initiation and execution of predictable and ...

Brain (1984), 107, 371-384

INITIATION AND EXECUTION OF PREDICTABLE AND UNPREDICTABLE MOVEMENTS IN PARKINSON'S DISEASE by C. A. BLOXHAM, T. A. MINDEL and C. D. FRITH {From the Division of Psychiatry, Clinical Research Centre, Watford Road, Harrow, Middlesex HA1 3UJ) SUMMARY

INTRODUCTION

The abnormalities of movement that characterize Parkinson's disease are well known and widely believed to be the consequences of damage to the nigrostriatal pathway and subsequent dopamine deficiency in the striatum. Each of the cardinal signs of this illness has received increasing attention and several attempts have been made to discover their possible pathophysiological basis (Marsden, 1982). The most disabling feature, however, is undoubtedly hypokinesia, held by many to be the primary symptom of basal ganglia dysfunction (Denny-Brown, 1968). Earlier studies were directed towards elucidating the different facets of this symptom such as initiation difficulties (Draper and Johns, 1964), slowness of movement (Joubert and Barbeau, 1969), and problems in stopping or changing motor patterns (Joubert and Barbeau, 1969, Angel et ai, 1970). More recently Hallett and Khoshbin (1980), emphasizing the difficulties parkinsonian patients have in performing ballistic movements described, on the basis of EMG studies, a possible 'derangement in energising muscles' that might represent a physiological mechanism underlying bradykinesia. Address for correspondence andreprints:DrC. D. Frith, CRC Psychiatry, Watford Road, Harrow HA 1 3UJ, UK.

Downloaded from by guest on August 7, 2015

Nine patients with Parkinson's disease were compared with aged-matched normal controls on a tracking and a reaction time task. In the tracking task, subjects had to follow a target moving either predictably or unpredictably at various frequencies. There were no differences between the groups on this task. Both groups were able to use the predictable target movements to eliminate the lag between the target movements and their own movements, indicating that they could emit preprogrammed responses which did not depend on visual feedback. The reaction time task also contrasted predictable and unpredictable movements since the subjects were sometimes told in advance which finger they would have to lift in response to the imperative signal. The patients were significantly less able than controls to use this predictability to reduce their reaction times. It is concluded that parkinsonian patients can use prior information to control the form of a preprogrammed movement, but cannot use this information to help initiate such a movement.

372

C. A. BLOXHAM, T. A. MINDEL AND C. D. FRITH

Downloaded from by guest on August 7, 2015

Other workers have interpreted the clinical features in terms of disturbed central control mechanisms, reflected by studies showing that parkinsonian subjects have difficulty in performing two actions simultaneously (Schwab et al, 1954). Indeed, from clinical observations of parkinsonian patients, Marsden (1982) concluded that the normal role of the basal ganglia is the 'automatic execution of learned motor plans'. In the same vein, Flowers (19786) reported that patients demonstrate a failure to employ the normal strategies of prediction during visual tracking tasks and cannot therefore utilize an 'internal model' of external events in carrying out motor activities. He suggested that this results in the loss of'skills' which leads to a greater dependency upon sensory, especially visual, information. A relevant concept is that of motor programs (Keele and Posner, 1968) which account for the improvement in performance of any movement in which an 'open loop' mode of control can be used, such as the tracking of a regularly moving target. In this situation the subject can plan and execute his movement in advance and in anticipation of the target's movement. In contrast, only feedback control, or closed loop strategies are possible when the target movement is irregular. Here the subject can only plan and execute his movement on the basis of how the target has just moved. The implications of Flowers' work is therefore that patients with Parkinson's disease are unable to make use of such programs during motor activity. However, clinical features such as kinesia paradoxica suggest that the specific difficulty may be in acquiring access to motor programs that in themselves are relatively intact, since in certain circumstances severely impaired patients can execute complex movements in a normal manner. We have attempted to demonstrate that the performance of parkinsonian subjects on two different tasks is consistent with the idea that they can execute preprogrammed movements, but have difficulty in initiating them without the help of an external trigger. During the first experiment predictable and unpredictable movements are compared in a pursuit task. To track a moving target using feedback the subject must first observe where the target has gone and then make the appropriate movement to catch up with it. As a consequence of his relatively long reaction time there will always be a detectable delay between an unexpected movement of the target and the corresponding corrective movement of the subject. If, however, the path of the target is predictable then the subject can initiate his movements in advance and in anticipation of the target. Young (1962) developed a simple method for generating predictable and unpredictable target movements and for assessing the lag in the subject's response. He demonstrated that for an unpredictable target there was an increasing lag in the eye movement response as the target velocity increased. For predictable targets, on the other hand, there was no increase in lag with increasing target velocity. Control of a movement in anticipation of the target cannot be achieved by feedback; it must be achieved by preprogramming the response. Thus if patients with Parkinson's disease can only control their

MOVEMENT IN PARKINSON'S DISEASE

373

movements by feedback they should not be able to benefit from predictable target movements by anticipating them. The second experiment involves the measurement of reaction times with and without prior knowledge of the movement required and concerns selection and initiation, rather than execution of a response; we would anticipate that if patients are unable to use prior information to select a motor program before execution of the task, they should not in turn be able to reduce their reaction times when the movement required is known in advance. EXPERIMENT 1 Method

Control of target movement. Movement of the target in the horizontal and vertical directions was controlled independently. For each of these directions of target movement the position of the target every 25 ms was determined from a combination of sine waves. For predictable tracks the vertical and horizontal components were single sine waves of the same frequency, but 90 deg out of phase so that the target moved in a circle in a clockwise direction. Frequencies ranged from 0.1 Hz to 1.2 Hz (see Table 1). For the unpredictable tracks three sine waves were combined for each direction of movement. These three sine waves were of unequal and unrelated frequencies (see Table 1). The irregular pattern of target movement that resulted was essentially unpredictable since it did not repeat for 5 min. The highest frequencies for the four unpredictable tracks were identical to the single frequency of the corresponding predictable track. Thus the lag shown by the subject for the same frequency in predictable and unpredictable tracks could be directly compared. Control of target movement was implemented by a PDP-11 computer. Recording of tracking response. Two voltages indicating the horizontal and vertical positions of the joystick were fed into the computer via A/D converters and thus controlled the position of the cross. Position of target and subject cross were recorded every 25 ms.

TABLE 1. TRACK COMPONENTS Predictable

Frequency 1 2 (Hz) 3 4

Unpredictable

Horizontal

Vertical

0.20 0.40 0.80 1.20

0.20 0.40 0.80 1.20

Horizontal 0.20 0.40 0.80 1.20

0.11 0.21 0.42 0.63

Vertical 0.04 0.09 0.18 0.27

0.20 0.40 0.80 1.20

0.09 0.19 0.38 0.57

0.06 0.11 0.22 0.33

Downloaded from by guest on August 7, 2015

Subjects had to track a target which moved in two dimensions on the face of a VDU with a maximum amplitude of 10 cm. The target was a 4 mm square and the subject-manipulated marker was a cross which fitted exactly inside the square. The subject's task was to match the movement of the target as closely as he could, keeping the cross, if possible, inside the square. The position of the cross was controlled by means of a 17 mm long joystick which the subject held in his preferred hand. The subject's elbow rested on a foam rubber pad and thus control of the joystick required movement of the wrist and forearm.

374

C. A. BLOXHAM, T. A. MINDEL AND C. D. FRITH

Procedure Subjects werefirstpresented with a very slow predictable track (0.1 Hz) which was used as a trial run. Thereafter each predictable track was followed by an unpredictable one with matching highest frequency. The tracks were presented in order from the slowest to the fastest. A single run with each track lasted one minute and the subject had a 3 min rest between each run. Subjects were told in advance whether a track would be predictable or not and were also told that the predictable track would be in a circle.

Subjects. Patients with a severe impairment of movement control due to Parkinson's disease would obviously find this tracking task extremely difficult to perform. As a consequence they would perform any version of this task, whether predictable or unpredictable, worse than age-matched controls. However, the demonstration of a global quantitative impairment in Parkinson's disease would be of little interest. Information about the precise nature of the problem with movement control can only come from the demonstration of qualitative differences between patients and controls. Our prediction from Flowers' experiments is that patients with Parkinson's disease are unable to execute the preprogrammed movements which are possible when tracking a target which moves in a predictable manner. Given this hypothesis the ideal selection strategy would be to choose a group of patients and controls who were matched on their performance when tracking an unpredictable target. If the hypothesis was correct, then, in spite of this close matching, the patients should be impairedrelativeto controls when tracking a predictable target. For this reason we deliberately excluded from our patient group those more severely impaired people who had considerable difficulty with the tracking task (i.e. less than 5 per cent time on target).

TABLE 2. CLINICAL ASSESSMENT OF DISABILITY IN THE PARKINSONIAN GROUP Tremor Nil L arm Rmiid

Rigidity Both arms Both arms

Brodykbtexia Moderate Slight

8 yrs

Nil

Moderate

Independent

NUat rest Nil Larm + Lleg

Slight, both arms Larm

Nn

Independent

Sinemet ( + propranotol) Smcroct

Larm L arm R^mild

Larm Slight

Independent Independent

Sinemet Nil

Shjht

Slight, berth

MUd

Independent

Sinemet

Subject 1 2

Sex M F

0") 56 70

Diraxioa (ttmefrom dtagnosii) 2yn Myn

3

M

50

Social (Stability Independent Independent

Drugs Sinemet Sine met Kemadrin

4

F

43

1J yn

5 6

M M

43 44

2yrs

7

F

74

2 mo.

8

F

69

4yrs

Nil

Larm

Nil

Independent

Sinemet ( + propranotol)

9

F

68

2yn

Nil

Both arms

Slight

Independent

NH

2yn

Downloaded from by guest on August 7, 2015

Analysis of performance. Thefirst30 s of each 1 min run were discarded, since the frequency analysis required that the sequence of samples should be in arelatively'steady' state. The following 512 samples (12.8 s) of both subject and target positions were subjected to a Fourier transform for the frequencies used to generate the target movement. This was done for the horizontal and vertical components separately. Frith (1973) had found that only those frequencies present in the target were present to any extent in the responses produced by subjects when performing this task. The phase lag of the subject was estimated from the differences between the phase of the target and the phase of the subject. Phase lag was measured in degrees such that 90 indicates a lag of \ cycle and 180 indicates a lag of J a cycle. As a consequence of relating lag to frequency the same delay in terms of time will appear as small lag at a low frequency and a large lag at a high frequency. The amplitude ratio (AR) was also measured. This gives a value of 1.0 when the amplitude of the movements produced by the subject is equal to that of the target. If the subject's amplitude is smaller than that of the target then the AR is less than 1.0. Groups and conditions were compared using analysis of variance with repeated measures (predictability and frequency).

MOVEMENT IN PARKINSON'S DISEASE

375

The parkinsonian group comprised 9 subjects, 4 male and 5 female. They were all outpatients with a history of the disease lasting at least eighteen months. Details of symptoms at the time of testing and the treatment regime are shown in Table 2. They were aged between 43 and 70 years with a mean of 57.4. All had been contacted through the Parkinson's Disease Society and had volunteered to take part in the experiment. All these volunteers were right-handed and they used this hand to perform the tracking task. The control group consisted of 11 volunteers from the League of Friends at Northwick Park Hospital. There were 6 females and 5 males aged between 43 and 74 years with a mean of 60.8. All but one female wereright-handed.All controls used their preferred hand to perform the tracking task. As a result of failure of apparatus, data from 1 subject in the Parkinson group and 2 subjects in the control group were lost. Informed consent was obtained from every subject.

TABLE 3. MEANS AND STANDARD ERRORS FOR AMPLITUDE RATIO AND PHASE LAG IN PARKINSONIAN PATIENTS AND CONTROLS

Unpredictable

Frequency (Hz) Patients Controls Patients Controls

0.2

0.4

0.8

1.2

0.93 ±0.04 0.94 ±0.06

0.95 ±0.04 0.85 ±0.03 21 ±4 18±4

0.74 ±0.08 0.80 ±0.07 51 ±9 41±6

0.64+0.09 0.59 ±0.09 75±7 91 ±9

0.79 ±0.08 0.66±0.07 10±3 6±5

0.57 ±0.08 0.53 ±0.07 26 ±14 51 ± 17

15±4 9±3

Predictable

Amplitude ratio Phase lag (deg)

Patients Controls Patients Controls

0.94±0.04 0.85 ±0.03

7±2 1±2

0.89 ±0.05 0.74 ±0.05 6±5 7±3

Results Table 3 shows the mean phase lags and amplitude ratios for the various conditions. For the raw phase lag data there was a strong relationship between mean and standard deviation with high means being associated with high standard deviations {see Table 3). This relationship was eliminated by taking the square root of the absolute phase lag, i.e. sign (square root (absolute phase)). By this means the data were rendered suitable for analysis of variance. (In fact exactly the same results were obtained by performing an ANOVA on the raw data.) There was a significant interaction between predictability and frequency (F(3,45) = 4.78, P (GreenhouseGeisser correction for repeated measures) < 0.01). This is illustrated in the figure. The lag is greater for the unpredictable tracks at all frequencies, but this difference is much greater at the high frequencies. Lag on the unpredictable tracks increases steadily with increasing frequency. Lag on the predictable tracks hovers around zero

Downloaded from by guest on August 7, 2015

Amplitude ratio Phase lag (deg)

376

C. A. BLOXHAM, T. A. MINDEL AND C. D. FRITH 90807060-

3 0

FIG. Mean phase lags for parlonsonian patients (8) and control subjects (9) combined. The y-axis has the phase lag of the subjects behind the target in deg, the x-axis the frequency of movement of the target in cycles/s. The closed circles are the phase lags for unpredictable targets, the open circles the phase lags for predictable targets. The solid lines are the bestfitcurves for the three lowest target frequencies assuming a constant lag in ms; this was 130 ms for the unpredictable targets and 0 ms for the predictable targets.

"

20100-100.2

0.4 0.8 Target frequency (Hz)

1.2

EXPERIMENT 2 Method Subjects responded to the command 'GO LEFT' or 'GO RIGHT' presented on a VDU by lifting their left or right indexfinger.Either 250 ms or 2000 ms prior to this command 1 of 3 possible warning signals appeared: 'READY', 'READY LEFT or 'READY RIGHT. With the warning signal 'READY' the subject did not know in advance which movement he would have to make, which is analogous to the unpredictable track in Experiment 1. With the warning signals 'READY LEFT' or 'READY RIGHT the subject knew in advance which movement he would have to make which is analogous to the predictable track. The subject's index fingers rested on two touch-sensitive pads. A relay was closed as soon as either finger ceased to be in contact with a pad and this closure was detected by a PDP-11 computer which also controlled the presentation of the warning and imperative signals. Procedure

The 8 conditions (warning x hand x delay) were presented in a balanced design so that in each block of 24 trials each condition appeared 3 times in a random order. Each subject first performed 1 block of trials for practice and then, when the task requirements were clear to him, performed 3 more blocks of trials the data from these being retained. The geometric mean of the reaction time for each condition was calculated. Wrong response times were not included in this mean and neither were excessively short responses (< 100 ms) or excessively longresponses(> 1.5 x current mean).

Downloaded from by guest on August 7, 2015

up to one cycle per second and then shows a sudden increase. There was absolutely no difference between the groups in the effect of predictability on lag (F(3,45) = 0.96), or any other effect. Essentially the same results were found for amplitude ratios. There was a decrease in the amplitude of movement as the target frequency increased {see Table 3), but there were no differences between parkinsonian patients and controls on this measure.

MOVEMENT IN PARKINSON'S DISEASE

377

Repetition effect. It is well known that if a subject has to respond to the same condition twice in succession then the second response is faster than the first (Bcrtelson, 1961; Rabbitt and Vyas, 1973). It was possible to look at this effect in our experiment since two conditions occurred in immediate succession on roughly 12 occasions for each subject. The geometric mean of the reaction time to the first and second of these repeats was calculated. It is clearly of interest to know whether this kind of response facilitation is affected in Parkinson's disease.

TABLE 4. THE EFFECT OF PRIOR INFORMATION ON REACTION TIME IN PATIENTS WITH PARKINSONS DISEASE AND CONTROLS

250 ms delay No information Mean Controls Patients

609 ± 623 ±

SEM 21 38

2000 ms delay

Information Mean 518 ± 600 ±

SEM 28 42

No information Mean

SEM

610 ± 2 4 611 + 40

Information Mean 447 ± 565 ±

SEM 28 56

Downloaded from by guest on August 7, 2015

Results Errors. The parkinsonian subjects made more errors than the controls (MannWhitney U-test, z=1.76, P