The literature on motor learning in patients with Parkinson's disease and the potential remedial effects of medication with L-dopa has yet to converge on an agreed message. Rather, many carefully controlled studies have produced a range of results beset with inconsistencies. Although, given the complexity of the neural networks implicated in generating adaptive movement, perhaps we should not be so surprised. Some of the important factors that interact to influence the variable outcomes of relevant experiments have already been identified.[1-3]

First, the concept of what constitutes, and the terms used to describe adaptive motor control can vary in different literatures. An important and clarifying distinction is between the acquisition of a motor skill (improvements in speed and accuracy), and motor adaptation (shifts in performance favoring speed or accuracy, or some form of sensorimotor recalibration).[4, 5] However, misunderstandings have arisen by researchers in different disciplines using different words for the same thing, and the same words for different things.[6] For example, the concepts of conscious, voluntary, explicit, controlled, goal-directed, and model-based motor learning overlap, but may not be identical. Similarly, unconscious, involuntary, implicit, automatic, habitual and model-free motor learning may represent similar phenomenon at different levels of description. But at our present stage of understanding, it may be unwise to consider them as synonymous. It is possible, therefore, that some of the seemingly paradoxical findings in the motor learning literature arise through a lack of an agreed nomenclature and taxonomy.

Secondly, many tasks of varying complexity have been used to investigate different aspects of motor learning in patients with Parkinson's disease. However, for understandably practical reasons (issues associated with testing elderly patients), most studies have confined themselves to investigations of the rapid, early stages of learning.[3, 5] The levels of practice required for automatized, habitual control to become fully established have usually been avoided. Consequently, the current motor skill literature with patients is likely to be biased towards their refinement of goal-directed movements.[6] Moreover, some more complex tasks could require variable mixtures of goal-directed and previously acquired habitual control. This is important because regionally segregated functional territories of the basal ganglia are differentially engaged during goal-directed and habitual control of behavior.[6, 7]

Thirdly, Parkinson's disease is a progressive disorder, which in itself has important implications for investigations of motor learning.[2] Because many of the tests measure speed of responding, the difficulties patients experience in initiating movement and bradykinesia at different stages of the disease could account for some of the variable results.[8] A further source of variability may be the progressive and sequential loss of dopamine from the regionally segregated functional territories of the basal ganglia.[9] Typically, dopaminergic innervation is lost first from the sensorimotor territories of the caudal putamen, which over time, extends forward and ventrally to encompass the associative and limbic territories. Insofar as habits are associated with the sensorimotor basal ganglia, and goal-directed actions with the associative territories,[7] patients tested at different stages of Parkinson's disease will vary in their ability to acquire, perform and refine goal-directed actions and habits.[6] Finally, when one part of the brain fails, other parts frequently adapt.[10] Tests conducted at different stages of the disease may therefore face a variable interplay between degenerative and adaptive mechanisms.[2]

Fourthly, a critically important factor is whether patients are tested ON or OFF dopamine-restoring medication, usually L-dopa and/or direct acting dopaminergic agonists.[11] Typically, patients receive sufficient medication to relieve their motor systems. In practice, this means a level of medication sufficient to restore dopaminergic transmission in the dopamine deprived sensorimotor basal ganglia. However, in patients where dopaminergic transmission in the associative and limbic territories is preserved, their medication is likely to produce supranormal levels of dopaminergic transmission.[12] This effect may be responsible for the impulsive behavioral control disorders experienced by some medicated patients.[13] Thus, for cognitive function it has been shown that too much or too little dopaminergic neurotransmission is detrimental to performance.[12] Assuming the same is true for motor control, the levels of medication will interact with the regional levels of dopamine depletion. Thus, depending on baseline levels of transmission, goal-directed learning in the limbic and associative basal ganglia may be enhanced, disrupted or unaffected by dopaminergic medication.[13] Alternatively, restoration of dopamine transmission to the sensorimotor territories is likely to have mainly beneficial effects on the acquisition and control of stimulus-response habits. Thus, to the extent that a particular task relies on goal-directed and/or habitual control dopaminergic medication is likely to have variable effects.[6]

Fifthly, Parkinson's disease is associated with the progressive loss of dopaminergic innervation, which means both tonic and phasic dopamine transmission will be impaired in affected regions.[14] The reinforcing effect of dopamine in motor learning has been linked to changes in the phasic release of dopamine evoked by the unpredicted occurrence, or absence, of biologically salient events.[15] Clearly, any rescue of learning and/or control associated with a pharmacological (tonic) restoration of dopaminergic transmission[16] is unlikely to result from normal sensory-evoked phasic release from absent dopamine neurones. It is therefore probably safer to assume that, for certain types of motor learning, tonic activation of dopamine receptors may be a necessary but not sufficient condition for motor learning. The contribution of other systems may be required.[1, 5]

Lastly, tonic alterations in dopaminergic transmission associated both with the disease itself and subsequent remedial medication may have further consequences for motor learning. It is possible that the magnitude of these tonic alterations in dopaminergic transmission approximates those achieved following the unpredicted occurrence or absence of reinforcing sensory events. In which case, targeted structures may be inappropriately “fooled” into reinforcing behavior (positively or negatively) that normally would be considered inconsequential.[3, 17]

With this number of multiply interacting factors it is perhaps small wonder that investigations of motor learning in Parkinson's disease have produced such a complex set of results. It is into these choppy waters that the study of Anderson and colleagues (in this issue) ventured.[18]

Anderson et al.[18] have investigated how patients with mild-to-moderate Parkinson's disease, ON and OFF their L-dopa medication, performed on 3 motor skills following extended periods of practice. In the light of recent findings with dopamine-depleted rodents,[19] the study focused on the question of whether skills acquired while under L-dopa medication would be retained if subsequently practiced in the OFF state. The 3 tasks used were: (i) finger-tapping where the subject has to switch back and forth between 2 counters as quickly and as accurately as possible; (ii) a center of mass shifting task (the same as the finger tapping task except the center of mass of the whole body was shifted back and forth); and (iii) a mirror-drawing task[4] in which the subject had to draw a star-shaped figure on the computer screen with the “mouse” reverse-programmed, ie, when the mouse moved left the cursor moved right. The patients were divided into 2 groups: (a) The ON–OFF group learnt the tasks during 4 daily sessions while ON their normal L-dopa medication. In the following week they had 4 more daily sessions where the tasks were practiced when the subjects were OFF their medication in a comparatively dopamine-deprived state. (b) The OFF–OFF group experienced an identical protocol, except they were OFF their L-dopa medication during the initial 4 learning sessions in week 1.

The principal findings of the study were as follows: For both groups, speed and accuracy improved over the initial 4 sessions on the finger tapping and body shifting tasks. Significantly, the medication had little effect on the rate of improvement and final level of performance in both tasks. In week 2, however, with both groups now OFF their medication, the performance of the ON–OFF group progressively deteriorated over the subsequent 4 practice sessions ending up close to or below that achieved at the outset. This result stood in stark contrast to the OFF–OFF group whose performance improved further, or was maintained at the asymptotic levels achieved during the first week. For the same patients, the results on the mirror-drawing task were rather different. From the outset the performance of patients OFF their medication was significantly better than that of the medicated group, and improved only slightly with further practice. In contrast, the ON–OFF group gradually improved their mirror drawing from a low start position until, by the end of week 1, their performance approached that of the unmedicated subjects. The performance of both groups was maintained during week 2. Notably, the performance of the ON–OFF group did not deteriorate with practice in the unmedicated state, as it did for the finger tapping and center of mass tasks.

Taking into consideration the factors identified in the first part of this editorial, how should the results of the Anderson et al.[18] study be interpreted? First, it must be noted that in each task the subjects did not have to learn “what” to do, rather, they had to learn “how” to do it better. This distinction is important because it may point to a differential involvement of neural systems implicated in behavioral control.[1, 5] Secondly, although not formally tested, it is probably correct to assume that the subjects were strongly goal-directed and not operating under automatic habitual control.[20] Thirdly, the finding that finger tapping and weight shifting improved with practice, independent of L-dopa medication, suggests that sufficient dopamine was present in relevant parts of the basal ganglia to allow these tasks to be performed. This suggestion would fit with a largely preserved dopamine innervation of the associative territories that support goal-directed control in patients that have mild to moderate Parkinson's disease.[6, 9] This, however, is not to imply that learning associated with skill refinement in these tasks necessarily occurs somewhere in the associative re-entrant loops. Rather, that sufficient activity in the associative basal ganglia may enable goal-directed selection of the finger tapping and body shifting tasks. In turn, this could allow other brain regions implicated in skill acquisition (hippocampus and cerebellum) to hone the subjects' performance on these tasks.[1, 5] The progressive decrement in performance of those subjects who, having perfected the task while on the ON state, then practiced in the OFF state, will be of particular interest to those who have suggested that learning plays a central role in the development of Parkinsonian symptomatology.[3, 11, 13, 17] The proposal is that dopamine denervation induces a state of aberrant corticostriatal plasticity that degrades established synaptic weights and replaces them with inappropriate, inhibitory learning that inverts the adaptive function of the basal ganglia. However, to account for practice-associated decrement in performance when subjects switched to the OFF state, the plasticity associated with skill improvement in the ON state would have to have occurred within the basal ganglia cortical or thalamic re-entrant loops. Also, the plasticity should be differentially influenced by the presence or absence of L-dopa. To test these suggestions, further work would be required to locate the practice-associated changes in neural activation within the distributed neural networks implicated in motor control (see Doyon[1] and Doyon et al[5]), while subjects performed the tasks used by Anderson et al.[18]

A different interpretation is required for Anderson and colleagues' finding with the mirror-drawing task.[18] The major result was that the performance of unmedicated patients with PD was close to asymptotic from the outset. Medicated patients reached this initial level of performance only after a week's practice. This finding is reminiscent of the report by Agostino et al.[4] that, during the first few trials, medicated patients with PD were better at mirror drawing than normal control subjects. It is a shame Anderson et al.[18] did not include a control group of normal subjects in their study. It would have been interesting to see if their medicated patients were better again at mirror drawing than normal subjects. It is, indeed, most unusual for neurological patients (especially those with Parkinson's disease) to have better performance on a motor control task than normal healthy subjects. How could this happen? A possible answer may lie with the formulation of Redgrave et al.,[6] which noted that sensorimotor territories of the basal ganglia implicated in habitual control[7] are particularly vulnerable to dopamine denervation in Parkinson's disease.[9] Consequently, their suggestion was that patients with Parkinson's disease with a differential loss of dopamine from the caudal putamen are deprived of their automatic habits and have to revert to conducting their lives in a voluntary goal-directed manner. In the case of mirror drawing, normal subjects, and possibly to some extent medicated subjects will come to the task with a well-established stimulus-response habit. When something controls something else, moving or turning the controller to the left will cause the thing that is controlled also to move to the left (exceptions are steering a boat with a rudder, and counter steering a motorcycle — both of which initially cause problems for the novice). In mirror drawing this long established dominant low-level habit will compete with the goal-directed instruction of “if you want the cursor to move to the left, move the mouse to the right.” If this analysis is correct, and habits in Parkinson patients are no longer competing options,[6] perhaps it is not surprising that people with Parkinson's disease are the kings of mirror drawing.

To conclude, despite the numerous pitfalls that have been identified with the interpretation of motor learning in patients with Parkinson's disease, the study of Anderson et al.[18] has in our view made a particularly significant contribution. First, their results give credence to the possibility that dopamine denervation induces a state of aberrant neuroplasticity. The implication might be that patients gradually “learn” to become akinetic.[3, 17] Secondly, they identify a motor task where unmedicated patients with Parkinson's disease seem to have an advantage. Perhaps this test, which is unconfounded by motor incapacity, could be used more widely as a diagnostic aid. Finally, the study confirms that a phenomenon observed in rodents[19] can also be demonstrated in humans. This is important because it increases the likelihood that future investigations of mechanism using animal models of Parkinson's disease will produce data that are relevant to human patients.


  1. Top of page
  2. References