Introduction

By simply thinking about playing golf, skiing, or writing, we realize that sequences of movements included in well-acquired motor skills can be executed smoothly and rapidly without conscious efforts, and with a minimal need to rely on environmental cues. Kinematic and dynamic parameters of these integrated movements are well internalized, and error-based adjustments remain minimal. Such motor skills must however be acquired through practice, and it is generally assumed that they follow two progressive stages of learning (Luft and Buitrago 2005). An early or first stage of learning is characterized by rapid improvements from one trial to another and by large fluctuations in performance, because subjects mainly rely on trial and error (Karni et al. 1998). Extended cortical structures, including the primary motor, dorsolateral prefrontal, anterior cingulate, posterior parietal, and cerebellar cortices, are thought to be involved in this early learning stage (Grafton et al. 1992; Jueptner et al. 1997; Penhune and Doyon 2002; Doyon et al. 2002; Puttemans et al. 2005). At the end of the early learning stage, more circumscribed activity has been described in the cerebellar dentate nucleus and basal ganglia (Doyon et al. 2002; Floyer-Lea and Matthews, 2005). The second stage of learning is characterized by slower incremental gain in performances over several sessions of practice. When there are a few or no variations of performances with the passage of time, or from one practice session to the other, motor skills can be considered optimal and stabilized. Both the neural (Kleim et al. 2004) and functional processes (Dudai 2004) which underly such stabilization are usually recognized as the long-term motor consolidation. The striatum and sensorimotor cortex have been found to be involved in this consolidation stage (Floyer-Lea and Matthews 2005).

Patients with diseases affecting the striatum, such as Parkinson’s disease (PD) and Huntington’s disease (HD), exhibit reduced learning capacity, especially during the second learning stage (Harrington et al. 1990; Doyon et al. 1997; Sarazin et al. 2002). In patients with schizophrenia, different learning patterns exist depending on the pharmacological profile of their antipsychotic medication. For example, patients treated with haloperidol show significant fluctuations of motor performance on a mirror drawing task, particularly during the late learning stage, while such fluctuations are not observed in patients treated with atypical antipsychotics such as clozapine (Bedard et al. 1996, 2000; Scherer et al. 2004). This effect has been attributed to the higher D2R affinity of the conventional drugs over atypical second-generation antipsychotic drugs (Bedard et al. 1996, 2000; Scherer et al. 2004). In order to test this hypothesis, Paquet et al. (2004) assessed schizophrenic patients treated either with olanzapine or haloperidol on the Computed Visual Tracking Task, in which subjects must track a moving target on a screen using a computer mouse. Poor motor learning, characterized by large inter-trial fluctuations and low learning rates, was observed in the haloperidol group, but not in the olanzapine group. Moreover, motor learning impairments in the haloperidol-treated group correlated well with the level of striatal D2R occupancy, as assessed with Single Photon Emission Computed Tomography (SPECT) and 123I-Iodo-benzamide (123I-IBZM).

These findings are concordant with results obtained in healthy volunteers acutely exposed to central dopamine enhancing or blocking drugs (Kumari et al. 1997). In a target-tracking task requiring subjects to touch a target that appeared successively in one of four quadrants on a touch-screen monitor, motor learning improved or worsened depending on whether subjects received a dopamine enhancer (amphetamine) or antagonist (haloperidol) respectively. Moreover, level of performance under haloperidol was characterized by large fluctuations from trial to trial. Taken together, results from these studies suggest a primary role of dopamine, and more specifically of the striatal D2R in motor consolidation.

Animal studies have suggested that the role of the striatum in motor learning lies in the combination or the chunking of discrete movements into motor sequences (Graybiel 1998; Jog et al. 1999). Chunking refers to a learning strategy where discrete items of information are grouped together in meaningful units (Miller 1956). The main advantage of this strategy is that it can expand the limit of short-term memory by increasing the number of items to retain by keeping only the grouping units. In motor learning studies, it was found that a motor sequence can also be organized into chunks, defined as isolated movements grouped into single units according to their kinematic and kinetic properties (Koch and Hoffmann 2000; Sakai et al. 2003; Verwey and Eikelboom 2003).

The present study is an effort to further clarify the effect of a selective D2R antagonist, such as raclopride, on the consolidation stage of a motor sequence learning task in two Cebus apella monkeys. It was expected that D2R antagonist will disturb consolidation by interrupting the movement chunking process.

Our first hypothesis was that raclopride will affect new sequence learning at the time of consolidation. This disturbance will be observed in persistent fluctuations of performance from trial to trial, leading to a delayed second stage of learning (consolidation). Distinction between the first and second stages of learning has been quantitatively determined, and a fluctuation score was created for this purpose. Our second hypothesis was that such a disturbed consolidation process would be related to difficulties in chunking a new movement that is added to a well-established one. A chunking score was specifically created for this purpose.

Materials and methods

Animals

Two Cebus apella monkeys (2–4 kg), J and L, were used in this study. They were kept in individual cages and exposed to a 12 h light–dark cycle. Monkeys were fluid-restricted during periods of training. Body weight and food consumption were monitored daily. All procedures were approved by the institutional Animal Research Ethics Review Board of the University of Montreal, in accordance with the standards of the Canadian Council on Animal Care.

Behavioral task

The two monkeys were trained to sit in a monkey chair facing a 25 × 25 cm box placed 30 cm in front of them, and on which there were five light-emitting push-buttons (3 × 3 cm) that could be lit independently (Fig. 1a). Monkeys were required to press on push-buttons to obtain the reward (drop of water). Button illumination and data acquisition were controlled by a computer equipped with custom-made software (BOX, DOCO Microsystems Inc. Montréal, Canada). The layout is shown in Fig. 1a and b. Motor sequences consisted of pressing successively three lit push-buttons. Monkeys had to press one push-button before illumination of the next. All sequences covered the same spatial length. For each correct sequence, reward was given after the third push-button was pressed. If the monkey did not press any push-button after 4 s, or if it pressed on an incorrect push-button, the trial was cancelled and a new trial began. Trials were administered at a random interval between 1 and 3 s. Before learning any sequence, monkeys were required to press on push-buttons randomly lit one at the time, in order to explore the entire box. When monkeys were able to press on each push-button with a success rate of at least 80%, sequence learning began. Monkey L was left-handed and was trained to learn the sequence clockwise, while monkey J was right-handed and learned the sequence counter-clockwise.

Fig. 1
figure 1

Task and main steps of the protocol. a Schematic representation of the experimental setup. The monkey is facing a box (25 cm × 25 cm) on which five push-buttons (3 cm × 3 cm) are placed in an “X” pattern. Illumination of the push-buttons was remotely controlled by a computer. b Representation of motor sequences that had to be learned by each monkey. The first portion (P1) of each sequence was always the same while the second portion (P2) changed according to each sequence. c Flow diagram showing the protocol parts. Monkeys first overlearned sequence A. Sequence B was then learned and sequence A recalled under no drug effect. In the second part of the experiment, monkeys had to learn a new sequence C and recall the sequence A under the effect of raclopride

Full size image

The first portion of each motor sequence (first and second push-button) was kept the same for all sequences, while the second portion (second and third push-button) was changed according to the experimental condition (Fig. 1b). By distinguishing the two components of the sequences, we aimed at better defining the chunking process of a new motor component (push buttons 2–3) to a well established one (push buttons 1–2). For each sequence, there were one 10–15 minute session per day, and each session contained between 80 and 120 trials. A first sequence (Seq. A) was learned and performed over 80 sessions, and considered as the over-learned baseline sequence. Thereafter, the study was divided in a first and a second part, which correspond respectively to the experimental conditions without and with drug. The first part involved learning and recalling motor sequences drug-free, while the second part involved learning and recalling motor sequences under the effect of raclopride (Fig. 1c). During the first part of the study, a new sequence (Seq. B) was learned in a drug-free condition for 20 sessions. The over-learned sequence (Seq. A OFF-drug) was then recalled during five sessions, in order to have a sufficient number of trials for comparisons between conditions. Then, the second part of the protocol was initiated, during which a new sequence (Seq. C) was learned for 20 sessions under the effect of raclopride. The overlearned sequence was then recalled (Seq. A ON-drug) for five sessions, under the effect of raclopride.

Dopamine antagonist exposure

For the purpose of studying motor sequence learning under a dopamine antagonist, monkeys received a daily subcutaneous injection of S(−)-Raclopride (+)-tartrate salt (Sigma–Aldrich, Montréal, Canada), dissolved in 0.9% saline solution. Raclopride has a high affinity for D2R, moderate affinity for D3R, and no affinity for the other dopamine receptor subtypes (Koepp et al. 1998; Rosa-Neto et al. 2004). This drug also has a short half-life, thus making drug accumulation resulting from daily dosing less likely (Ahlenius et al. 1991). Doses of 0.01–0.02 mg/kg are known to induce striatal D2R occupancy up to 77%, without motor side effects (Wadenberg et al. 2000).

In order to determine the optimal individual dose that would induce a high rate of D2R occupancy under the threshold for motor or systemic side effects, the monkeys first received test doses of vehicle and raclopride at 2–3 days interval. They were given random doses ranging from 0.005 to 0.03 mg/kg, until they showed hyperkinetic or hypokinetic movements. A neurologist (P.B.) trained to evaluate the extrapyramidal symptoms in Cebus apella monkeys, and blind to the injection protocol, scored the behavior of the animals on a motor scale during four hours following the injection. This scale was an adapted version of the “Abnormal Involuntary Movements Scale” (AIMS) used in humans, and reminiscent of another scale used to rate dyskinesias in parkinsonian monkeys (Samadi et al. 2003). Dyskinetic movements affecting each body part were rated on a severity scale ranging from 0 to 4. The monkey’s overall activity was also monitored with a digital camera connected to a computer equipped with custom-made software. This allowed us to identify any symptom of bradykinesia or sedation that may have appeared following an injection of raclopride. Following this dose selection method, the monkeys were administered a daily s.c. dose of 0.02 mg/kg (monkey L) and 0.005 mg/kg (monkey J) 30 min before each testing session. Since raclopride is known to reach peak effect and saturation of D2R at 30–45 min after administration (Koepp et al. 1998), the monkeys began testing at that time and for a maximum duration of 15 min, during which they completed 80–120 successful trials.

Data analysis

Learning stages

The first and second motor learning stages in Seq.B and Seq.C were defined by using the K-clustering method (MacQueen 1967), which allowed us to separate two sets of execution time (ET) data. ET was defined as the time required to execute the whole sequence, starting at the moment where the monkey pressed the first push-button to the moment where it pressed the last. ET values during the last 10 days of the overlearned sequence (Seq.A) were used as the optimal stage 2 level of performance. When 85% of trials within a day reached the stage 2 level for five consecutive days, we assumed that the monkey had reached the second and optimal learning stage.

Fluctuation score

A fluctuation score was developed based on the difference between the ET value for each trial and its corresponding value on a filtered curve. To obtain an appropriately fitted curve, ET values were low-pass filtered (Gaussian window 2.5–5 Hz cut-off, 50 values) to remove individual trial fluctuations. The median fluctuation value was indicative of the degree of dispersion of the ET values around the filtered curve, reflecting the fluctuation of performance. The higher the fluctuation score was, the higher the variation of performance was along trials. This fluctuation score allowed us to test the first hypothesis that a selective D2R antagonist would induce fluctuation of performances when monkeys would have to learn new motor sequences. This variable was also used to compare fluctuation of performance between the first and the second learning stages of motor sequences.

Chunking score

Chunking score was used in order to test our second hypothesis that fluctuations in learning would be related to the inability to group movements into sequences. First, the difference between the ET for the first part of the sequence (T1) (time between the first and the second push-button) and the ET for the second part of the sequence (T2) (time between the second and the third push-button) was calculated. This yielded a score for each trial. Secondly, these scores were filtered, using the same method described above for the fluctuation score. Then, by calculating the difference between the original scores and the filtered scores, we obtained a median chunking score for each learning stage of each condition. We assumed that, as learning progresses, the difference between T1 and T2 would diminish and stabilize as the separate movements are chunked into a single sequence. According to our analysis, this should give a high chunking score for the first learning stage and a low chunking score for the second stage. More precisely, if a motor sequence is not well chunked, as in the first learning stage, the difference between T1 and T2 should fluctuate with time, because of monkey’s inability to execute the sequence with constant kinematic parameters. But as the subject reaches the second learning stage, the chunking score should decrease, because the motor sequence is progressively executed at the same speed from trial to trial, and the second part is then fully integrated with the first overlearned part.

For the purpose of statistical analysis, data were log-transformed (Fluctuation scores: Log10 (x + 10), Chunking scores: Log10 (x + 15)) to meet parametric test criteria. Differences between conditions and learning stages within a condition were identified by Scheffé post-hoc tests, and significance threshold was set at P < 0.05.

Results

Baseline overlearned sequence (Seq.A)

Raw data and filtered curves for the last 10 days of training to Seq.A are shown for the two monkeys in Fig. 2a and d. Fluctuation scores were significantly lower during the latter days than during the former, for monkey L (F = 69.941, P < 0.001) and monkey J (F = 15.724, P < 0.001). The last 10-day performances were therefore used as the baseline overlearned performance, that is to say the consolidated performance. The first 10 day performances were not used as baseline learning, given that this period corresponded to the first exposition of the monkeys to the task, and therefore was influenced by other factors not related to learning per se (e.g., familiarity to this novel task).

Fig. 2
figure 2

Performances for SEQA recalled with no drug and with raclopride. Performances during the last 10 days of SEQ A for monkey L (a) and monkey J (d). The black line represents the filtered curve (see Materials and methods). Performances for SEQ A OFF-drug for monkey L (b) and monkey J (e). Performances for SEQ A ON-drug for monkey L (c) and monkey J (f). Fluctuation scores for each condition and each monkey. For each trial, the execution time value and its corresponding point on the filtered curve were subtracted. A median value of all the differences was then calculated to obtain the Fluctuation score (g). There is no significant difference between each condition (SEQ A, SEQA OFF-drug, SEQA ON-drug), indicating that monkey can successfully recall an overlearned sequence, even under raclopride

Full size image

Part one: OFF-drug conditions

Learning of a new sequence (Seq.B) without drug

Motor sequence learning without drug was assessed with a new sequence (Seq. B). The switching from the first to the second learning stage in Seq. B occurred near the 600th trial for monkey L and the 200th trial for monkey J (Fig. 3a, c). The fluctuation score was lower during stage 2 of learning compared to stage 1 for both monkey L (P < 0.001) and monkey J (P < 0.001) (Fig. 3e). Chunking scores comparisons also revealed significantly lower values during stage 2 than during stage 1 in both monkey L (P < 0.001) and monkey J (P < 0.001) (Fig. 3f). Motor sequence learning without drug is therefore characterized by a significant reduction of performance fluctuations from stage 1 to stage 2. A successful incorporation of the second to the first movement portion was also observed, as revealed by a significant decrease of the chunking score in stage 2 compared to stage 1.

Fig. 3
figure 3

Performances for new sequences learned with no drug and with raclopride. Performances for SEQ B (new learning OFF-drug) for monkey L (a) and monkey J (c). The vertical line separates learning stage 1 from learning stage 2, as determined by the K-clustering method. Performances for SEQ C (new learning ON-drug) for monkey L (b) and monkey J (d). e. Histogram of the fluctuation scores presented separately for each learning stage, during SEQ B and SEQ C. This score was always lower during the second learning stage, whether or not the sequence was learned under raclopride. f Histogram of the chunking score presented separately for each learning stage during SEQ B and SEQ C. A low chunking score reflects a well-chunked motor sequence. This score was also significantly lower for the second learning stage. *P < 0.001

Full size image

Recall of baseline overlearned sequence without drug (Seq.A OFF-drug)

Figure 2b and e show that, following the learning of Seq. B, the recall of Seq. A is accomplished without any change from what was observed at baseline. Fluctuation scores did not differ between the two experimental conditions for monkey L (P = 1.000) and monkey J (P = 0.991) (Fig. 2g). Thus, there was no retroactive interference of Seq. B on the recall of Seq. A, that is to say that monkeys can recall an old sequence, even after they have learned a new sequence meanwhile.

Part two: on-drug conditions

Learning of a new sequence under raclopride

Effects of raclopride on the new sequence learning were assessed with Seq.C (Fig. 3b, d). Results showed that monkeys were able to learn this sequence as in the condition without drug. Fluctuation scores during Seq.C were found to significantly decrease from stage 1 to stage 2 in both monkey L (P < 0.001) and monkey J (P < 0.001) (Fig. 3e). It was also found that the chunking score improved significantly during stage 2 as compared with stage 1, for monkey L (P < 0.001) as well as for monkey J (P < 0.001) (Fig. 3f).

Despite such a preserved improvement of fluctuation and chunking scores during motor learning under raclopride, it seems that stage 2 was not reached as fast in the raclopride condition as in the OFF-drug condition (Fig. 3a–d). Monkey L needed twice as much trials to reach the second stage and monkey J needed four times more trials, under the D2R antagonist. Raclopride seems therefore to delay the stage 2, by inducing prolonged fluctuations of performance.

The effect of raclopride on the fluctuation score was analyzed across the experimental and drug conditions, in the two monkeys separately, by using a general seven-levels one-way ANOVA with Scheffé post-hoc tests (Seq.A Baseline [stage2]; Seq.A OFF-drug [stage2]; Seq.A ON-drug [stage2]; Seq.B [stage1]; Seq.B [stage2]; Seq.C [stage1]; Seq.C [stage2]). Results revealed significant differences between fluctuation scores for both monkey L (F = 77.729, P < 0.001) and monkey J (F = 89.110, P < 0.001). Post-hoc comparisons for both monkeys revealed that the fluctuation score in stage 2 is greater in Seq.C than in Seq.B (monkey L: P < 0.001; monkey J: (P < 0.001) (Fig. 4c)). In stage 1, the fluctuation score is also greater in Seq.C than in Seq.B, but this was observed only for monkey L (P < 0.001), not for monkey J (P = 0.809) (Fig. 4a).

Fig. 4
figure 4

Fluctuation and chunking scores for new sequences learned with no drug and with raclopride. a Histograms showing the fluctuation scores for the first learning stage. Significant differences were found between SEQ B (OFF-drug) and the SEQ C (ON-drug) in monkey L but in monkey J. b Histograms of the chunking scores for the first learning stage. No differences were found between the two experimental conditions. c Histogram of the fluctuation scores for the second learning stage. Raclopride induced more fluctuations in the two monkeys. d Histogram of the chunking scores for the second learning stage. Chunking scores are significantly higher with raclopride than without, for the two monkeys, indicating that movements are not well chunked into a single unit under raclopride. *P < 0.001

Full size image

Chunking scores were also analyzed across the experimental and drug conditions in the two monkeys separately. Comparisons were carried out on the two stages (stages 1 and 2) of the two sequences requiring motor learning (Seq. B and Seq. C). The 4-levels one-way ANOVA (Seq.B [stage1]; Seq.B [stage2]; Seq.C [stage1]; Seq.C [stage2]) revealed significant differences (monkey L (F = 62.913, P < 0.001; monkey J (F = 142.715, P < 0.001). Scheffé post-hoc comparisons revealed no chunking score difference between Seq. B and Seq. C during stage 1 for both monkey L (P = 1.000) and monkey J (P = 0.810) (Fig. 4b). However, significant differences were observed between Seq. B and Seq. C during stage 2, in the two monkeys (monkey L (P < 0.001); monkey J (P < 0.001) (Fig. 4d). This indicates that, at the time of consolidation (stage 2), raclopride was associated with a difficulty in joining the second and new portion of a movement to a first and well established one that is to form a single sequence unit.

Recall of the baseline overlearned sequence under raclopride (Seq.A ON drug)

Monkeys had to recall the overlearned sequence A for five consecutive sessions under the effect of raclopride (Seq.A-ONdrug) (Fig. 2c, f). Fluctuations scores did not differ from baseline in monkey L (P = 1.000), nor in monkey J (P = 0.840), supporting the view that the raclopride-induced fluctuations observed during a new sequence learning (Seq.C) do not generalized to a well established motor sequence (Seq.A) (Fig. 2g).

Discussion

In this study, we investigated the ability of two monkeys to learn and recall motor sequences under the influence of raclopride, a selective D2R antagonist. Our results suggest that, although raclopride did not prevent monkeys from learning, it induced prolonged fluctuations of performances, which in turn delayed the occurrence of the second stage of learning, or consolidation. Even during the second stage of learning, at a time when performances should remain constant from trial to trial, raclopride-induced fluctuations were still present, indicating a difficulty to consolidate long-term memories of the learned sequence. Moreover, the chunking score in the two monkeys was significantly affected by raclopride during the second stage of learning, but not during the first one. This specific effect might be related to an inability to combine movements into a single sequence at the time of consolidation.

The main limit of our study is the absence of a control condition. Saline injections during the OFF-drug condition would have completely ruled out the possibility that the effects of raclopride are due to stress induced by the injections. However, since the performances for the three recall conditions (Seq. A, Seq. A recall OFF-drug, Seq. A recall ON-drug) of the overlearned sequence are comparable, factors such as stress can be partially ruled out as a main cause of motor consolidation deficits.

One might also argue that learning per se is also affected by raclopride, since monkeys needed two to four times more trials to reach the second learning stage and that a review of the literature shows that D2 blocking substances have a deleterious effect on motor learning (Bedard et al. 1996; Kumari et al. 1997; Scherer et al. 2004; Paquet et al. 2004). However, since we have observed a significant decrease of total execution time values and fluctuation scores, and an increased chunking efficiency along trials whether animals were in ON or in OFF drug conditions, we reasoned that the capacity to learn new motor sequences is still maintained under the raclopride treatment. In other words, monkeys show an incapacity to consolidate new motor sequences rather than to specifically reach a good level of performance.

Despite this clear effect on motor sequence consolidation, there was no effect of raclopride on the ability to recall an already well-learned motor sequence. As a whole, our results demonstrate that, even without any extrapyramidal manifestation, the administration of a selective D2R antagonist may affect motor learning consolidation, suggesting the high sensitivity of the latter function following a dopamine blockade.

Fluctuations and chunking difficulties are always expected at the beginning of motor sequence learning, because of the trial and errors learning process that takes place at this time. However, as the motor sequence becomes well integrated, fluctuations and chunking difficulties should decrease with time. This is what was observed in the current study during the second stage of learning, whether monkeys were treated or not with raclopride (Seq.B and Seq.C), thus confirming that raclopride does not prevent from learning. However, fluctuations and chunking disturbances were nevertheless greater under raclopride (Seq.C) than during the OFF-drug condition (Seq.B). The deleterious effect of raclopride in the second but not in the first stage of learning suggests a specific effect at the time of consolidation and not at the time of learning per se. Given that D2R are mostly located in the striatum and that this structure has been found to be involved in motor learning consolidation (Floyer-Lea and Matthews 2005), we believe that results of the present study might be explained by a deleterious effect of raclopride in the striatum. Therefore, these results and their interpretation are concordant with an accumulating body of evidences regarding the relationship between motor learning and dopamine, and strongly support previous results obtained in patients with schizophrenia treated with D2R antagonists (Bedard et al. 1996, 2000; Scherer et al. 2004).

Results obtained by others are concordant with our view that the consolidation of new motor skills may involve a striatal and dopamine dependant chunking process of new individual movements to old and well-established ones. Using a task similar to ours, Matsumoto et al. (1999) have prevented the learning of a motor sequence task by a selective and unilateral lesion of the dopamine nigrostriatal fibers of primates. In their study, multiple two component sequences had to be learned, all identical in their first (well established) portion, but different in their second (new) portion. Performances in the first portion were always the same whether the monkeys were lesioned or not. However, performances in the second portion improved only with the arm ipsilateral to the site of lesion, not with the controlateral one. This suggests that monkeys had problems in chunking new movements to old and already learned ones, or to form a single motor sequence.

In rats, the striatum seems to encode the serial order of syntactic sequences of natural behaviors, such as grooming. This behavior in rats always follows predictable patterns, with up to 25 forelimb strokes and body lick movements combined into a four-phase syntactic chain of actions that always lasts approximately 5 s. Dopamine depletion in the neostriatum, such as the one observed following 6-OHDA exposure, disrupts the serial structure of the chain pattern, without disrupting the execution of single movement components of the grooming behavior executed outside a sequential context (Berridge and Whishaw 1992; Cromwell and Berridge 1996). Recording studies have also demonstrated that during the grooming behavior, an important number of cells in the dorsolateral and ventromedial neostriatum code the sequential pattern of syntactic chains and only few neurons code simple motor properties of grooming movements (Aldridge and Berridge 1998). For instance, neurons in the dorsolateral region increase firing only during syntactic chains. These findings suggest that neuronal activity in the striatum is related to the execution of specific motor sequences and that dopamine may be involved in the process where movements are linked together.

Evidence for a role of the striatal dopaminergic system in movement chunking may also come from PD patients. It has been shown that, compared to treated PD patients, non-treated patients have difficulties in performing movements embedded in a sequential context, whereas they can perform isolated movements as a discrete task (Benecke et al. 1987). In a finger-tapping task in which different sequence lengths had to be executed, Stelmach et al. (1987) have also observed a dissociation of the first and subsequent taps in PD patients, as revealed by a pattern of progressively increasing errors with longer tap sequences in Parkinsonians. Similar conclusion can be drawn from the results of Yaguez et al. (2006) who used a Corsi Block Tapping-test, that required the subject to point sequentially to the targets in the visuospatial order of their illumination. In this task, sequence length increases with trials, and patients with PD had difficulties in adding new items to the basic three-component sequence. Although attention control and short-term memory probably play a significant role in performing such sequential movements, all these authors concluded that there is a deficit in the programming and execution of movement sequences in PD. We believe rather that a dopamine dependant chunking difficulty may play a role in such sequencing deficits.

It might be suggested that the effect of raclopride in the present study may be related to an unspecific attentional, alertness, or a motivational deficit induced by dopaminergic blockade. Such deficits have been reported in animal with dopamine depletion (Salamone et al. 1995; Cousins and Salamone 1996), in PD patients (Gotham et al. 1988; Brown and Marsden 1991; Cronin-Golomb et al. 1994; Cools et al. 2001; Tamura et al. 2003), in patients with schizophrenia treated with antipsychotics (Purdon et al. 2001; Saeedi et al. 2006), and in healthy controls tested after an acute dose of sulpiride (Mehta et al. 2005). However, if monkeys had such non-specific deficits, it is likely that these deficits would have been observed not only during Seq.C, but also during the recall of Seq.A ON-drug. It can further be argued that the non-specific effects of raclopride are more likely to occur during Seq.C, because this is the only condition under raclopride that contains a new element (second portion of the sequence), for which attention and alertness are necessary. However, the effect of raclopride on Seq.C is mainly observed during the second stage of learning. The maximal attention and alertness capacities are required during the first learning stage, at a time when novelty must be detected and readjustment of movements must be done. A raclopride-induced lack of motivation can also be ruled out in the present study, given that the two monkeys performed approximately the same number of trials at each session, and drank the same amount of water whether they were in the OFF-drug or ON-drug conditions. Therefore, we strongly believe that the results obtained with raclopride here are the consequence of a dopamine blockade, which produces in turn a specific difficulty in integrating portions of movements into fluent sequences, during the consolidation process. Further studies are warranted to document the neural substrate underlying this chunking process, and to determine whether other dopamine antagonist substances, with no D2R affinity would produce the same effect.