Sequence learning in Parkinson’s disease: The effect of spatial stimulus–response compatibility

Abstract

Patients with Parkinson’s disease (PD) have repeatedly demonstrated reduced sequence-specific learning effects in serial reaction time tasks (SRTs). Previous research with PD patients has mainly employed the ‘classical’ SRT task, involving a spatially compatible assignment of stimuli and responses. From cognitive research, it is known that spatial compatibility triggers rapid, automatic responses in the direction of the stimulus. Automatic responding has shown to be disinhibited in PD patients and may therefore interfere with stimulus anticipation during the learning process. The aim of the present study was to test this hypothesis by investigating if reduced sequence-specific learning depends on spatial stimulus–response compatibility. PD patients and age-matched controls were examined either with an SRT variant involving central stimulus presentation, thereby preventing automatic linking of stimulus and response locations, or with a spatially compatible SRT task. Patients showed reduced sequence-specific learning effects only when the stimulus–response assignment was spatially compatible. This pattern of results confirms the hypothesis that sequence learning deficits in PD may result from a predominance of automatic response activation over learning-based stimulus anticipations during the learning phase.

Introduction

During the past 20 years, much research effort has been devoted to the investigation of the cognitive alterations in Parkinson’s disease (PD) and their relationship to the disease’s characteristic motor symptoms. A main topic of cognitive research in PD has been procedural learning, the acquisition of a motor or nonmotor skill by practice, as this type of learning has been supposed to depend on intact basal ganglia function (Mishkin, Malamut, & Bachevalier, 1986; Squire, 1987). Although a general deficit of procedural learning was expected, empirical evidence has been quite heterogeneous. While initial empirical evidence seemed to support this hypothesis (e.g., Frith, Bloxham, & Carpenter, 1986; Wallesch et al., 1990), a considerable number of studies did not show any impairments in PD patients as compared to age-matched control groups (e.g., Allain et al., 1995; Bondi & Kaszniak, 1991; Heindel, Salmon, Shults, Walicke, & Butters, 1989; Huberman, Moscovitch, & Freedman, 1994; Koenig, Thomas-Anterion, & Laurent, 1999; Morris et al., 1988; Peigneux, Meulemans, Van der Linden, Salmon, & Petit, 1999; Reber & Squire, 1999; Smith, Siegert, & McDowall, 2001; Vakil & Herishanu-Naaman, 1998). The heterogeneity of previous research suggests that there are different forms of implicit learning (cf. Seger, 1997), some of which may be disturbed in PD, while others remain intact (Haaland, Harrington, O’Brian, & Hermanowicz, 1997; Soliveri, Brown, Jahanshahi, & Marsden, 1992b). Thus, Reber and Squire (1999) raised the hypothesis that deficits in PD may be restricted to procedural learning tasks that provide motor learning indices.

The serial reaction time task (SRT; Nissen & Bullemer, 1987) is considered to be a standard tool for the assessment of procedural learning by means of motor learning indices. In this task, visual stimuli appear successively at different screen locations, and a response is given by pressing spatially corresponding keys. Unbeknown to the participants, a complex, continuously cycling sequence of stimulus locations is presented. During several blocks of training of the sequence, reaction times (RTs) usually decrease. When switching to a random sequence thereafter, RTs increase again. Both the RT decrease during the training phase and the RT increase in random blocks can occur without or with only fragmentary explicit knowledge about the sequence.

The SRT involves two different cognitive functions which are characteristic for implicit learning tasks. First, it entails learning of the associations between the stimuli and motor responses, in the following named ‘visuomotor learning.’ Secondly, it involves learning of a specific sequential pattern, which will be termed ‘sequence-specific learning.’ In SRTs, sequence-specific learning causes a decrement of reaction times (RTs) beyond visuomotor learning (Nissen & Bullemer, 1987). Most SRTs provide two learning indices which allow to differentiate between visuomotor learning and sequence-specific learning. The decrement of reaction times (RTs), which usually occurs during initial training of the sequence, is due to both visuomotor learning and sequence-specific learning. The RT increase typically observed when switching to the random sequence, purely reflects sequence-specific learning. Thus, the difference between RTs in random blocks and preceding sequence blocks indicates whether the specific regularities of the sequence have previously been learned during the training phase.

Patients with Parkinson’s disease have repeatedly been investigated in behavioral studies employing the SRT paradigm (Doyon et al., 1997; Ferraro, Balota, & Connor, 1993; Jackson, Jackson, Harrison, Henderson, & Kennard, 1995; Pascual-Leone et al., 1993; Smith et al., 2001; Sommer, Grafman, Clark, & Hallet, 1999; Westwater, McDowall, Siegert, Mossman, & Abernethy, 1998). Some, but not all of these studies found reduced sequence-specific learning, as indicated by smaller RT increases in the random block (Ferraro et al., 1993; Jackson et al., 1995; Westwater et al., 1998; but see also Smith et al., 2001). Others found a smaller RT decrease during the sequence blocks in the absence of reduced sequence-specific learning (Pascual-Leone et al., 1993; Sommer et al., 1999). This pattern suggests that visuomotor learning, but not sequence-specific learning is disturbed. Nevertheless, the specific nature of the underlying deficit still remains unclear.

As a possible explanation for the deficits PD patients exhibit not only in SRTs, but in a variety of other cognitive tasks, the hypothesis has been raised that PD causes a failure to suppress automatic response tendencies (Jackson & Houghton, 1995). This hypothesis converges with everyday observations of ‘stimulus-bound’ behavior in PD patients. The phenomenon of ‘paradoxic kinesis,’ that PD patients with severe motor disabilities are able to get up and run quickly in situations of emergency, impressively shows that they can more readily initiate movements in response to external stimuli than voluntarily. The finding that visual markers can be therapeutically useful to increase stride length in PD patients with gait hypokinesia (Martin, 1967; Morris, Iansek, Matyas, & Summers, 1994) may serve as another example for the predominance of external over internal cues in movement initiation.

The observation of ‘overshooting’ automatic responses triggered by external visual cues points to the view that PD may cause a failure to inhibit automatic attentional processes. Consequently, attention is easily ‘catched’ by external stimuli, whereas self-initiated disengagement and re-engagement of attention is disturbed. Empirical evidence for this assumption comes from studies employing different cognitive tasks such as negative priming (Filoteo, Rilling, & Strayer, 2002), the Posner, Snyder, and Davidson (1980) paradigm (Wright, Geffen, & Geffen, 1993), and externally versus self-cued reaction times (Siegert, Harper, Cameron, & Abernethy, 2002). These studies showed that PD patients are not only unimpaired, but sometimes even outperform healthy subjects when the attentional focus is guided by external stimuli. Taken together, there is ample evidence for a “hyperreflexive” orientation of attention in PD patients.

From research in healthy participants, it is known that attentional shifts and, consequently, automatic responses are easily triggered by spatially compatible stimuli (DeJong, Liang, & Lauber, 1994; Kornblum, Hasbroucq, & Osman, 1990). A standard paradigm for measuring the effects of spatial compatibility is the Simon task (Simon, 1969). In this choice-reaction task, subjects typically respond faster when the relative spatial positions of stimulus and response match (S–R compatibility), compared to when the positions do not match, even if the spatial position is irrelevant for responding. In PD patients, it could recently be demonstrated that a disinhibition of automatic responses occurs in PD patients performing the Simon task (Praamstra & Plat, 2001). The RT gain in compatible trials, called the “Simon effect,” was larger in PD patients than in controls, and accompanied by an enhancement of the attention-related electrophysiological N200 component.

Regarding the SRT paradigm, nearly all previous studies with PD patients employed the ‘classical’ SRT task by Nissen and Bullemer (1987), which involves spatially compatible stimuli and responses. As an exception, Smith et al. (2001) employed a task variant, in which a spatial stimulus sequence was combined with verbal instead of manual responses, in order to exclude possible effects of the patients’ motor deficits. PD patients showed normal sequence learning compared to healthy controls. The second exception is a study by Helmuth, Mayr, and Daum (2000), who employed a spatially incompatible assignment of stimuli and responses to examine whether spatial and object sequences can be learned independently from each other. Participants had to respond to a sequence of different objects (numbers 1–4) appearing at different locations, with the object sequence being response-relevant. Earlier, Mayr (1996) had demonstrated that young, healthy subjects can learn spatial and object sequences independently from each other. The results of Helmuth et al. (2000) showed that elder controls learned the response-relevant object sequence. However, as healthy participants did not show any learning of the response-irrelevant spatial sequence, it could not be ruled out that a failure to learn the spatial and object sequences simultaneously may simply result from the fact that elder participants focused on the response-relevant object sequence. A further problem was that patients’ blockwise RT means varied considerably throughout the task, which made it difficult to interpret the RT difference between random and sequence blocks. Thus, the mere effect of spatial compatibility could not be derived from the findings of this study, and might be examined more conclusively by a between-variation of spatially compatible and incompatible stimulus sequences.

Taken together, in most studies sequence learning was reduced in PD patients performing the ‘classical,’ spatially compatible SRT task, but the specific nature of the underlying deficit is yet unclear. Recent empirical evidence points to the view that a failure to inhibit automatic response tendencies, as triggered by spatial S–R compatibility, might account for patients’ deficits in a variety of cognitive tasks. However, the impact of S–R compatibility on PD patients’ performance in SRT tasks has not been directly addressed yet. The present study intended to focus on this aspect by investigating PD patients’ performance: (a) in a task variant involving central stimulus presentation, thereby preventing automatic linking of stimulus and response locations and (b) in a spatially compatible SRT.