Elsevier

Brain Research

Volume 1137, 16 March 2007, Pages 117-130
Brain Research

Research Report
The striatum and probabilistic implicit sequence learning

https://doi.org/10.1016/j.brainres.2006.12.051Get rights and content

Abstract

The distinction between implicit (unconscious) and explicit (conscious) learning is controversial. Some argue that explicit learning is dependent on the medial temporal lobes, whereas implicit learning is mediated by the basal ganglia and others propose that all learning is explicit. The purpose of the present study was to investigate the involvement of the basal ganglia in implicit learning by examining learning of a probabilistic sequence of targets, in patients with Parkinson's disease (PD) and controls. Following learning, we assessed participants' awareness of the sequence structure by asking them to generate or withhold sequence consistent responses (process dissociation procedure) and to perform a recognition test in which measures of priming and recognition were obtained concurrently. Although the PD group demonstrated evidence of probabilistic sequence learning in this study, learning was significantly attenuated compared to controls. Residual learning in the PD group was genuinely implicit in nature because (a) PD patients were not able to control the expression of their acquired knowledge, and (b) their knowledge supported subsequent priming of sequence-consistent responses but recognition ratings were at chance. In contrast, following learning controls were capable of above chance recognition indicating that their sequential knowledge was acquired in a more explicit way. The results support the view that (i) the basal ganglia contribute to probabilistic implicit sequence learning (ii) that such learning can occur implicitly without explicit knowledge in PD patients.

Introduction

It has been suggested that striatal structures with their cortical projections support implicit (unconscious) habit learning whereas the cortico-limbic-diencephalic structures are considered the substrate for explicit (conscious) learning (e.g., Butters et al., 1985, Cohen and Squire, 1980). There is now considerable evidence for the role of the striatum in habit learning from animal lesion studies (Fernandez-Ruiz et al., 2001, Jog et al., 1999, McDonald and White, 1993, Packard and McGaugh, 1992, Packard et al., 1989), clinical studies of probabilistic category learning and sequential learning in patients with Parkinson's disease (PD) or Huntington's disease, disorders that involve striatal pathology (Brown et al., 2001, Brown et al., 2003, Doyon et al., 1997, Doyon et al., 1998, Ferraro et al., 1993, Filoteo et al., 1998, Helmuth et al., 2000, Jackson et al., 1995, Kelly et al., 2004, Knopman and Nissen, 1991, Knowlton et al., 1996a, Knowlton et al., 1996b, Pascual-Leone et al., 1993, Sage et al., 2003, Shin and Ivry, 2003, Shohamy et al., 2004a, Shohamy et al., 2004b, Smith et al., 2001, Sommer et al., 1999, Westwater et al., 1998, Witt et al., 2002, Willingham and Koroshetz, 1993) and neuroimaging (Doyon et al., 1996, Grafton et al., 1995, Peigneux et al., 2000, Poldrack et al., 1999, Poldrack et al., 2001). Furthermore, a recent meta-analysis of the findings of six previous studies of implicit sequence learning in PD concluded that implicit learning is significantly impaired in PD (Siegert et al., 2006). An alternative view is that the basal ganglia may play no special role in the development of unconscious knowledge. This position is supported by evidence indicating that in addition to implicit learning deficits, PD patients are impaired at specific explicit tasks such as the Wisconsin Card Sorting Task (e.g., Monchi et al., 2004). In addition, other evidence indicates that the same rather than different brain regions are recruited in healthy participants during implicit and explicit learning (e.g., Willingham et al., 2002). Indeed, some researchers reject the notion that learning can be implicit and argue that all human learning relies on a unitary explicit system (e.g., Wilkinson and Shanks, 2004).

One task which has been primarily been used to examine implicit sequence learning is the serial reaction time (SRT) task. In the standard task, on each trial, a target appears in one of four boxes on a computer screen and participants respond as quickly as possible to the target by pressing a corresponding key on the keyboard. Participants typically perform numerous blocks of 100 trials. Unknown to them, the targets appear in a pre-determined sequence of locations, commonly about 12 locations in length (e.g., 3-4-2-3-1-2-1-4-3-2-4-1). Reaction times (RTs) are measured for each block and usually in a transfer block, a different and pseudo-random sequence is introduced. If RTs are greater for this transfer block than for the preceding training blocks, then it can be inferred that participants learned the trained sequence (Destrebecqz and Cleeremans, 2001, Destrebecqz and Cleeremans, 2003, Nissen and Bullemer, 1987, Perruchet and Amorim, 1992, Reber and Squire, 1994, Reber and Squire, 1998, Reed and Johnson, 1994, Shanks and Channon, 2002, Shanks and Johnstone, 1998, Shanks and Johnstone, 1999, Shanks et al., 2003, Shanks et al., 2006, Wilkinson and Shanks, 2004). It is often concluded that knowledge acquired during the SRT is implicit or unconscious. For instance, amnesic patients perform normally on the SRT task, although declarative memory for the training episode is impaired (Nissen and Bullemer, 1987, Reber and Squire, 1994, Reber and Squire, 1998). In contrast, PD patients are impaired on both motor (Brown et al., 2003, Doyon et al., 1997, Doyon et al., 1998, Jackson et al., 1995, Shin and Ivry, 2003, Sommer et al., 1999, Kelly et al., 2004, Werheid et al., 2003) and verbal (Smith et al., 2001, Westwater et al., 1998) versions of the SRT task. Although, in some cases learning was attenuated rather than entirely absent in PD patients (Ferraro et al., 1993, Pascual-Leone et al., 1993) and in one study the experimenters did not observe an impairment of learning on the verbal version of the SRT in PD (Smith et al., 2001). Furthermore, Helmuth et al. (2000) reported that while PD patients were impaired at learning sequences of stimulus–motor responses, they were unimpaired at learning a sequence of spatial locations. Individuals with Huntington's disease (HD) also show impairments on the SRT in some studies (Knopman and Nissen, 1991, Willingham and Koroshetz, 1993) but preserved performance in another study (Brown et al., 2001).

Differences in sample characteristics (stage of illness and disease severity, medication state of the patients) and key features of the SRT (length and structure of the sequence, duration of the response–stimulus interval, development of explicit knowledge) are some of the factors that are likely to determine the magnitude of implicit learning deficits on the SRT in PD and HD (Kelly et al., 2004) and may explain the inconsistent pattern of findings described above.

In the standard SRT paradigm, target location repeatedly follows a second order conditional (SOC) 12-item sequence in a fixed or deterministic fashion during the sequence blocks (e.g., SOC 1: 3-4-2-3-1-2-1-4-3-2-4-1). Some have employed a dual task methodology such as requiring participants to count tones while performing the SRT task, which was believed to reduce the availability of explicit resources and encourage the development of implicit learning (e.g., Grafton et al., 1995, Kelly et al., 2004). However, this method is particularly taxing for PD patients (Brown et al., 1993, Brown and Marsden, 1991). Others have added an element of noise to the sequence blocks so that at any point in the training phase the probability that the target appears predictably is 0.85 whereas there is a 0.15 probability that the target will be unpredictable (Shanks et al., 2003, Shanks et al., 2006, Wilkinson and Shanks, 2004). To date, no studies of SRT in PD have employed probabilistic sequence learning therefore, in the present study, a probabilistic sequence generation method for SOC sequences was adopted. Using probabilistic rather than fixed deterministic sequences in the SRT has several advantages: (i) it minimizes the likelihood of explicit sequence learning; (ii) a probabilistic sequence can be presented under single task conditions and therefore, is preferable to other methods of fostering implicit knowledge such as the dual-task methodology because it is less likely that PD patients will find it taxing; and (iii) an on-line measure of sequence knowledge is available by comparing RTs to predictable (probable) and unpredictable (improbable) targets.

There has been much debate regarding whether or not the learning acquired during the SRT is genuinely implicit in nature (Shanks and St. John, 1994). In general, healthy participants and patients with PD find the knowledge they acquire during the SRT difficult to verbalize (Brown et al., 2003, Helmuth et al., 2000; Nissen and Bullemer, 1987, Smith et al., 2001, Willingham et al., 1993). However, healthy participants perform well on more sensitive tests of awareness like free generation and recognition (Doyon et al., 1997, Doyon et al., 1998, Helmuth et al., 2000, Pascual-Leone et al., 1993, Perruchet and Amorim, 1992, Shanks and Johnstone, 1998, Shanks and Johnstone, 1999, Shanks et al., 2003, Sommer et al., 1999, Wilkinson and Shanks, 2004). However, it is possible that implicit processes contaminate performance on these latter tests (Destrebecqz and Cleeremans, 2001, Destrebecqz and Cleeremans, 2003, Merikle, 1992, Merikle et al., 2001). To determine whether performance on explicit tests are contaminated by implicit knowledge, two main methodologies have recently been employed.

Jacoby (1991) and Jacoby et al., 1989, Jacoby et al., 1993 argue that the expression of conscious knowledge is controllable. He therefore devised the ‘process dissociation procedure’ (PDP) to separate uncontrollable (implicit) and controllable (explicit) processes in task performance. The PDP has been applied to previous SRT studies of healthy participants (e.g., Destrebecqz and Cleeremans, 2001, Destrebecqz and Cleeremans, 2003, Wilkinson and Shanks, 2004). In Wilkinson and Shanks' (2004) experiment, participants observed short sequences of targets taken from the training sequence to which they must first respond to, just as they had done in the previous blocks of SRT trials. Following their response to the short sequence, participants were instructed to produce a single generation response under either: inclusion instructions, which required them to recall and generate the next item in the sequence. Or exclusion instructions, where they had to produce a key press that did not overlap with the training sequence instructions. Wilkinson and Shanks' (2004) observed that normal participants were able to control the expression of acquired sequence knowledge.

In the present study of SRT in PD, the trial-by-trial test initially devised by Wilkinson and Shanks was included to determine whether generation performance is under intentional control (and therefore explicit in nature), or alternatively, whether it is outside conscious control and can be described as implicit. In both PD and control groups, a within-participant design was employed and participants performed the generation task under both inclusion (I) and exclusion (E) instructions. The critical predictions are that I = E > B, which would imply that knowledge is wholly implicit (i.e., outside intentional control), and I > B > E, which would imply that knowledge is wholly explicit, where I and E refer to measures of the extent to which sequences generated under inclusion and exclusion conditions overlap with the training sequence, and B refers to an appropriate baseline level of generation. Therefore, positive evidence of implicit knowledge would come from either the null effect I = E or from the finding that E > B.

A closer look at recognition performance suggests that in some cases there are at least two processes involved. In this test, participants are presented with 24 test sequences of each composed of six targets to which they must respond, just as they did in the previous blocks of SRT trials. Half of the test sequences come from the training sequence and are therefore called ‘old’ and the remaining sequences are pseudo-random and therefore called ‘new’. In addition, participants must give a recognition judgment to each sequence. In previous studies of normal participants, old chunks were not only more likely to be judged as ‘old’ but were also executed significantly more quickly than new sequences (Shanks and Perruchet, 2002, Shanks et al., 2003). Therefore, it is possible to examine two processes concurrently during the recognition test: recognition ratings and an effect of fluency revealed by an RT difference between old and new chunks. The latter effect will be referred to as a priming effect. In contrast to the view that the recognition test exclusively measures conscious knowledge, it is possible that recognition judgments are not a pure reflection of awareness because they are in part based on the above priming effect rather than on actual old/new item discrimination. For instance, perhaps participants execute old sequences more rapidly via the unconscious retrieval of a motor program and then they infer that the sequence is old because it was executed rapidly. In all previous SRT studies in PD patients, RT performance was measured during the training stage whereas awareness was measured subsequently during the test phase. This design is problematic because when RT performance is observed in the absence of awareness, it is difficult to determine whether the participants performed at chance on the awareness test because they were unaware of the sequence or because the intervening interval between performance and the test stage led to the decay of their memory of the sequence, before the awareness test was undertaken. In contrast, the examination of the possible influence of motor fluency on recognition performance allows the experimenter to measure performance and awareness concurrently. If it is demonstrated that participants showed a priming effect combined with poor recognition performance, then the possibility of the detrimental effect of an interfering interval between the training and test phase is eliminated and the finding would constitute good evidence of implicit learning. To date, such a pattern of results has not been observed, and the concurrent recognition and priming method has not yet been employed to examine awareness in patients with PD.

The study had two objectives. First, we aimed to determine whether probabilistic sequence learning is impaired in PD or not. Patients with PD and age and IQ matched controls were assessed on 15 blocks of the probabilistic SRT task. We predicted that if the contribution of the striatum to implicit learning on the probabilistic SRT is essential, learning on this task in PD patients would be significantly reduced or absent compared to controls.

Given our prediction that sequence learning would be reduced or absent in PD, our second aim was to assess whether intact knowledge acquired by controls during the probabilistic SRT and any residual knowledge, which may be acquired by PD patients, is genuinely implicit or not. Therefore, we tested participants' awareness of their acquired knowledge with two tests of awareness. We predicted that if learning on the probabilistic SRT is implicit, then (a) the expression of participants' acquired knowledge would be outside intentional control (PDP method), and (b) recognition performance would be at chance and during the recognition test, priming and recognition would be dissociable.

Section snippets

Results

In this and all subsequent analyses: (i) RTs or errors for participants trained on one of the two possible sequences were combined. (ii) RTs or errors to the first two targets of each block were excluded because their locations cannot be predicted. (iii) If there was a violation of the sphericity assumption, Pillai's multivariate test of significance was employed (V). Thus, if the Greenhouse–Geisser was less than 1.0, Pillai's exact F is reported. (iv) A significance criterion of á = 0.05 is

Probabilistic implicit sequence learning in PD

In the present study both healthy controls and patients with PD demonstrated evidence of learning a probable sequence, as shown by the RT difference between probable and improbable trials. However, the magnitude of learning was significantly reduced in the patient group. We therefore conclude that PD patients show evidence of sequence learning, albeit at a significantly attenuated level relative to matched controls. Our findings are consistent with previous research indicating that

Participants

Fourteen individuals with a diagnosis of idiopathic PD (9 male, 5 female) aged between 45 and 73 (M = 62.79, SD = 8.07) were recruited. The patients were recruited from the movement disorders clinic at the National Hospital for Neurology and Neurosurgery. All patients met PDS Brain Bank diagnostic criteria for PD (Hughes et al., 1992). Informed consent was obtained prior to participation in the study. The study was approved by The National Hospital for Neurology and Neurosurgery and Institute of

References (94)

  • G.M. Jackson et al.

    Serial reaction time learning and Parkinson's disease: evidence for a procedural learning defect

    Neuropsychology

    (1995)
  • L.L. Jacoby

    A process dissociation framework: separating automatic from intentional uses of memory

    J. Mem. Lang.

    (1991)
  • S. Kelly et al.

    Learning of hybrid and ambiguous sequences by patients with Parkinson's disease

    Neuropsychologia

    (2004)
  • D.S. Knopman et al.

    Procedural learning is impaired in Huntington's disease—evidence from the serial reaction–time–task

    Neuropsychologia

    (1991)
  • P.M. Merikle et al.

    Perception without awareness: perspectives from cognitive psychology

    Cognition

    (2001)
  • M.J. Nissen et al.

    Attentional requirements of learning: evidence from performance measures

    Cogn. Psychol.

    (1987)
  • D. Shohamy et al.

    The role of dopamine in cognitive sequence learning: evidence from Parkinson's disease

    Behav. Brain Res.

    (2005)
  • J. Smith et al.

    Preserved implicit learning on both the serial reaction time task and artificial grammar in patients with Parkinson's disease

    Brain Cogn.

    (2001)
  • K. Werheid et al.

    Sequence learning in Parkinson's disease: the effect of spatial stimulus–response compatibility

    Brain Cogn.

    (2003)
  • A.R. Aron et al.

    Human midbrain sensitivity to cognitive feedback and uncertainty during classification learning

    J. Neurophysiol.

    (2004)
  • Beck, A.T., Steer, R.A., Brown, G.K., 1996. Manual for the Beck Depression Inventory. Psychological Corporation, San...
  • D.G. Beiser et al.

    Model of cortical-basal ganglionic processing: encoding the serial order of sensory events

    J. Neurophysiol.

    (1998)
  • D.J. Brooks

    Functional imaging studies on dopamine and motor control

    J. Neural Trans.

    (2001)
  • R.G. Brown et al.

    Dual task-performance and processing resources in normal subjects and patients with Parkinson's-disease

    Brain

    (1991)
  • R.G. Brown et al.

    The execution of bimanual movements in patients with Parkinson, Huntington and cerebellar disease

    J. Neurol., Neurosurg. Psychiatry

    (1993)
  • R.G. Brown et al.

    Dissociation between intentional and incidental sequence learning in Huntington's disease

    Brain

    (2001)
  • R.G. Brown et al.

    Pallidotomy and incidental sequence learning in Parkinson's disease

    NeuroReport

    (2003)
  • M. Carbon et al.

    Learning networks in health and Parkinson's disease: reproducibility and treatment effects

    Hum. Brain Mapp.

    (2003)
  • N.J. Cohen et al.

    Preserved learning and retention of pattern analysing skill in amnesia: dissociation of knowing how and knowing that

    Science

    (1980)
  • R. Cools et al.

    Enhanced or impaired cognitive function in Parkinson's disease as a function of dopaminergic medication and task demands

    Cereb. Cortex

    (2001)
  • T. Curran

    Effects of ageing on implicit sequence learning: accounting for sequence structure and explicit knowledge

    Psychol. Res.

    (1997)
  • A. Destrebecqz et al.

    Can sequence learning be implicit? New evidence with the process dissociation procedure

    Psychon. Bull. Rev.

    (2001)
  • A. Destrebecqz et al.

    Temporal effects in sequence learning

  • J. Doyon et al.

    Functional anatomy of visuomotor skill learning in human participants examined with positron emission tomography

    Eur. J. Neurosci.

    (1996)
  • J. Fernandez-Ruiz et al.

    Visual habit formation in monkeys with neurotoxic lesions of the ventrocaudal neostriatum

    Proc. Natl. Acad. Sci. U. S. A.

    (2001)
  • A.P. Field

    Discovering Statistics Using SPSS

    (2005)
  • J.V. Filoteo et al.

    Probabilistic category learning in patients with amnesia, Huntington's disease, or Parkinson's disease: the role of the hippocampus and basal ganglia

    J. Cogn.

    (1998)
  • M.J. Frank et al.

    By carrot or by stick: cognitive reinforcement learning in parkinsonism

    Science

    (2004)
  • M.J. Frank

    Dynamic dopamine modulation in the basal ganglia: a neurocomputational account of cognitive deficits in medicated and nonmedicated Parkinsonism

    J. Cogn. Neurosci.

    (2005)
  • A.M. Gotham et al.

    ‘Frontal’ cognitive function in patients with Parkinson's disease ‘on’ and ‘off’ levodopa

    Brain

    (1988)
  • S.T. Grafton et al.

    Functional mapping of sequence learning in normal humans

    J. Cogn. Neurosci.

    (1995)
  • E. Hazeltine et al.

    Attention and stimulus characteristics determine the locus of motor-sequence encoding—a PET study

    Brain

    (1997)
  • M.M. Hoehn et al.

    Parkinsonism: onset, progression and mortality

    Neurology

    (1967)
  • M. Honda et al.

    Dynamic cortical involvement in implicit and explicit motor sequence learning: a PET study

    Brain

    (1998)
  • A.J. Hughes et al.

    Accuracy of clinical diagnosis of idiopathic Parkinson's disease: a clinico-pathological study of 100 cases

    J. Neurol. Neurosurg. Psychiat.

    (1992)
  • L.L. Jacoby et al.

    Separating conscious and unconscious influences of memory: measuring recollection

    J. Exp. Psychol. Gen.

    (1993)
  • L.L. Jacoby et al.

    Becoming famous without being recognized: unconscious influences of memory produced by dividing attention

    J. Exp. Psychol. Gen.

    (1989)
  • Cited by (60)

    • Effects of practice and delays on learning and retention of skilled tool use in Parkinson's disease

      2017, Neuropsychologia
      Citation Excerpt :

      In contrast, if executive dysfunction was responsible for the deterioration of motor skill performance after a 3-week delay, as reported by Roy et al. (2015), then one might expect there would be little or no decline in performance after shorter delays because less prefrontal cortical activation and executive function would be required (Zhao et al., 2015). Second, participants were give more practice to investigate whether individuals with PD would still demonstrate some overall learning across sessions as a result of more extensive training, despite forgetting between sessions (Shohamy et al., 2004, 2008; Wilkinson and Jahanshahi, 2007; Wilkinson et al., 2009). Declarative and procedural memory systems interact during motor sequence learning (Albouy et al., 2013), and it is possible that more practice may result in greater use of procedural memory.

    • Levodopa medication improves incidental sequence learning in Parkinson's disease

      2016, Neuropsychologia
      Citation Excerpt :

      However, it is not possible to differentiate, on the basis of these studies, the degree to which impairments were a consequence of the disease or its medical treatment with dopamine replacement therapy. Only two studies have specifically reported data on SRTT learning in PD patients who were either not taking levodopa (Muslimovic et al., 2007) or tested off-medication after a washout period (Wilkinson and Jahanshahi, 2007). In the first study, Muslimovic and colleagues (2007) assessed learning in a large sample of PD patients (n=95) performing a 10 item SRTT.

    • Probing implicit learning in obsessive-compulsive disorder: Moderating role of medication on the weather prediction task

      2016, Journal of Obsessive-Compulsive and Related Disorders
      Citation Excerpt :

      For this reason, the WPT has been used extensively to examine the neurocircuitry to support implicit learning (Price, 2009). Functional neuroimaging data support dependence of the WPT on intact striatal function (Wilkinson & Jahanshahi, 2007; Wilkinson, Khan, & Jahanshahi, 2009). The striatum (right caudate) is activated during WPT task completion, as are cortical areas that project to it (Aron, Gluck, & Poldrack, 2006; Poldrack, Prabhakaran, Seger, & Gabrieli, 1999; Seger & Cincotta, 2005).

    View all citing articles on Scopus
    View full text