Abstract

Patients with cerebellar stroke are impaired in procedural learning. Several different learning mechanisms contribute to procedural learning in healthy individuals. The aim was to compare the relative share of different learning mechanisms in patients and healthy controls. Ten patients with cerebellar stroke and 12 healthy controls practiced a visuomotor serial reaction time task. Learning blocks with high stimulus–response compatibility were exercised repeatedly; in between these, participants performed test blocks with the same or a different (mirror-inverted or unrelated) stimulus sequence and/or the same or a different (mirror-inverted) stimulus–response allocation. This design allowed to measure the impact of motor learning and perceptual learning independently and to separate both mechanisms from the learning of stimulus–response pairs. Analysis of the learning blocks showed that, as expected, both patients and controls improved their performance over time, although patients remained significantly slower. Analysis of the test blocks revealed that controls showed significant motor learning as well as significant visual perceptual learning, whereas cerebellar patients showed only significant motor learning. Healthy participants were able to use perceptual information for procedural learning even when the rule linking stimuli and responses had been changed, whereas patients with cerebellar lesions could not recruit this perception-based mechanism. Therefore, the cerebellum appears involved in the accurate processing of perceptual information independent from prelearned stimulus–response mappings.

INTRODUCTION

The cerebellum is essential for motor learning and the performance of skilled movements. Damage to the cerebellum is associated with impairments on a range of tasks, which often involve motor responses, such as force field adaptation (Smith & Shadmehr, 2005), habit learning (Witt, Nuhsman, & Deuschl, 2002), postural control (Diedrichsen, Verstynen, Lehman, & Ivry, 2005), and associative motor learning (Tucker et al., 1996). Less attention has been directed to nonmotor functions of the cerebellum such as cognition and the processing of sensory information and emotion (Schmahmann & Pandya, 1997).

Procedural learning is an important form of motor learning, which can be studied with the serial reaction time task (SRTT; Nissen & Bullemer, 1987). In this established paradigm, participants are presented with a series of visual stimuli, which indicate particular responses (e.g., right stimulus indicates right keypress). Unbeknown to the participants, the stimuli and, therefore, the responses appear either randomly or in a fixed sequence, which repeats cyclically after a number of trials. Procedural learning is evident by a progressive shortening of RTs in sequence trials with the participants typically unaware of it, compared with longer RTs where the stimuli are random. An important feature of the original SRTT is its similarity to everyday skill learning. Participants respond to visual stimuli according to simple rules with high stimulus–response (S–R) compatibility. This is relevant when insights gained in SRTT experiments are generalized to real life.

In SRTT experiments, several learning processes are thought to occur in parallel. Moreover, learning remains mostly implicit so that participants cannot simply report what or how they learn. To overcome this limitation and find out whether learning of the stimulus sequence or the movement sequence predominates, researchers have studied modified paradigms in which participants, for example, were required to integrate information provided by a combination of visual and acoustic stimuli (Dennis, Howard, & Howard, 2006; Goschke, Friederici, Kotz, & van Kampen, 2001), to merely watch stimuli but withhold motor responses (Kelly & Burton, 2001; Howard, Mutter, & Howard, 1992), to respond to verbal commands instead of sensory cues (Keele, Jennings, Jones, Caulton, & Cohen, 1995), or to learn tasks with low S–R compatibility (Gheysen, Gevers, Schutter, van Waelvelde, & Fias, 2009; Spencer & Ivry, 2008; Bapi, Doya, & Harner, 2000; Koch & Hoffmann, 2000; Willingham, Wells, Farrell, & Stemwedel, 2000). Although valuable insights were gained, researchers still disagree whether everyday procedural learning is primarily or exclusively response based (Bischoff-Grethe, Goedert, Willingham, & Grafton, 2004; Koch & Hoffmann, 2000; Willingham et al., 2000) or whether perceptual learning is also relevant (Abrahamse, Lubbe, & Verwey, 2009; Deroost & Soetens, 2006; Clegg, 2005; Robertson & Pascual-Leone, 2001; Keele et al., 1995).

Several studies have demonstrated that damage to the cerebellum impairs procedural learning. Whereas some SRTT studies found a complete lack of procedural learning (Shin & Ivry, 2003; Pascual-Leone et al., 1993), others found these abilities weakened but preserved (Timmann et al., 2004; Gómez-Beldarrain, García-Moncó, Rubio, & Pascual-Leone, 1998; Doyon et al., 1997; Molinari et al., 1997). It is not clear, however, how patients with cerebellar disease achieve their residual performance. The patients may employ the same or different strategies as healthy individuals, and specific mechanisms of procedural learning might be compromised, preserved, or perhaps even enhanced to compensate for lost functions.

SRTT experiments in healthy individuals that examined learning mechanisms often tested one or several conditions very dissimilar from everyday motor skills (e.g., Dennis et al., 2006; Goschke et al., 2001; Kelly & Burton, 2001; Howard et al., 1992). This may have modulated the way movement sequences have been learned. For example, when experimental constraints made one learning strategy less likely to succeed (e.g., learning of the stimulus sequence in tasks with low S–R compatibility), participants might have adopted an alternative strategy (e.g., learning of the movement sequence). In this case, the results of the experiment would not reflect a mechanism normally at work but rather a mechanism available under specific (perhaps unnatural) conditions.

The aim was to examine the contribution of different forms of procedural learning in patients with cerebellar stroke and matched healthy controls. Although we expected healthy controls to show motor as well as perceptual learning, the hypothesis was that patients are impaired in at least one form of learning. We used a standard SRTT to assure high similarity of the experimental task and everyday motor skills (Dirnberger & Novak-Knollmueller, 2013). Consequently, we did not manipulate the SRTT learning blocks but for these used a classical paradigm with high S–R compatibility. Between the learning blocks, participants completed test blocks that assessed motor learning and perceptual learning independently.

METHODS

Participants

We studied 10 patients with isolated ischemic infarction in the left (n = 3), right (n = 4), or bilateral (n = 3) cerebellum (Figure 1). Patients were selected according to the following criteria: (1) chronic phase >6 months after infarction (mean = 2.3 years, SD = 1.3 years, range = 0.5–5.0 years), (2) neurological examination indicating only cerebellar pathology, (3) no other cerebral pathology on a recent MRI or CT scan, and (4) no history of any other neurological or psychiatric disease.

Figure 1. 

Overlay lesion plots for the three lesion subgroups. The gray scale is devised so that the regions that were damaged with maximal frequency within a patient subgroup are shown in black. B = bilateral; L = left; R = right.

Figure 1. 

Overlay lesion plots for the three lesion subgroups. The gray scale is devised so that the regions that were damaged with maximal frequency within a patient subgroup are shown in black. B = bilateral; L = left; R = right.

The patients were compared with a group of 12 healthy controls. Both samples were tested with the Mini Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975), the Wortschatztest (WST; a German-language vocabulary test providing an estimate of premorbid verbal IQ; Schmidt & Metzler, 1992), the Advanced Progressive Matrices (APM; Raven, Court, & Raven, 1996), and the Beck Depression Inventory (BDI; Beck, Ward, Mendleson, Mock, & Erbaugh, 1961). Informed consent was obtained from all participants following the guidelines of the local ethics committee.

SRTT

Participants were seated approximately 60 cm away from a 17-in. computer screen. A white fixation cross appeared in the middle of the screen and remained visible for the entire experiment. Two empty white boxes on each side of that cross represented the four possible response positions. The stimulus was the whole of the box changing color from black to white. This was the signal for the participants to press the appropriate response key. The stimulus remained on screen until a response was made. Motor responses were detected with a response box, which had four keys that could be reached comfortably with the index and middle fingers of the left and right hand. The S–R allocations varied during the experiment (see below); the valid allocations were always indicated by capital letters on the left and right edge of the screen. Participants were instructed to rest all four fingers on the appropriate keys during the entire experiment and to respond to stimuli as quickly as they could while still retaining accuracy. The dependent variable (RT) was the time between stimulus onset and keypress. The response–stimulus interval was 200 msec.

Participants performed an SRTT in which 17 blocks with repetitive or pseudorandom sequences were alternated (Figure 2). There were two types of blocks with repetitive sequences:

  1. In learning blocks (L), participants performed the same 10-item stimulus sequence 45 times (i.e., 450 keypresses). The sequence was BACDACBDAD, with letters A–D representing the four boxes from left to right. In all learning blocks, the participants had to press the key right under the stimulus (i.e., standard rule: A = left middle finger, B = left index finger, etc.). There were four consecutive learning blocks. The same sequence and the standard rule were used in each block.

  2. In test blocks (T), participants also practiced a 10-item sequence for 45 times. Three different stimulus sequences were used in the test blocks: the standard learning sequence, the mirror-inverted learning sequence, and a new sequence (Table 1). The S–R allocation was also altered: In one test block, participants had to press the key right under the stimulus just as during the learning blocks (standard rule), whereas in the remaining three test blocks, they had to press keys on the opposite side of the stimulus (mirror-inverted rule: left index finger for mid-right box, right middle finger for far-left box, etc.). The following combinations were tested: (A) mirror-inverted learning sequence/standard rule (control condition I), used in conjunction with a later test to check whether participants had become aware of the mirror inversion of the learning sequence; (B) standard learning sequence/mirror-inverted rule, testing perceptual learning in the absence of motor learning—the standard stimulus sequence together with the inverted rule combined to a new motor sequence; (C) mirror-inverted learning sequence/mirror-inverted rule, testing motor learning in the absence of perceptual learning1—because the inverted stimulus sequence together with the inverted rule combined to the prelearned motor sequence, participants in this condition practiced the same movement sequence as in the learning condition but were not presented with the same stimulus sequence; (D) new sequence/mirror-inverted rule (control condition II), which, just as test conditions B and C, made high executive demands because of low S–R compatibility but, in contrast to conditions B and C, assessed neither motor learning nor perceptual learning that may have occurred in the learning blocks. The new sequence was CDBCABACBD.

All three sequences (learning sequence, inverted learning sequence, and new sequence) had a similar degree of dissimilarity to one another as indicated by the transition matrices. The order of test blocks A–D was counterbalanced between participants.

Figure 2. 

Experimental design of the SRTT. After practice (P), learning blocks (L) and test blocks (T) were arranged in a strictly alternating fashion. A random block (R) was performed before and after each learning and test block. Blocks represented by large fonts (L, T) had 450 keypresses, and blocks represented by small fonts (R) had 90 keypresses.

Figure 2. 

Experimental design of the SRTT. After practice (P), learning blocks (L) and test blocks (T) were arranged in a strictly alternating fashion. A random block (R) was performed before and after each learning and test block. Blocks represented by large fonts (L, T) had 450 keypresses, and blocks represented by small fonts (R) had 90 keypresses.

Table 1. 

Types of Blocks Used in the Experimental Design

Block
Stimulus Sequence
Rule
Test of Motor Learning
Test of Perceptual Learning
Use
Learning Standard Standard – – Main task 
Mirror-inverted Standard No No Control condition I (awareness of sequence reversal) 
Tests Standard Mirror-inverted No Yes Test of perceptual learning 
Mirror-inverted Mirror-inverted Yes No Test of motor learning 
New Mirror-inverted No No Control condition II (executive demands) 
Block
Stimulus Sequence
Rule
Test of Motor Learning
Test of Perceptual Learning
Use
Learning Standard Standard – – Main task 
Mirror-inverted Standard No No Control condition I (awareness of sequence reversal) 
Tests Standard Mirror-inverted No Yes Test of perceptual learning 
Mirror-inverted Mirror-inverted Yes No Test of motor learning 
New Mirror-inverted No No Control condition II (executive demands) 

Stimulus sequences: The standard stimulus sequence had four types of stimuli: far-left, mid-left, mid-right, and far-right. For the construction of the mirror-inverted stimulus sequence, every far-left stimulus was replaced by a far-right stimulus; every mid-left stimulus, by a mid-right stimulus; every mid-right stimulus, by a mid-left stimulus; and every far-right stimulus, by a far-left stimulus.

Rules: The standard rule required participants to press keys with the finger immediately below the displayed stimulus (i.e., far-left stimulus → left middle finger, mid-right stimulus → right index finger). The mirror-inverted rule required participants to press keys with the same finger of the other hand (i.e., far-left stimulus → right middle finger, mid-right stimulus → left index finger).

In between the learning blocks and test blocks, participants practiced blocks with a nonrepetitive pseudorandom sequence (R; 90 keypresses), which had the same item probability and transition matrix as the learning sequence and were therefore similar to the learning blocks (Figure 2). The standard (not mirror-inverted) rule was applied in these blocks. The R blocks were used to control for effects of probabilistic learning of item-by-item transitions. For this purpose, relative RTs were calculated as ratio (percentage) of performance in each learning block as compared with the preceding R block (Dirnberger, Novak, Nasel, & Zehnter, 2010; Molinari et al., 1997).

Procedure

After a demonstration and a practice (P), participants began with an R block. They then went on with the first learning block (L), followed by another R block, then the first test block (T), and so on (Figure 2). Four learning blocks and four test blocks were tested in a strictly alternating fashion.2 To indicate the currently valid S–R allocation, two capital letters “I” were displayed on the left and right edge of the screen during test blocks with the mirror-inverted rule, and two capital letters “N” during all other blocks. After completion, participants were asked whether they had noted any regularities/repetitions, any relation between the different phases of the experiment, or any other peculiarities. These questions were followed by a recall and, finally, a recognition test in which participants had to select the learning sequence out of three alternatives each presented on a separate sheet of paper.

Data Analysis

Median RTs were calculated from correct responses for each participant and block. For learning blocks (L), a repeated-measures ANOVA (rmANOVA) was calculated with median RTs as the dependent variable, the within-subject factor Time, and the between-group factor Group. Two supplementary rmANOVAs with the same design were calculated, one on relative RT data of learning blocks (referenced to preceding R block; Dirnberger et al., 2010; Molinari et al., 1997) to test for effects of probabilistic learning, and another one on median RTs of test blocks (T) across type of block to describe effects of temporal order. For the seven patients with unilateral lesions, a further rmANOVA was calculated on RTs ipsilateral versus contralateral to the side of lesion, with Side (ipsilateral, contralateral) as an additional within-subject factor. For all rmANOVAs, Huynh–Feldt corrected univariate F values are reported. As an independent indicator of learning in the learning blocks, we computed the rebound between each learning block and the subsequent random block using paired-samples t tests.

To test the main hypotheses of this experiment, paired-samples t tests were calculated for the test blocks (T) separately for each group. A first test compared the “motor learning” test block with control condition II, and a second test compared the “perceptual learning” test block with control condition II. The data from control condition I (the only test block with standard, i.e., not mirror-inverted, S–R allocation) were not entered into statistical analysis and were used for demonstrative purposes only. Independent samples t tests were calculated to compare RTs between samples (e.g., patients vs. controls). For all tests, the level of significance was set to alpha = 0.05.

RESULTS

Neuropsychological Data

Patients and controls showed no significant difference in their demographic data and a similar level of performance for all neuropsychological tests (Table 2; Mann–Whitney test, all Us > 25, all ps > .05). Across tests, both groups performed in the normal range. Performance was similar for patients with left, right, or bilateral lesions, although subsamples were too small for a formal statistical comparison.

Table 2. 

Demographic and Neuropsychological Data


Cerebellar Patients
Controls
Sex 5 men, 5 women 7 men, 5 women 
Age (years) 47 ± 15 (19–67) 43 ± 14 (21–61) 
MMSE (raw score) 29 ± 1 (27–30) 29 ± 1 (27–30) 
WST (premorbid verbal IQ) 104 ± 14 (85–133) 112 ± 9 (95–125) 
APM set 1 (spatial IQ) 110 ± 17 (81–132) 114 ± 14 (90–132) 
BDI (raw score) 9 ± 5 (1–15) 8 ± 4 (2–15) 

Cerebellar Patients
Controls
Sex 5 men, 5 women 7 men, 5 women 
Age (years) 47 ± 15 (19–67) 43 ± 14 (21–61) 
MMSE (raw score) 29 ± 1 (27–30) 29 ± 1 (27–30) 
WST (premorbid verbal IQ) 104 ± 14 (85–133) 112 ± 9 (95–125) 
APM set 1 (spatial IQ) 110 ± 17 (81–132) 114 ± 14 (90–132) 
BDI (raw score) 9 ± 5 (1–15) 8 ± 4 (2–15) 

All variables are given as mean ± SD followed by the range (in parentheses). For one patient, the WST could not be administered because his native language was not German. MMSE = Mini Mental State Examination; APM = Advanced Progressive Matrices; BDI = Beck Depression Inventory; IQ = intelligence quotient.

SRTT Learning Blocks

Figure 3 illustrates the time course of motor learning separately for patients with cerebellar lesions and healthy controls. In the beginning, both groups show a similar level of performance, although patients are slightly slower than controls. Although both groups improve their performance over time, this improvement is stronger for the controls.

Figure 3. 

Time course of original RTs across all types of block (top) and time course of relative RTs in learning blocks 1–4 (bottom), separately for patients with cerebellar lesions (solid line) and healthy controls (broken line). L = learning block, T = test block (across type), R = random block. Error bars at the top indicate 50% of standard error. Error bars at the bottom indicate standard error.

Figure 3. 

Time course of original RTs across all types of block (top) and time course of relative RTs in learning blocks 1–4 (bottom), separately for patients with cerebellar lesions (solid line) and healthy controls (broken line). L = learning block, T = test block (across type), R = random block. Error bars at the top indicate 50% of standard error. Error bars at the bottom indicate standard error.

In the statistical analysis of the learning data (L), a main effect of Time indicated a decrease of RTs from earlier to later blocks for both patients and healthy controls (F(3, 60) = 11.99, p < .001). An interaction of Time × Group (F(3, 60) = 3.99, p = .013) indicated that this decrease was stronger in healthy controls (F(3, 33) = 12.99, p < .001) than patients with cerebellar lesions (F(3, 27) = 2.97, p = .049).3 Supplementary analyses of relative RT data confirmed a significant main effect of Time (F(3, 60) = 10.11, p < .001), whereas the interaction Time × Group (F(3, 60) = 2.06, p = .127) was no longer significant. The latter analysis indicates that motor learning in a narrow sense (i.e., corrected for effects of probabilistic learning; Dirnberger et al., 2010; Molinari et al., 1997) also occurred for patients and controls.

As indicated by the increase in RTs from each learning block to the subsequent random block, control participants exhibited procedural learning that became stronger over time (paired t tests: t(11) = 1.77, 2.72, 3.01, 5.30; p = .104, p = .020, p = .012, p < .001), whereas the patients with cerebellar stroke, in accordance with previous reports (Spencer & Ivry, 2008; Gómez-Beldarrain et al., 1998; Doyon et al., 1997; Molinari et al., 1997), showed weaker and less consistent signs of procedural learning (paired t tests: t(9) = 0.86, 2.36, 0.98, 9.20; p = .412, p = .043, p = .353, p < .001).

Further analyses showed that, for patients with unilateral lesions, differences in RTs ipsilateral (mean = 424 msec, SD = 54 msec) and contralateral (mean = 414 msec, SD = 47 msec) to the side of lesion were not significant (F(1, 6) = 0.79, p = .407), nor was there a significant interaction effect with the side of lesion.

SRTT Test Blocks

In the statistical analysis of test blocks (T) across the type of block, a main effect of Time indicated that both groups performed faster in later blocks (F(3, 60) = 3.38, p = .036). A main effect of group and the Time × Group interaction were not significant.

The average RTs in control condition I (mirror-inverted learning sequence/standard rule) were, as expected, somewhat higher (controls: mean = 358 msec, SD = 142 msec; patients: mean = 457 msec, SD = 125 msec) than the average RTs in learning blocks, which were repeated four times so that more improvement was possible (controls: mean = 354 msec, SD = 140 msec; patients: mean = 430 msec, SD = 121 msec).

For the controls, the comparison specific to motor learning (i.e., mirror-inverted learning sequence/mirror-inverted rule vs. control condition II) indicated significantly shorter RTs in the motor learning condition (paired t test: t(11) = 2.33, p = .040; Figure 4). Similarly, the comparison specific to perceptual learning (i.e., original learning sequence/mirror-inverted rule vs. control condition II) indicated as well significantly shorter RTs for the perceptual learning condition (paired t test: t(11) = 2.67, p = .021). Therefore, controls displayed significant performance benefits from the re-exercise of the prelearned movement sequence as well as from the re-exercise of the prelearned stimulus sequence. The difference of these performance benefits were not significant (paired t test: t(11) = 0.53, p = .959), suggesting effects of similar magnitude.

Figure 4. 

Effects of motor and perceptual sequence learning, separately for healthy controls (white bars) and patients with cerebellar lesions (black bars). For each group, shorter RTs relative to control condition II indicate learning. Healthy controls showed significant benefits from the re-exercise of the prelearned movement sequence (motor learning) as well as the re-exercise of the prelearned stimulus sequence (visuoperceptual learning), whereas patients only benefited from the re-exercise of the prelearned movement sequence. Error bars indicate standard error. *p < .05.

Figure 4. 

Effects of motor and perceptual sequence learning, separately for healthy controls (white bars) and patients with cerebellar lesions (black bars). For each group, shorter RTs relative to control condition II indicate learning. Healthy controls showed significant benefits from the re-exercise of the prelearned movement sequence (motor learning) as well as the re-exercise of the prelearned stimulus sequence (visuoperceptual learning), whereas patients only benefited from the re-exercise of the prelearned movement sequence. Error bars indicate standard error. *p < .05.

For the patients, the comparison specific to motor learning indicated significantly shorter RTs in the motor learning condition (paired t test: t(9) = 2.57, p = .030; Figure 4), whereas the comparison specific to perceptual learning was not significant (paired t test: t(9) = 0.84, p = .425). Patients therefore displayed a performance benefit only from the re-exercise of the prelearned movement sequence but not from the re-exercise of the prelearned stimulus sequence. It should be reiterated that all the test conditions compared here applied the mirror-inverted rule and therefore made the same executive demands.

Assessment of Awareness

In the assessment of declarative memory, five patients (50%) and seven controls (58%) gave an answer indicating awareness of some sequential nature of the stimuli. This difference was not significant (chi square test: χ2 = 0.16, df = 1, p = .796). No participant could recall the sequence, and both groups performed at chance when asked to identify the sequence out of three alternatives (patients: 50% correct, controls: 33% correct; chi square test: χ2 = 0.63, df = 1, p = .429). None of the participants became aware that a pattern of visual stimuli ever was the mirror image of a previously displayed pattern.

DISCUSSION

The aim was to re-evaluate procedural learning in patients with cerebellar lesions and to compare the contribution of different learning mechanisms in patients and healthy controls. Results indicate that patients and controls learned a simple visuomotor task, although learning was reduced for the patients. The healthy controls could learn the motor sequence and the stimulus sequence independently. In contrast, the patients showed a specific inability to learn the sequence of visual stimuli, whereas their ability to learn the sequence of movements was retained.

The patients' deficit in procedural learning is consistent with previous neuropsychological studies (Shin & Ivry, 2003; Gómez-Beldarrain et al., 1998; Molinari et al., 1997), which have established that the integrity of the cerebellum is essential for such tasks. Patients and controls were well matched for verbal and spatial IQ, and none showed signs of general cognitive impairment or relevant depression. The cerebellar patients had well-defined lesions without any extracerebellar damage and were all in the chronic stage of their disease in which all nonessential functions of the cerebellum could be assumed compensated.

Acute cerebellar lesions lead to severe motor deficits predominantly ipsilateral to the side of lesion. In accordance with our findings, subsequent recovery is more complete for motor than nonmotor functions (Schmahmann & Pandya, 1997). The mechanisms of neural plasticity active during this recovery are yet unclear. They may act predominantly within the damaged cerebellar hemisphere or involve recruitment of resources from the intact cerebellar hemisphere. In consideration that our findings indicate persistent deficits in the perceptual and not the motor domain, possible compensation of impaired motor functions by the intact cerebellar hemisphere appears less relevant for the interpretation of our results.

Environmental constraints have been shown to modulate the way movement sequences are learned (Kirsch & Hoffmann, 2012). When experimental constraints make one learning strategy more exhausting or less likely to succeed, an alternative strategy might take over (Kovacs, Han, & Shea, 2009). However, this alternative strategy might be irrelevant in real life. Consequently, when experimental paradigms apply complex stimuli or learning rules, the observed mechanisms may not be the same that participants call upon outside the laboratory. Studies with more naturalistic approaches can provide an important adjunct here (Ingram & Wolpert, 2011). In everyday life, visual objects are mostly sufficient as stimuli (e.g., door knob, buttons of a shirt) and trigger straightforward responses without the requirement of further information. In our study, we therefore did not manipulate the main visuomotor learning task—the learning blocks—but in these employed a standard SRTT with high S–R compatibility. Accordingly, in the present experiment, all forms of procedural learning available in real life (i.e., response-based, stimulus-based, etc.) could occur in parallel, and no strategy or mechanism was put at a disadvantage.

Over the last decades, several studies have explored the mechanisms mediating procedural learning in healthy participants and the way in which the acquired skills are represented (e.g., Abrahamse, Jimenez, Verwey, & Clegg, 2010). Many studies have focused on response-based and perception-based models of procedural learning. Although both mechanisms were found to be relevant in various modified SRTT paradigms, researchers still disagree whether response-based mechanisms (Bischoff-Grethe et al., 2004; Koch & Hoffmann, 2000; Willingham et al., 2000) or perceptual learning (Abrahamse et al., 2009; Clegg, 2005; Keele et al., 1995) predominate in everyday procedural learning. Further research found that another mechanism, the learning of S–R associations, is also relevant. According to these findings, the activation of S–R pairs in close succession forms lasting associations between these pairs, which contribute to the representation of procedural knowledge (Schwarb & Schumacher, 2010; Spencer & Ivry, 2008; Ziessler, 1998).

In the test conditions of the present experiment, we could examine response-based and perception-based learning independently. In addition, the S–R pairs in all test conditions with a mirror-inverted rule were different from those in the learning blocks; yet, when healthy controls re-exercised the stimulus or motor sequence in one of these test conditions, their performance was better than in a control condition where both stimulus and motor sequence were new. Learning of S–R mappings therefore cannot explain these benefits. Our results indicate that, in addition to S–R mappings, other mechanisms of motor and perceptual learning are also relevant in healthy individuals. This confirms previous experiments that described procedural learning in a paradigm where S–R associations were as well altered with every trial (Goschke et al., 2001).

Spencer and Ivry (2008) examined patients with cerebellar ataxia on two versions of the SRTT in which the S–R mappings were manipulated. In a symbolic cue condition, the responses were based on the color of the stimuli presented in the center of the screen, whereas in a direct cue condition, the response location was specified by a stimulus at that location. Coding of the S–R pairs was therefore more difficult in the symbolic cue condition where participants had to learn the “meaning” of each color. The patients exhibited a severe impairment in the symbolic cue condition but not the direct cue condition. The conclusion was that (complex) S–R mappings are maintained by a network involving the cerebellum.

Our results in patients with cerebellar lesions extend these findings to cerebellar learning mechanisms independent of S–R mappings. As outlined above, healthy participants could make use of perceptual information in the test blocks even when novel S–R mappings were required (inverted rule), demonstrating the action of a perception-based learning mechanism independent of S–R mappings. In contrast, patients with cerebellar lesions were not able to use the same perceptual information, suggesting that the cerebellum plays a role in the processing of perceptual information regardless of S–R mappings.

In line with this finding, studies with the MMN (an event-related brain potential derived from the averaged EEG sensitive to the regularity/irregularity of a sequence of stimuli) in patients with cerebellar lesions have demonstrated that the processing of incoming somatosensory information is abnormal ipsilateral to the side of the lesion (Restuccia, Della Marca, Valeriani, Leggio, & Molinari, 2007). No motor response is prepared or executed in the MMN paradigm, strengthening the connection between sensory inputs and cerebellar functions. Further studies have shown that cerebellar patients are also impaired in their ability to process sequences of visual stimuli. Leggio and collaborators (2008) presented cerebellar patients and healthy controls with sets of cartoon pictures from the picture arrangement subtest of the Wechsler Memory scale (Wechsler & Stone, 1987). The pictures of each set depicted scenes of one story and were presented to the participants in a mixed-up order. The instruction was to rearrange the pictures in temporal order. Results showed that the patients with cerebellar lesions scored lower than healthy controls. To test whether cerebellar influences on sequential processing are dependent on the nature of the visual stimuli, in the same study, the authors examined another sample of cerebellar patients and controls with sets of cartoon drawings and verbal stimuli (sentences), which had to be arranged into a meaningful order so that the stimuli combined to a coherent sequence or story. Again, the patients scored lower than controls in all conditions indicating a deficit in the sequencing of visually presented stimuli.

Franz, Ivry, and Helmuth (1996) examined unimanual and bimanual tapping in patients with unilateral cerebellar lesions. Whereas during unimanual tapping, within-hand variability was larger for the ipsilesional than the contralesional hand, variability in the ipsilesional hand was greatly reduced when patients tapped with both hands together. This suggests that, during bimanual tapping, two otherwise independent oscillators are functionally coupled so that the intact cerebellar hemisphere can compensate for the lesioned hemisphere. However, in contrast to bimanual tapping, the procedural learning task in our experiment involved only unimanual movements so that such immediate coupling was not an option to improve performance.

An alternative concept can explain our findings. Ramnani (2006) proposed that the human cerebellum is not only concerned with the preparation, timing, or control of specific movements but also with the abstract goals of movements. It appears conceivable that impairments to set goals during the performance of complex sequential movements are harder to compensate than timing deficits of simple taps. This may explain why compensation is less complete in our demanding bimanual motor learning task compared with simple bimanual tapping.

As outlined above, our observations have relevance for theories that postulate a cerebellar role in cognition in addition to its traditional engagement in motor control (Ramnani, 2006; Schmahmann & Pandya, 1997). Our method and findings might also have practical value for the assessment of motor and perceptual functions in clinical settings. For example, at present, it is not clear whether cerebellar patients perform better when focusing on visual cues (i.e., try to improve impaired perceptual learning) or on their movements (i.e., attend to preserved motor learning). The patients' optimal strategy may in fact change in the course of procedural learning. Future studies could compare the time course and the attainable level of performance associated with different learning strategies in patients with cerebellar disease and other disorders.

Conclusion

Healthy participants make use of response-based and perception-based mechanisms of procedural learning. Because any progress made during learning remains intact when the rule linking stimuli and responses is changed, these mechanisms appear independent from S–R mappings. In contrast, for patients with cerebellar lesions, perception-based mechanisms of procedural learning are compromised. As outlined above, the control participants' ability and, therefore, the patients' deficit to make use of perceptual sequence information are independent from the rule linking stimuli and responses. Other stimulus-independent, response-based mechanisms of procedural learning are preserved in patients with cerebellar lesions. These mechanisms, which are also independent from prelearned S–R mappings, enable patients to achieve a moderate improvement of performance over time.

Acknowledgments

The study was in part supported by the Austrian Science Fund (Fonds zur Förderung wissenschaftlicher Forschung), Grant P13772. We are grateful to Dr. Claudia Mongini and Dr. Ingrid Schweeger for the analysis of the cerebellar lesions and the design of the lesion templates.

Reprint requests should be sent to Georg Dirnberger, Department of Clinical Neuroscience, Danube University, 3500 Krems, Austria, or via e-mail: georg.dirnberger@donau-uni.ac.at.

Notes

1. 

In the current study, movement effectors (fingers) and movement locations (key positions) are always linked. We therefore cannot differentiate between these two, which are summarized under motor learning.

2. 

After the last R block, participants were tested with a free generation task (Nissen & Bullemer, 1987), which provides a combined measure of implicit and explicit awareness. For the sake of brevity, we omit the methods and results of this task, which indicated most similar poor levels of awareness for both groups. Differences between groups were not significant.

3. 

Reduced executive functions in the patients may explain why their RTs seem to increase from learning block 6 to learning block 10, whereas for healthy controls, there is a consistent decrease of RTs from earlier to later learning blocks. The repeated change between different movement sequences and S–R allocations might be more demanding for the patients (Dirnberger et al., 2010). We note that overall procedural learning in cerebellar patients and healthy controls was not significantly correlated with several standard measures of executive functions (Dirnberger et al., 2010). This suggests that the executive functions called upon when individuals switch between different motor tasks are different from those tested with nonmotor paradigms. However, intercorrelations between various executive tests are known to be weak (Miyake, Friedman, Emerson, Witzki, & Howerter, 2000).

REFERENCES

Abrahamse
,
E.
,
Jimenez
,
L.
,
Verwey
,
W.
, &
Clegg
,
B.
(
2010
).
Representing serial action and perception.
Psychonomic Bulletin and Review
,
17
,
603
623
.
Abrahamse
,
E.
,
Lubbe
,
R.
, &
Verwey
,
W.
(
2009
).
Sensory information in perceptual-motor sequence learning: Visual and/or tactile stimuli.
Experimental Brain Research
,
197
,
175
183
.
Bapi
,
R.
,
Doya
,
K.
, &
Harner
,
A.
(
2000
).
Evidence for effector independent and dependent representations and their differential time course of acquisition during motor sequence learning.
Experimental Brain Research
,
132
,
149
162
.
Beck
,
A.
,
Ward
,
C.
,
Mendleson
,
M.
,
Mock
,
J.
, &
Erbaugh
,
J.
(
1961
).
An inventory for measuring depression.
Archives of General Psychology
,
4
,
561
571
.
Bischoff-Grethe
,
A.
,
Goedert
,
K.
,
Willingham
,
D.
, &
Grafton
,
S.
(
2004
).
Neural substrates of response-based sequence learning using fMRI.
Journal of Cognitive Neuroscience
,
16
,
127
138
.
Clegg
,
B. A.
(
2005
).
Stimulus-specific sequence representation in serial reaction time tasks.
Quarterly Journal of Experimental Psychology
,
58A
,
1087
1101
.
Dennis
,
N.
,
Howard
,
J.
, &
Howard
,
D.
(
2006
).
Implicit sequence learning without motor sequencing in young and old adults.
Experimental Brain Research
,
175
,
153
164
.
Deroost
,
N.
, &
Soetens
,
E.
(
2006
).
Perceptual or motor learning in SRT tasks with complex sequence structures.
Psychological Research
,
70
,
88
102
.
Diedrichsen
,
J.
,
Verstynen
,
T.
,
Lehman
,
S.
, &
Ivry
,
R. B.
(
2005
).
Cerebellar involvement in anticipating the consequences of self-produced actions during bimanual movement.
Journal of Neurophysiology
,
93
,
801
812
.
Dirnberger
,
G.
,
Novak
,
J.
,
Nasel
,
C.
, &
Zehnter
,
M.
(
2010
).
Separating coordinative and executive dysfunction in cerebellar patients during motor skill acquisition.
Neuropsychologia
,
48
,
1200
1208
.
Dirnberger
,
G.
, &
Novak-Knollmueller
,
J.
(
2013
).
Motor and perceptual sequence learning: Different time course of parallel processes.
NeuroReport
,
24
,
578
583
.
Doyon
,
J.
,
Gaudreau
,
D.
,
Laforce
,
R.
,
Castonguay
,
M.
,
Bedard
,
P.
,
Bedard
,
F.
,
et al
(
1997
).
Role of the striatum, cerebellum, and frontal lobes in the learning of a visuomotor sequence.
Brain and Cognition
,
34
,
218
245
.
Folstein
,
M.
,
Folstein
,
S.
, &
McHugh
,
P.
(
1975
).
“Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician.
Journal of Psychiatry Research
,
12
,
189
198
.
Franz
,
E.
,
Ivry
,
R.
, &
Helmuth
,
L.
(
1996
).
Reduced timing variability in patients with unilateral cerebellar lesions during bimanual movements.
Journal of Cognitive Neuroscience
,
8
,
107
118
.
Gheysen
,
F.
,
Gevers
,
W.
,
Schutter
,
E.
,
van Waelvelde
,
H.
, &
Fias
,
W.
(
2009
).
Disentangling perceptual from motor implicit sequence learning with a serial color-matching task.
Experimental Brain Research
,
197
,
163
174
.
Gómez-Beldarrain
,
M.
,
García-Moncó
,
J.
,
Rubio
,
B.
, &
Pascual-Leone
,
A.
(
1998
).
Effect of focal cerebellar lesions on procedural learning in the serial reaction time task.
Experimental Brain Research
,
120
,
25
30
.
Goschke
,
T.
,
Friederici
,
A.
,
Kotz
,
S.
, &
van Kampen
,
A.
(
2001
).
Procedural learning in Broca's aphasia: Dissociation between the implicit acquisition of spatio-motor and phoneme sequences.
Journal of Cognitive Neuroscience
,
13
,
370
388
.
Howard
,
J.
,
Mutter
,
S.
, &
Howard
,
D.
(
1992
).
Serial pattern learning by event observation.
Journal of Experimental Psychology: Learning Memory and Cognition
,
18
,
1029
1039
.
Ingram
,
J.
, &
Wolpert
,
D.
(
2011
).
Naturalistic approaches to sensorimotor control.
Progress in Brain Research
,
191
,
3
29
.
Keele
,
S.
,
Jennings
,
P.
,
Jones
,
S.
,
Caulton
,
D.
, &
Cohen
,
A.
(
1995
).
On the modularity of sequence representation.
Journal of Motor Behavior
,
27
,
17
30
.
Kelly
,
S.
, &
Burton
,
A.
(
2001
).
Learning complex sequences: No role for observation?
Psychological Research
,
65
,
15
23
.
Kirsch
,
I.
, &
Hoffmann
,
J.
(
2012
).
Stimulus-dependent modulation of perceptual and motor learning in a serial reaction time task.
Advances in Cognitive Psychology
,
8
,
155
164
.
Koch
,
I.
, &
Hoffmann
,
J.
(
2000
).
Patterns, chunks, and hierarchies in serial reaction-time tasks.
Psychological Research
,
63
,
22
35
.
Kovacs
,
A.
,
Han
,
D.
, &
Shea
,
C.
(
2009
).
The representation of movement sequences is related to task characteristics.
Acta Psychologica
,
132
,
54
56
.
Leggio
,
M.
,
Tedesco
,
A.
,
Chiricozzi
,
F.
,
Clausi
,
S.
,
Orsini
,
A.
, &
Molinari
,
M.
(
2008
).
Cognitive sequencing impairment in patients with focal or atrophic cerebellar damage.
Brain
,
131
,
1332
1343
.
Miyake
,
A.
,
Friedman
,
N.
,
Emerson
,
M.
,
Witzki
,
A.
, &
Howerter
,
A.
(
2000
).
The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis.
Cognitive Psychology
,
41
,
49
100
.
Molinari
,
M.
,
Leggio
,
M.
,
Solida
,
A.
,
Corra
,
R.
,
Misciagna
,
S.
,
Silveri
,
M.
,
et al
(
1997
).
Cerebellum and procedural learning: Evidence from focal cerebellar lesions.
Brain
,
120
,
1753
1762
.
Nissen
,
M.
, &
Bullemer
,
P.
(
1987
).
Attentional requirements for learning: Evidence from performance measures.
Cognitive Psychology
,
19
,
1
32
.
Pascual-Leone
,
A.
,
Grafman
,
J.
,
Clark
,
K.
,
Stewart
,
B.
,
Massaquoi
,
S.
,
Lou
,
J.
,
et al
(
1993
).
Procedural learning in Parkinson's disease and cerebellar degeneration.
Annals of Neurology
,
34
,
594
602
.
Ramnani
,
N.
(
2006
).
The primate cortico-cerebellar system: Anatomy and function.
Nature Reviews Neuroscience
,
7
,
511
522
.
Raven
,
J. C.
,
Court
,
J.
, &
Raven
,
J.
(
1996
).
Manual for Raven's Progressive Matrices and Mill Hill Vocabulary Scales.
Oxford
:
Oxford Psychologists Press
.
Restuccia
,
D.
,
Della Marca
,
G.
,
Valeriani
,
M.
,
Leggio
,
M.
, &
Molinari
,
M.
(
2007
).
Cerebellar damage impairs detection of somatosensory input changes. A somatosensory mismatch-negativity study.
Brain
,
130
,
276
287
.
Robertson
,
E.
, &
Pascual-Leone
,
A.
(
2001
).
Aspects of sensory guidance in sequence learning.
Experimental Brain Research
,
137
,
336
345
.
Schmahmann
,
J.
, &
Pandya
,
D.
(
1997
).
The cerebrocerebellar system.
International Review of Neurobiology
,
41
,
31
60
.
Schmidt
,
K.
, &
Metzler
,
P.
(
1992
).
Wortschatztest.
Weinheim, Germany
:
Beltz Test GmbH
.
Schwarb
,
H.
, &
Schumacher
,
E.
(
2010
).
Implicit sequence learning is represented by stimulus–response rules.
Memory and Cognition
,
38
,
677
688
.
Shin
,
J.
, &
Ivry
,
R.
(
2003
).
Spatial and temporal sequence learning in patients with Parkinson's disease or cerebellar lesions.
Journal of Cognitive Neuroscience
,
15
,
1232
1243
.
Smith
,
M.
, &
Shadmehr
,
R.
(
2005
).
Intact ability to learn internal models of arm dynamics in Huntington's disease but not cerebellar degeneration.
Journal of Neurophysiology
,
93
,
2809
2821
.
Spencer
,
R. M.
, &
Ivry
,
R. B.
(
2008
).
Sequence learning is preserved in individuals with cerebellar degeneration when the movements are directly cued.
Journal of Cognitive Neuroscience
,
21
,
1302
1310
.
Timmann
,
D.
,
Drepper
,
J.
,
Calabrese
,
S.
,
Burgerhoff
,
K.
,
Maschke
,
M.
,
Kolb
,
F.
,
et al
(
2004
).
Use of sequence information in associative learning in control subjects and cerebellar patients.
Cerebellum
,
3
,
1
8
.
Tucker
,
J.
,
Harding
,
A.
,
Jahanshahi
,
M.
,
Nixon
,
P.
,
Rushworth
,
M.
,
Quinn
,
N.
,
et al
(
1996
).
Associative learning in patients with cerebellar ataxia.
Behavioral Neuroscience
,
110
,
1229
1234
.
Wechsler
,
D.
, &
Stone
,
C.
(
1987
).
Wechsler Memory Scale—Revised.
San Antonio, TX
:
The Psychological Corporation
.
Willingham
,
D.
,
Wells
,
L.
,
Farrell
,
J.
, &
Stemwedel
,
M.
(
2000
).
Implicit motor sequence learning is represented in response locations.
Memory and Cognition
,
28
,
366
375
.
Witt
,
K.
,
Nuhsman
,
A.
, &
Deuschl
,
G.
(
2002
).
Dissociation of habit-learning in Parkinson's and cerebellar disease.
Journal of Cognitive Neuroscience
,
14
,
493
499
.
Ziessler
,
M.
(
1998
).
Response-effect learning as a major component of implicit serial learning.
Journal of Experimental Psychology: Learning Memory and Cognition
,
24
,
962
978
.