Overview of the evidence for implicit learning following stroke

Controlled trials investigating the ability to engage in implicit learning have evaluated whether stroke survivors have been as effective at learning an implicit motor task as a control group of healthy volunteers. Studies have shown that whilst learning did occur, stroke survivors were unable to learn as well as healthy volunteers, and that this was statistically significant (Winstein et al., (1999), P<0.001; Pohl et al. (1999), P<0.002; Boyd and Winstein (2004), P=0.003; Boyd et al. (2007), P<0.002); Vidoni and Boyd (2009), P=0.005). In these trials participants are commonly given a motor learning task such as a serial reaction time task, in which they are asked to respond to a series of lights by pressing a corresponding button when each light is lit. The sequence in which the series of lights are illuminated is repeated and can become implicitly learnt by the participant. Learning is demonstrated by a reduction in the time taken to complete the test sequences before and after the experiment (e.g.

Boyd et al., 2007). Using a similar methodology, other studies have evaluated the accuracy of reaching the target rather than the time taken (e.g. Pohl and Winstein, 1999), thus learning is demonstrated by an improvement in the accuracy of reaching the target. Arguably these aspects are only one

characteristic of movement performance and these studies have not evaluated elements such as stability and adaptability of the learnt task (Magill, 2006).

Generalisability of the findings from the studies referred to above is limited by small sample sizes (Altman, 1990). Meta-analysis of the findings from these studies would be able to provide an effect size and estimate of statistical significance, this form of summary analysis might address this limitation (Borenstein et al., 2010). Participants included in these studies were all able to follow the instructions for completion of the task inherent within the study design (Boyd et al., 2007, Boyd and Winstein, 2004a, Gomez-Beldarrain et al., 1998, Orrell et al., 2007, Pohl and Winstein, 1999, Shin et al., 2005, Vidoni and Boyd, 2009, Winstein et al., 1999). In order to do this, it implies that the participants had little or no cognitive impairment. 38% of stroke survivors have some cognitive impairment three months after stroke (Patel et al., 2002) and this is likely to impact on implicit learning (Shumway-Cook and Woollacott, 2007).

Studies evaluating the capacity for implicit motor learning in stroke survivors with cognitive impairments are therefore also required.

Another limitation to these studies arises because of the decision to test implicit motor learning through reaction time tests or accuracy tests with the ipsilesional or unaffected arm. (Boyd et al., 2007, Boyd and Winstein, 2004a, Gomez-Beldarrain et al., 1998, Orrell et al., 2007, Pohl and Winstein, 1999, Shin et al., 2005, Vidoni and Boyd, 2009, Winstein et al., 1999). Following stroke the contralesional or affected arm may display movement impairments that could lead to a difficulty in carrying out a serial reaction time task. These impairments may lead to slower times on a reaction time test or inaccuracies in reaching a target. Assumptions may subsequently be made about the participant’s ability to learn, when the results may have occurred because of movement difficulties experienced as a consequence of the stroke. The decision to use the ipsilesional arm during both practice and outcome assessment avoids this confounder. Whilst this would seem appropriate as a means of evaluating the behavioural ability to learn a motor task, it is not appropriate as a means of investigating CNS activity during the learning of that task by a stroke survivor.

Movement in each side of the body is predominantly controlled by the

contralateral structures of the CNS. Therefore neural activity within the damaged hemisphere, occurring in response to motor learning of the impaired limb, may manifest itself differently to learning involving the unaffected limb.

There is little published work identifying CNS activity before and after an implicit motor learning task using the impaired limb following stroke (Hosp and Luft, 2011). A small study by Meehan et al. (2001) attempted to identify the underlying CNS activity in response to a learning task in a small group of participants diagnosed with stroke (n=9) compared to that of healthy volunteers (n=9). Functional Magnetic Resonance Imaging (fMRI) was conducted during early practice of the task (day 1) (baseline), practice took place over days 2-6 and then a second fMRI was taken at day 7 (retention).

Behavioural tests indicated that both groups were able to learn the tracking task used in the study and fMRI revealed that during the early phase of learning both groups showed increased activity in the fronto-parietal regions as expected (notably dorsolateral prefrontal cortex (DLPFC) and premotor cortex).

Activity in the DLPFC subsequently reduced during the second imaging session (day 7 – retention) for the group of healthy volunteers, demonstrating the need for less attentional resources as the activity became more familiar (learnt). This shift away from activity in the DLPFC was not, however apparent when the participants diagnosed with stroke were scanned at retention. In contrast to the group of healthy volunteers, Meehan et al. (2011) found increased activity remained within both the dorsolateral pre frontal area and the pre motor cortex in the group of stroke survivors. This suggests that the participants diagnosed with a stroke required continued attention to the motor tasks even after behavioural assessment would suggest that the task had been learnt. This was a small study which included stroke survivors who were at least twelve months post onset, and therefore, the findings cannot be generalised to all individuals diagnosed with stroke. More research is required, but this study does suggest that there are differences in the physiological response to learning in stroke survivors that may not be apparent in healthy volunteers. The continued assumption that physiological changes underpinning learning after stroke are the same as those that occur during learning in healthy participants

may be inaccurate (Hosp and Luft, 2011). The potential for different physiological responses, together with the findings from the previously discussed behavioural studies of implicit learning, support the theory that motor learning principles may not be able to be applied to movement rehabilitation of