patients during a limb positioning task (Demirci et al., 1997). They proposed a model of hypometria (Fig. 7) in which both the motor command and efference copy of a desired movement were underscaled. In this model, the reduced kinaesthetic feedback arising from the reduced movement amplitude matches the underscaled efference copy, resulting in the regulation of movement amplitude around a lower range. With the addition of visual cues, mismatches between kinaesthetic and visual feedback of the desired movement would be evident, indicating a failure in the production of the desired movement. This mismatch would then result in an increase in the motor command and efference copy in order to rescale movement amplitude accordingly. It was suggested that `efference copy probably gives the feeling of motor effort' and so the resultant increase in expected movement from kinaesthetic feedback would also be accompanied by an increase in the perceived effort of the movement. The spatiotemporal and perceived exertion data associated with the SMLD condition in the current study appear to support the model proposed by Demirci and colleagues (Demirci et al., 1997). The apparent anomaly in this study are the results obtained for the patient group in the SL marker condition, which do not support the model outlined above. Despite the fact that patients increased stride length and gait velocity by around 30% with the use of SL markers, their perception of effort was not increased above baseline levels. The lack of change in effort may indicate that the patients perceived that they were generating an adequate stride length in baseline conditions, i.e. the efference copy was intact, even though the actual SL was somewhat reduced. This would indicate a discrepancy between the motor command and the expected movement, possibly arising from either: a deficiency in kinaesthetic feedback resulting in an exaggeration of perceived movement amplitude, or an inability to respond appropriately to the discrepancies between the intended movement specified by the efference copy and the actual movement generated as indicated by kinaesthetic feedback.

Fig. 7

Model of hypometria in Parkinson's disease. Magnitude of motor commands and sensory information is indicated by ★ symbols. MC = motor centre; C1, C2 = comparator centres. Adapted from Demirci and colleagues (Demirci et al., 1997).

A possible explanation accounting for the differences in reaction time and perceived exertion results evident between the two visual cues in this study may arise from differences in the degree of optic flow, or the dynamic flow of information across the visual field (Azulay et al., 1999). Since the lines projected by the SMLD moved along with the subject they would appear as static cues to the subject and would not increase optic flow. In contrast, walking over the SL markers would increase dynamic visual information. From the study by Azulay and colleagues utilizing stroboscopic lighting to suppress dynamic visual cues, it was concluded that the `perceived motion of stripes, induced by the patient's walking, is essential to improve the gait parameters' (Azulay et al., 1999). As increases in SL were evident in the SMLD condition in the present study, it is obvious that static visual cues can enhance Parkinson's disease gait. However, the increased attentional demand, as assessed by probe reaction time, and perceived effort of walking when using the SMLD may reflect that this condition did not benefit from visuomotor facilitation, whereas this facilitation was present and may have aided gait performance in the SL marker condition. It is also possible that this flow-induced visual information obtained when walking over a striped runway may be more useful in the regulation of gait velocity and that the improvements evident in SL may partly be arising secondary to this. However, the results concerning the regulation of gait velocity in this study did not discriminate between the static and dynamic cue types.

In conclusion, both static and dynamic visual cues are able to improve the hypokinetic spatiotemporal gait pattern of individuals with Parkinson's disease. From the results of this study, it is suggested that improvements in the Parkinson's disease gait pattern with the use of visual cues may arise through the patients' ability to utilize visual feedback to regulate movement amplitude, reducing their reliance on kinaesthetic feedback.


The authors wish to thank Professor Meg Morris, Professor Richard Faull and two anonymous reviewers for insightful comments on previous versions of this manuscript. The authors also wish to thank Lorraine Macdonald, RN, of the Movement Disorders Clinic, Auckland Hospital for the referral of patients. Funding was provided by a grant from the Auckland University Staff Research Fund.


View Abstract